WO1998005165A1

WO1998005165A1 - Method and device for detecting moving body

Info

Publication number: WO1998005165A1
Application number: PCT/JP1997/002590
Authority: WO
Inventors: Hideyuki Matsumoto; Akinori Kooketsu; Akio Yoshino; Shunnichi Komaki; Masakazu Saitoh
Original assignee: Usc Co., Ltd.
Priority date: 1996-07-26
Filing date: 1997-07-25
Publication date: 1998-02-05

Abstract

A method and device by which a moving body can be easily and surely detected without increasing the manufacturing cost. In the method, input signals are alternately quantized by performing first and second quantization. The first quantization is performed by setting input signals having luminance equal to or larger than a prescribed value at a high level and those having luminance of a value smaller than the prescribed value at a low level. The second quantization is performed by setting input signals having luminance equal to or larger than a prescribed value at a low level and those having luminance of a value smaller than the prescribed value at a high level. A moving body is extracted by superimposing a picture frame (1) subjected to the first quantization upon a picture frame (2) successively subjected to the second quantization after the first quantization so as to make the picture elements of the frames (1 and 2) having the same luminance compensate each other.

Description

TECHNICAL FIELD The present invention relates to a moving object detection method and a moving object detection device that extract only a moving object from a series of moving images, and can be used as, for example, an intruder monitoring device. Furthermore, the present invention relates to a moving object detection method and a moving object detection device for the purpose of improving accuracy and speed, improving robustness, and the like of processing relating to detection of various kinds of moving objects. Technical background When detecting a moving object from a moving image, in the past, a person continuously looked at the moving image projected on the display, and the moving object was detected based on his judgment. For example, in order to detect or record a trespasser, a monitoring device using a monitoring camera has been conventionally used. Such a monitoring device projects an image in a predetermined area photographed by a surveillance camera, and a surveillant who monitors this image determines whether or not an illegal intruder exists. Such monitoring work requires labor and may overlook the presence of an intruder, depending on the degree of fatigue of the monitor. In other words, the task of detecting moving objects was cumbersome, sometimes overlooked, and used for certainty.

SUMMARY OF THE INVENTION The present invention has been made in view of the above situation, and a purpose thereof is to provide a moving object detection method and a moving object capable of easily and reliably detecting a moving object such as an unauthorized intruder without requiring labor. It is intended to provide a low-cost detection device. Further, it is another object of the present invention to improve the reliability of moving object detection, for example, by improving the robustness in detecting the moving object as described above. Disclosure of the invention Among the moving object detection method and the moving object detection device according to the present invention, the invention according to the moving object detection method according to claim 1 is characterized in that the quantization direction of the input frame obtained by the imaging unit is alternately inverted. A feature is to extract a moving body portion by alternately outputting a positive frame and a negative frame and by equalizing the density value of a stationary portion common to the positive frame and the negative frame adjacent in the forward direction.

Further, in the invention described in claim 1 described above, it may be configured to perform contrast correction. That is, as described in claim 2, for the first frame on the time axis of the input frame group consisting of a desired number of input frames, the brightness is set to a predetermined value for the first frame on the time axis when the brightness is higher than the desired value. Re-quantize by applying contrast correction for correction, and re-quantize the remaining frames of the input frame group using the contrast correction information obtained in the first frame. Configuration. Alternatively, as described in claim 3, with respect to the first frame on the time axis of the input frame group including the desired number of input frames, the brightness of the frame whose brightness is equal to or more than the desired value is determined. Re-quantization is performed by performing contrast correction for correcting to a predetermined value, and for a frame following this frame, re-quantization is performed using the contrast correction information in the frame preceding that frame. Can be configured.

Furthermore, in the invention described in claim 1, skip processing can be performed. That is, as described in claim 4, at least one of the input frames is set as a search target frame, and a maximum density value and a minimum intensity value are calculated from all pixels constituting the search target frame. Then, each is set as a first threshold value and a second threshold value, and a difference between the maximum density value and the minimum density value is calculated to obtain a third 闆 value. In addition, the remaining frames other than the search target frame among the input frames are set as template frames, and the maximum density value and the minimum density value are calculated from all the pixels constituting the template frames. Then, it is determined whether the maximum density value and the minimum density value obtained from these template frames exceed the first K value and the second! ¾ value, respectively. If none of the density values exceeds the first and second threshold values, it is determined that no moving object exists in the template frame. On the other hand, when the maximum density value and the minimum degree value obtained from the template frame both exceed the first and second 1 第二 values. Then, the difference between the maximum density value and the minimum density value obtained from the template frame is compared with the third ¾ value, and if the difference is smaller than the third threshold value, the above template is compared. It is determined that no moving object exists in the frame. On the other hand, when the maximum density value and the minimum density value obtained from the template frame both exceed the first and second 閱 values, and the difference is equal to or greater than the third threshold value. Then, it is determined that there is a moving object in the above template frame.

Further, in the invention described in claim 1 described above, the vector connection processing may be performed. That is, as described in claim 5, a change in coordinate data in each input frame of a pixel which is estimated to form a moving object portion among pixels constituting the input frame and a pixel in the vicinity thereof, And the moving direction of the input image and its neighboring image in each input frame are detected, and the object consisting of all the pixels whose change in the coordinate data coincides with the moving direction and the object composed of the neighboring pixels is regarded as one moving object. I reckon.

Furthermore, if all the above-described processes are performed, a more preferable moving body detection method can be obtained. That is, as described in claim 6, for the first frame on the time axis of the input frame group consisting of the desired number of input frames, the brightness is set for the frame whose brightness is equal to or more than the desired value. Performs re-quantization by applying contrast correction to correct a predetermined value, and re-quantizes the remaining frames of the input frame group using the contrast correction information obtained in the first frame. First, the first contrast correction is performed. Next, the following skip processing is performed as one of the so-called pattern matching. That is, at least one of the moving object detection frames obtained by performing the first contrast correction is set as a search target frame, and the maximum density value and the minimum density value are obtained from all the pixels constituting the search target frame. Are calculated as the first 閱 value and the second threshold value, and the difference between the maximum density value and the minimum density value is calculated as the third 閎 value. At the same time, the remaining frames other than the search target frame in the moving object detection frames are used as template frames, and the maximum density value and the minimum density value are calculated from all the pixels constituting the template frame. Whether the maximum density value and the minimum intensity value obtained from the template frame exceed the first and second values, respectively. If the maximum density value and the minimum density value do not exceed the first and second values, it is determined that no moving object exists in the template frame. When the maximum density value and the minimum density value obtained from the template frame both exceed the first and second levels, the maximum density value and the maximum density value obtained from the template frame are used. The difference from the minimum density value is compared with the third ¾ value. If the difference is smaller than the third 闼 value, it is determined that no moving object exists in the template frame. On the other hand, when the maximum density value and the minimum strain value obtained from the template frame both exceed the first and second threshold values, and the difference is equal to or greater than the third value. Then, it is determined that a moving object exists in the template frame.

In the moving object detection frame, a portion where there is no movement with respect to the temporally previous frame has an intermediate degree value (for example, around 128 in an 8-bit grayscale image). Utilizes such characteristics to detect whether or not there are pixels in the pixels composing the template frame that are more prominent than the first and second reference values obtained in advance. In such a detection, first, it is checked whether or not a density value exceeding the first and second 存在 values exists in the pixels of the template frame. If the values do not exceed each value, it is determined that no moving object exists in the template frame.If the values exceed the first and second 閩 values, the IE determines the peak value. That is, the third M value obtained by the frame to be searched Is compared with the difference obtained by the template frame, and if this difference value is smaller than the third 閱 value, it is determined that no moving object exists in the template frame. Since skip processing is performed in two stages, the speed of pattern matching processing can be increased and the mouth-cutting performance can be improved.

At the same time, the following vector connection processing is performed. That is, a change in coordinate data of each of the pixels constituting the moving object detection frame in the moving object detection frame, and a moving direction of each of the pixels in the moving object detection frame are detected. The object composed of all the pixels that almost coincide with the moving direction is regarded as one moving object, and vector connection processing is performed.

In the invention described in claim 6, as described in claim 7, A second contrast correction may be performed instead of the first contrast correction. That is, contrast correction for correcting the luminance of a first frame on the time axis of a group of input frames including a desired number of input frames to a predetermined value for a frame having a luminance higher than a desired value. And performs a second contrast correction for re-quantizing the frame following this frame using the contrast correction information in the frame preceding the frame.

Further, among the moving object detecting method and the moving object detecting device according to the present invention, the moving object detecting device according to claim 8 includes a photoelectric conversion means for converting an optical input into a charge signal and storing the signal. The photoelectric conversion means quantizes the input optical information to a high level when the luminance thereof is equal to or more than a predetermined value and a low level when the luminance is less than a predetermined value; The above-mentioned image frames which have been subjected to the first quantization are alternately subjected to a low level and quantized to a low level below a predetermined value as a high level. By comparing the image frame subjected to the second quantization following the quantization and canceling out the pixels with the same luminance on both sides, it is regarded as a moving object. Function.

As the photoelectric conversion unit, an imaging unit such as a CCD camera can be adopted as described in claim 9, and further, a detection unit for detecting a moving object extracted by the imaging unit can be provided. Further, as described in claim 10, a notifying means which operates when the presence of a moving object is detected by the detection means and notifies the presence of the moving object can be provided. Further, as described in claim 11, it is also possible to provide a recording unit that operates when the presence of a moving object is detected by the detection unit and records the moving object.

Further, in each of the above-mentioned inventions, a contrast correction means may be additionally provided as described in claim 12. That is, of each pixel in a state where the light input is quantized, a contrast correction unit for performing a contrast correction for correcting a luminance of a pixel whose luminance is equal to or more than a desired value to a predetermined value is provided. The contrast correction means performs the contrast correction on the first frame on the time axis of the input frame group composed of a desired number of input frames, requantizes the frame, and re-quantizes the input frame group. The rest of us It has a function to requantize the frame using the contrast correction information obtained in the first frame. Of course, the photoelectric conversion means can freely convert the light input corrected by the contrast correction means into a charge signal. Alternatively, a contrast correction means as described in claim 13 may be provided. That is, the contrast correction means performs the above-described contrast correction on the first frame on the time axis of the input frame group composed of a desired number of input frames, requantizes the frames, and outputs the frames. It has a function to requantize the subsequent frame using contrast correction information in the frame preceding the frame.

Furthermore, in the invention described in claim 8 to claim 11, it may be configured to include first contrast correction means, skip processing means, and vector connection processing means. good. That is, as described in claim 14, a first contrast correction means, a skip processing means, and a vector connection processing means are provided. The first contrast correction means performs the contrast correction on the first frame on the time axis of the input frame group including a desired number of input frames, and requantizes the first frame. It has a function to requantize the remaining frames in the group using the contrast correction information obtained in the first frame. The skip processing means includes, as a search target frame, at least one of the moving object detection frames obtained by performing the contrast correction by the first contrast correction means, The maximum density value and the minimum density value are calculated from all the pixels constituting the search target frame, and are respectively set as the first threshold value and the second ia value. Is calculated as the third threshold value. In addition, the remaining frames other than the search target frame among the above-mentioned motion detection frames are used as template frames, and the maximum density value and the minimum density value are calculated from all the pixels constituting the template frame. I do. Then, it is determined whether or not the maximum density value and the minimum density value obtained from these template frames exceed the first K value and the second それぞれ value, respectively. If the density value does not exceed the first and second 1¾ values, it is determined that no moving object exists in the template frame. On the other hand, when the maximum density value and the minimum density value obtained from the template frame both exceed the first and second minimum values, The difference between the maximum density value and the minimum density value obtained from the template frame is compared with the third threshold value.If the difference is smaller than the third 閱 value, the template frame is compared. Determines that there is no moving object. On the other hand, when the maximum density value and the minimum density value obtained from the template frame both exceed the first and second threshold values and the difference is equal to or greater than the third threshold value, It has a function of determining that a moving object exists in the template frame. In addition, the vector connection processing means includes: a change in coordinate data of a pixel constituting the moving object detection frame in each moving object detection frame; and a moving direction of the pixel in each moving object detection frame. And has a function of detecting an object composed of all the pixels in which the change in the coordinate data and the movement direction almost coincide with each other as one moving object. .

Incidentally, instead of the first contrast correction means of the moving object detection device according to claim 14, a second contrast correction means may be provided as described in claim 15. The second contrast correction means performs the above-described contrast correction on the first frame on the time axis of the input frame group composed of a desired number of input frames, requantizes the frame, and executes the subsequent frames. It has a function to requantize the frame using the contrast correction information in the frame before the frame. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic diagram for explaining a basic principle of a moving object detection method.

FIG. 2 is a flowchart for explaining the basic principle of the moving object detection processing.

FIG. 3 is a diagram showing the charge accumulation amount and the refresh timing. FIG. 4 is a block diagram showing the first embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a moving object detection processing circuit.

FIG. 6 is a block diagram showing a second embodiment of the present invention.

FIG. 7 is a diagram for explaining a third embodiment of the present invention.

FIG. 8 is a diagram for explaining the first contrast correction.

FIG. 9 is a diagram for explaining pattern matching. FIG. 10 is a flowchart for explaining the skip processing.

FIG. 11 is a diagram for explaining the vector connection process.

FIG. 12 is a diagram for explaining an undesired vector connection process. FIG. 13 is a diagram showing a modification of the third embodiment.

BEST MODE FOR CARRYING OUT THE INVENTION First Embodiment

Next, an embodiment of the present invention will be described.

First, the basic configuration and operation of the present invention will be briefly described, and then the first embodiment of the present invention will be described.

By alternately inverting the quantization direction of the input frame obtained by the imaging means, a positive frame 1 and a negative frame 2 are output alternately as shown in FIG. Then, the moving object detection frame 4 is output by equalizing the density value of the stationary portion 3 common to the positive frame 1 and the negative frame 2 which are adjacent in the forward direction. In the moving body detection frame 4, the moving body part 5 emerges because the density value of the stationary part 3 is uniform. Note that “sum / 2” in FIG. 1 indicates a process of subtracting the Macao value of each of the frames 1 and 2. In other words, when the above-mentioned moving object detection frame 4 is output, simply overlaying the positive frame 1 and the negative frame 2 has density information for two screens, so that an image having an extremely high density value is obtained. Become. Therefore, an appropriate amount of the constant value is subtracted.

The principle described above is shown in FIG. As shown in FIG. 2, the imaging means (photoelectric conversion means) converts the input optical information. First, the density value is converted into a charge by positive logic, and then the density is stored by negative logic while storing the charge. Converts a value to a charge.

Then, the surface image frame subjected to the first quantization (converted to electric charges by positive logic) and the second quantization subsequent to the first quantization (subjected to negative logic) By comparing the image frame (which has been converted to electric charge) with the pixel having the same luminance in both sides, pixels having different luminances can be freely extracted in both cases, which are regarded as moving objects. After completing this series of processing, the charge is released and the same processing is repeated. The amount of charge accumulation and the release of charge (refresh) timing in such processing are as follows. Figure 3 shows.

With the configuration described above, of the positive frame 1 and the negative frame 2 obtained by alternately inverting the quantization direction, the adjacent positive frame 1 and negative frame 2 By superimposing and equalizing the density value of the stationary portion common to each of the frames 12, the stationary portion is canceled out, so that the moving body portion 5 can be extracted.

Next, FIG. 4 is a block diagram showing a moving object detection device according to the present embodiment. The device according to the present embodiment is provided with a photoelectric conversion unit 10 that converts an optical input into a charge signal and stores the charge signal, similarly to a conventionally known CCD camera. As shown in FIG. 4, the photoelectric conversion means 10 includes a lens 11 for inputting optical information, an image sensor (CCD) 12, a signal processing circuit 13, a synchronization circuit 14, And a driving unit 15 for driving the element 12. The signal processing circuit 13 includes a CDS 16 which is a circuit for removing noise and extracting only a video signal. Such a configuration is the same as a conventionally known CCD camera.

In particular, in the moving object detection device according to the present embodiment, the signal processing circuit 13 is provided with a moving object detection processing circuit 17 which is a feature of the present invention. As shown in FIG. 5, the moving object detection processing circuit 17 includes an A / D converter 18 that performs A / D conversion of input optical information, and before performing various preprocessing for performing the moving object detection processing. A processing circuit 19, a frame memory 20 for converting the density value into electric charge by positive logic, and storing the image frame converted into the electric charge, and an inversion circuit for converting the density value into electric charge by negative logic. Path 21, a synthesizing circuit 22 for synthesizing the image frames obtained by the frame memory 20 and the inverting circuit 21, respectively, and a signal obtained by the synthesizing circuit 22 is D / A converted and output. And a D / A converter 23. Reference numeral 24 denotes a distributor for transmitting a part of the input signal as a through output signal. The frame memory 20 converts the A / D-converted signal sent through the pre-processing circuit 19 into a high-level signal having a luminance equal to or higher than a predetermined value and a signal having a luminance lower than the predetermined value. Performs the first quantization to quantize (in positive logic) as a verbal level. Then, the image data subjected to the first quantization is stored.

On the other hand, the inversion circuit 21 converts the signal after the A / D conversion into a signal whose luminance is equal to or less than a predetermined value. The upper one is set to the low level, and the one below the predetermined value is set to the high level, and the second quantization is performed (with negative logic). That is, in the frame memory 20, the input signal is converted into a charge value by positive logic into electric charge. The inverting circuit 21 converts an input signal into negative electric charge by negative logic.

Then, the image frame stored in the frame memory 20 and the image frame subjected to the second nesting by the inverting circuit 21 are sent to the synthesizing circuit 22, and the synthesizing circuit 22 The two are synthesized and compared. In the synthesizing circuit 22, the pixels having the same luminance are canceled by the two. In other words, the density of both stationary parts is made uniform. The processing in this case is performed based on the threshold value given from the preprocessing circuit 19.

In the output signal that has been subjected to the D / A conversion by the processing in the synthesizing circuit 22, the output of the stationary portion is made uniform, and the luminance that is considered to be a moving object differs between the two. Are extracted. Therefore, in the output signal, the moving object is in a raised state.

After transmitting the image frame signal to the synthesizing circuit 22, the frame memory 20 and the inverting circuit 21 receive the next signal and perform the same processing. The synthesizing circuit 22 sends an output signal to wait for a new input.

Since the moving object detection device according to the present embodiment is configured as described above, the input optical information is subjected to the first quantization and the second quantization alternately, and these are superimposed. By doing so, the stationary part is canceled and the moving part is raised as a result. Therefore, the moving object detection operation can be performed easily and reliably.

In addition, in the case of the moving object detection device of the present embodiment, the structure can be easily achieved only by adding the moving object detection processing circuit 17 to the signal processing circuit 13 of a conventionally known CCD camera. Such a moving object detection processing circuit 17 can be incorporated into the above-mentioned CCD camera, or it can be manufactured as an external adapter and used by attaching it to an existing CCD camera. Therefore, the manufacturing cost does not increase significantly.

Note that, in the moving object detection device according to the present embodiment, the input image is a gray scale. Although moving objects can be detected even if the input image is a full-color image, extra information is added to the output image because the full-color image can express small color differences and saturation differences. Would. For this reason, a gray scale is preferable as the input image.

When a moving object is detected by the above-described apparatus of the present embodiment, the position, angle, and focal length of the apparatus need to be fixed. This is because if these change, all the shifts due to this change will be extracted. It should be noted that the present embodiment can be implemented using a computer based on data captured by a conventional CCD camera. Second form example

Next, a second embodiment of the present invention will be described. This embodiment is applied as a monitoring device. In the case of the present embodiment, as shown in FIG. 6, a CCD camera 25 for extracting a moving object from an input signal and a detecting means (motion-diode) for detecting the moving object extracted by the CCD camera 25 Detector) 26, an alarm 27 that operates when the detection means 26 detects a moving object such as an unauthorized intruder, and a monitor 28 that displays the input signal one by one. And a timelabs video 29 that also operates as a trigger when the detection means 26 detects a moving object such as an unauthorized intruder. Reference numeral 30 denotes a switcher, which controls the operation and stoppage of the CCD camera.The CCD camera 25 converts the input signal to a high level with a luminance higher than a predetermined value. A first quantization that quantizes a luminance less than a predetermined value as a mouth level and a second quantization that quantizes a luminance higher than the predetermined value as a low level and a luminance lower than the predetermined value as a high level. It has a function of extracting a moving object portion by alternately applying the image signals obtained by the first and second quantizations and canceling out a static portion having no change in luminance. Such a function employs a CCD camera 25a incorporating a circuit for performing such a function, and a conventionally known CCD camera 25b with an adapter 35 described below. May be adopted.

Although FIG. 6 shows both cameras 5a and 5b, it is only necessary to provide one of them. Circuits for performing the above functions are the same as those in FIGS. 4 and 5 showing the first embodiment described above.

The moving object extracted by the CCD camera 25 as described above is detected by the detecting means (motion 'detector) 26. The detecting means 26 detects the moving object by detecting a luminance that is regarded as a moving object or a luminance different from the luminance of the stationary portion. Such a detecting means 26 can be manufactured by using a conventionally known circuit. In addition, an alarm 27 is provided which operates when the detection means 26 detects a moving object such as an illegal intruder or the like as a trigger. In the case of the present embodiment, the alarm 27 is incorporated in the detection means 26.

Further, in the case of the present embodiment, a monitor 28 for displaying the input signals one by one is provided, and similarly to the conventional monitoring device, the moving object can be detected also by the video projected on the monitor 28. Further, when the detecting means 26 detects a moving object such as an illegal intruder, an evening video 29 is provided which operates using this detection signal as a trigger. With this, it is possible to record an image related to the moving object as the moving object is detected. Various other conventionally known recording means can be employed. Since the moving object detection device of this embodiment is configured as described above, the moving object is extracted by the CCD camera 25a. That is, as shown in FIG. 5, the frame memory 20 stores the signal after the A / D conversion transmitted through the pre-processing circuit 19 and the signal whose luminance is equal to or more than a predetermined value. The first quantization is performed in which quantization is performed at a high level and that below a predetermined value is defined as a low level (positive logic). Then, the image data subjected to the first quantization is stored.

On the other hand, the inverting circuit 21 converts the signal after the A / D conversion into a low level when the luminance is equal to or higher than a predetermined value and a high level when the luminance is lower than the predetermined value (by negative logic). And perform the second quantization. That is, in the frame memory 20, the input signal is converted into a density value into a charge by positive logic. The inverting circuit 21 converts the density of the input signal into electric charges by negative logic.

Then, the image frame stored in the frame memory 20 and the image frame subjected to the second quantization by the inverting circuit 21 are sent to the synthesizing circuit 22, and the synthesizing circuit 22 22 Combine the two and compare them. In this synthesis circuit 22, Cancel the pixels having the same luminance. In other words, the density of both stationary parts is made uniform. The processing in this case is performed based on the threshold value given from the preprocessing circuit 19. As a result, moving objects are extracted.

This moving object is detected by the detecting means 26. Therefore, unlike the above-described conventional monitoring device, the monitor does not require effort such as monitoring the video to find the presence of an unauthorized intruder, and does not overlook the presence of a moving object. Further, in the case of the present embodiment, an alarm device 27 is provided as a notification means that operates in conjunction with the detection of the moving object by the detection means 26 described above, so that the person in charge of monitoring the presence of this moving object is provided. Etc. can be notified promptly. In addition, since the timelabs video 9 records an image of the moving object in accordance with the detection of the moving object by the detection means 26, an image of only the moving object such as an illegal intruder can be stored. It is preferable that the position and angle of the camera and the focal length are fixed, as in the case of the structure of the first example described above. Third example

Next, a third embodiment of the present invention will be described. In the third embodiment, a moving object detection device is configured using a personal computer that has been widely used in the past. That is, as shown in FIG. 7, the constituent circuits 13 and 14 shown in FIG. 1 are assembled on one board 33, and this board 33 can be freely attached to an expansion slot of a personal computer. I do. The board 33 has a function of performing pre-processing and pattern matching as shown in FIG. 9, including first contrast correction, skip processing, and vector connection processing, which are characteristic parts of the present invention. Provided. Note that this embodiment is characterized in that it performs contrast correction, skip processing, and vector connection processing, and other configurations and operations are the same as those of the first, second, and second embodiments described above. It is. Therefore, overlapping description will be omitted or simplified, and the following description focuses on the features of the present embodiment.

In the moving object detection device according to the present embodiment, a contrast correction circuit is provided in the preprocessing circuit 19 (see FIG. 5). This contrast correction circuit, as shown in Figure S, First, an LUT (Look Up Table) is created for the first frame 30 on the time axis of each GO P (Group Of Picture), and the LUT (Look Up Table) is created. The remaining frame 31 functions to quantize using the LUT created in the first frame 30 above. The GOP is a force which is a concept in the MPEG, which is one standard of the compression processing. This G0P corresponds to a frame group described in the claims. Further, the contrast correction circuit constitutes a first contrast correction means described in the claims.

The above-mentioned contrast correction circuit is provided for the following reason. That is, quantization (first and second quantization) for obtaining a positive frame (image frame subjected to the first quantization) and a negative frame (image frame subjected to the second quantization). ), Quantization is performed in advance with the default settings, luminance is corrected for the frames quantized with the default settings, and the image displayed on a display such as a monitor is monitored as described above. So that it is easy for people to see. Such correction is performed in the conventional structure. That is, for example, in the case of a monitoring device that configures a security system, the monitoring is often performed in darkness, and the video projected on the display is extremely large as it is. It is difficult for people to monitor. For this reason, it is necessary to perform contrast correction.

By the way, the conventional contrast correction first performs quantization with default settings, then creates an LUT (Look Up Table) by scanning all pixels, and then re-references this LUT. Perform quantization. The LUT refers to an operation process of converting a luminance equal to or higher than a desired value among pixels constituting a frame input and quantized with default settings to a predetermined luminance to obtain a desired contrast. However, since the above-described conventional contrast correction needs to follow the above-described procedure, it is inevitable that the processing time for image processing increases. The contrast correction according to this embodiment was created in view of the current situation.

In the present embodiment, the above-described skip processing means is provided on the board 33. This skip processing means is constituted by an arithmetic circuit (MPU) not shown. Can be The effect is as shown in the flowchart of Fig.10. That is, at least one of the preprocessed image frames is set as the search target frame 34, and the maximum density value (SH) and the minimum density value (SH) are obtained from all the pixels constituting the search target frame 34. SL), and use them as the first threshold SH and the second threshold SL, respectively. At the same time, the difference (SPP) between the maximum density value (SH) and the minimum density value (SL) is calculated and used as the third threshold value SPP (step 1).

In addition, as step 2 in parallel with step 1, the remaining frames of the above image frames other than the search target frame 34 are set as template frames 35, and the maximum of all the pixels constituting the template frames 35 is set. Calculate the glow value DH and the minimum density value DL. As described above, in the search target frame 34 and the template frame 35, if the respective values SH, SL, DH, and DL are obtained, as step 3, the values obtained from these template frames 35 are obtained. It is determined whether the maximum Pu value DH and the minimum concentration value D exceed the first threshold value SH and the second BB value SL, respectively. If none of the maximum DH and the minimum concentration DL exceed the first and second thresholds SH and SL (DH <SH, and DL> SL), the template frame 3 It is determined that no moving object exists in 5 and the vector generation processing is skipped (step 4).

On the other hand, both the maximum concentration value DH and the minimum concentration value DL force obtained from the above template frame 35 exceed the above first and second 闼 values SH and SL (DH> = SH, (DL <= SL), proceed to the next step 5, and calculate the difference DPP between the maximum density value DH and the minimum density value DL obtained from the template frame 35 and the third 闞 value SPP. Compare,. The above difference DPP is calculated in advance in step 2. If the difference DPP is smaller than the third threshold value SPP (DPP + SPP), it is determined that no moving object exists in the template frame 35, and the process jumps to the step 4 to generate the vector. To skip.

On the other hand, when the maximum density value DH and the minimum value DL obtained from the template frame 35 exceed the first and second thresholds SH and SL, respectively (DH> = If SH and DL <= SL) and the difference DPP is greater than or equal to the third 閱 value SPP (DPP> = SPP), the template frame 35 It is determined that there is a moving object in, and the process proceeds to the vector generation processing in step 6.

The reason for providing such skip processing means will be described below.

Performs quantization (first and second quantization) to obtain a positive frame (image frame subjected to the first quantization) and a negative frame (image frame subjected to the second quantization), When extracting a stationary portion common to each of these frames, a process called so-called pattern matching is performed. This pattern matching is widely used for positioning of various components, character recognition, defect inspection, etc., and is an imaging technology that is the core of FA (Factory-Automation). Such pattern matching is performed to quickly determine whether or not a moving object is present in an image frame (moving object detection frame) in which the positive frame and the negative frame are superimposed. The processing is performed by using the correlation according to 0. Such pattern matching is composed of skip processing, vector generation processing, and determination processing as shown in Fig. 9. Among the above, the skip processing is This is performed so that the next frame generation processing step can be omitted for an image frame in which no moving object is assumed to exist due to the shading degree value, etc. Therefore, if this skip processing can be performed reliably, the moving object The time required for the detection is reduced, and the effective skip processing contributes to the improvement of the robustness, and is also preferable from this point. The skip processing described above is performed based on such a viewpoint.

Further, after the above-described vector generation processing, a force for performing the vector connection processing is provided. For this purpose, a vector connection processing means is provided. This vector connection processing means can also be constituted by an arithmetic circuit (the same as the skip processing means), like the above-mentioned skip processing means. The vector connection processing means detects a change in coordinate data in each moving object detection frame among the pixels constituting the moving object detection frame. That is, it detects how each pixel in one moving object detection frame changes its coordinate position with time. For the coordinate position (state data), XY orthogonal coordinates are virtually set in the moving object detection frame, and the position is detected by this. At the same time, the moving direction of each of these pixels in each moving object detection frame is detected. That is, the movement vector of each pixel is detected. This vector is the coordinate data of each pixel in the adjacent motion detection frame. Can be easily obtained from Then, an object consisting of all the pixels that almost coincide with the change in the coordinate data, the moving direction, the force, and the 'is regarded as one moving object.

Explaining with reference to the drawings, in the vector connection processing of the conventional structure, the moving object detection frames 31 and 31 as shown in FIGS. 11 (A) and (B) are used as shown in FIG. 11 (C). However, only the frame 32 in which a single moving object 42 exists can be obtained. On the other hand, in the moving object detection method according to the present embodiment, the moving direction is detected along with the change in the coordinate data, and it is determined whether the moving object exists in the image.

From the frames 31 and 31 of 1 (A) and (B), two moving objects 43 and 44 to be detected are accurately detected as shown in FIG. Of course, the influence of disturbances due to the above-mentioned lighting devices and wind and rain is eliminated with the small order, and therefore, it is not recognized as a moving object.

The reason for providing the above-described vector connection processing means will be described below. The vector connection process is a process for determining the shape and size of the moving body portion. To determine the shape and size of the moving body portion in this manner, the input of each pixel constituting the input frame is performed. It is conceivable to detect a change in coordinate data in a frame and regard a portion constituted by a pixel having a change and pixels in the vicinity thereof as a moving object. Specifically, by detecting a change in the coordinate data of each of the pixels constituting the input frame in the input frame, the pixels having the change are indicated by a diagonal grid in Fig. 12 (A). Assuming that the pixels that do not change are those on a white background, it is judged that there are three moving objects A, B, and C as shown in Fig. 11 (B) in the case of the vector concatenation process, which is usually considered. I do.

However, such vector connection processing has the following inconveniences. In other words, pixels other than the detection target (for example, a suspicious person) may be included in the pixels determined to be a part constituting the moving object due to the change in the coordinate data. For example, disturbances caused by a lighting device such as a fluorescent lamp, disturbances caused by natural phenomena such as wind and rain, etc. may determine that a moving object similar to the detected object exists even for objects other than the detected object. In this way, even if it is something other than the detection target However, assuming that the same worker as the object to be detected exists> The processing required for detecting the moving object is unnecessarily increased, and the processing speed is reduced. It also impedes reliable detection of moving objects.

Furthermore, when judging the presence or absence of a moving object simply by changing the coordinate data, even if there are two or more independent detection targets, these detection targets are in an adjacent state (for example, a scrum ), The plurality of detected objects may be determined to be one. That is, in the vector connection process that is generally considered, only one moving object A shown in FIG. 11 (C) is detected from the frames shown in FIGS. 11 (A) and 11 (B). Thus, it is not preferable that the number of detected moving objects is reduced in spite of the fact that there are actually a plurality of moving objects. In view of the circumstances described above, the vector connection processing according to the present embodiment was devised and provided.

Since the moving object detection device according to the present embodiment is configured as described above, the process for determining the presence or absence of a moving object can be performed quickly and reliably. In other words, the contrast correction circuit creates an LUT for the first frame 30 on the time axis of each GOP, and creates the LUT for the remaining frame 31 of each GOP. Quantize using the LUT created in frame 30. Therefore, quantization with default settings and creation of LUT by scanning all pixels, which were conventionally required for contrast correction, can be omitted. As a result, the processing procedure can be shortened, and the processing speed can be improved accordingly. The above LU is equivalent to the contrast correction information described in the claims.

Further, in the skip processing means, the skip processing is performed in two stages, so that the speed of the pattern matching processing can be increased and the mouth-cutting performance can be improved. Further, since it is detected whether or not each density value of the template frame 35 exceeds each threshold value on both the high density side and the low density side, the possibility of accidental data squibbing can be eliminated. . Further, since the first, second, and third IM values are obtained from all the pixels of the search target frame 34, each of these ^ values themselves can take a value suitable for the input image condition.

The first, second, and third values SH, SL, and SPP are calculated by the following equations. SH = Ave + Hi (1)

S H = A v e-HI (2)

S P P = ρ + 2 (3)

In Equations (1), (2), and (3), Ave represents the average value when n is the total number of pixels in the search target frame and X is the density value of each pixel. p represents the variance. The average value Ave is given by the following equation (4), and the variance p is given by the following equation (5). The above DH, DL and DPP are calculated in the same manner.

A v e-(Σ χ) η (4)

ρ = (η Σ χ - ( Σ χ) 2) / χ ι · · · (5)

Since the present embodiment is configured as described above, it is prevented from being recognized as a moving object similar to the object to be detected, despite the fact that the object is not the object to be detected due to disturbance or the like as described above. Also, despite the existence of a plurality of detection targets, the number of moving objects to be detected does not become smaller than the actual one. As a result, unnecessary data processing is omitted, and the speed and reliability of the moving object detection process can be improved. Further, the moving object detection device according to this embodiment can be configured by incorporating the above-described processing means into an arithmetic device (MPU, CPU) or the like provided in the preprocessing circuit 19 or the like. It is needless to say that conventionally known buses such as the ¾, the board 33 having the configuration shown in FIG. 7 and the bus using the PCI bus can be used. In FIG. 7, reference numeral 36 denotes a display. Needless to say, the board 33 has a function of performing the preprocessing and the pattern matching (skip processing and vector connection processing).

Note that, also in the moving object detection device according to the present embodiment, the input image is grayscale. When a moving object is detected by the above-described apparatus of the present embodiment, the position, angle, and focal length of the apparatus need to be fixed. These reasons are as described in the first embodiment.

In addition, a configuration as shown in FIG. 13 can be adopted based on the same concept as the configuration of the third embodiment described above. That is, a video control circuit and the like are provided on the board 33a together with the above-mentioned constituent circuits 13 and 14, and a personal computer is provided. By storing software for video control etc. in the RAM, it is possible to freely output judgment results. In the case of this example, the through output signal is supplied to the video deck 37, and the video is displayed on the monitor display 38. Further, the image signal subjected to the above-described contrast correction and the like is sent to the display 36, and the image is displayed on the display 36. In the case of the configuration shown in Fig. 13, the monitor display 38 displays the image as it is projected by the CCD camera; the display 36 of the personal computer is contrast-corrected. An image with high brightness and brightness is displayed. Since the correction process at this time is performed promptly as described above, a large time lag does not occur on the display screen of the monitor display 38.

More specifically, a camera adapter is connected to the CCD camera. From this camera dub, recording is performed as a still image or the like on recording means such as a hard disk via the board 33a. At the same time, an image is sent from the camera adapter to the monitor display 38 via the video deck 37. A conventionally known VCR 37 can be used, and the video is also recorded by the VCR 37. Further, in the case of this example, a function that can more effectively detect a moving object can be incorporated into the board 33a. For example, the detection range can be surrounded by a red line, or the moving direction of the moving object can be indicated by an arrow. Such a function can be easily performed by a conventionally known image processing technology.

In the third embodiment described above, an LUT is created for the first frame on the time axis of each G 0 P, and the first frame is created for the remaining frames of each GOP. The example using the contrast correction circuit (first contrast correction means), which is created in the frame and quantized using the UT, has been described. The circuit is configured to create an LUT for the first frame on the time axis-For frames following the above frame, quantize using the LUT in the frame before the relevant frame (Second contrast correction means) Yes. When such a contrast correction circuit is provided, the processing speed is inferior to the contrast correction circuit of the embodiment described above, but the default setting required for the contrast correction is used. Quantization and scanning of all pixels in each frame LUT creation can be simplified, and the processing procedure can be shortened, resulting in an improvement in processing speed.

The moving object detection method and the moving object detection device according to the present invention are not limited to the above-described example, and can take various configurations. For example, it can be an independent device without being incorporated into the board 3333a (without using a personal computer). It is also possible to select and execute any of the above-described contrast correction, skip processing, and vector connection processing. INDUSTRIAL APPLICABILITY As described above, the moving object detection method according to claim 1 is configured as described above, and is easy and reliable despite the small increase in manufacturing cost. The excellent effect that the moving object detection can be performed is obtained.

Furthermore, in addition to the above effects, the moving object detection method according to claims 2 and 3 can improve detection performance and processing speed under low illuminance conditions, and can quickly detect moving objects.

Furthermore, the moving object detection method described in claim 4 can achieve a reduction in the time required for moving object detection and an improvement in mouth-bustness, in addition to the effects described in claim 1.

Furthermore, the moving object detection method according to claim 5 has the effect described in claim 1 and also prevents the object from being recognized as a detection target despite being not a detection target due to disturbance or the like. It is possible to ensure and speed up moving object detection. In particular, in the case of this method, only a person who enters the room or only a person who leaves the room can be detected, which has a great practical effect.

Furthermore, the moving object detection method according to claims 6 and 7 has all of the above effects. That is, the processing speed can be improved, and quick and reliable moving object detection can be performed. In addition, the robustness can be improved, and the possibility of accidental skipping can be eliminated.

Furthermore, since the moving object detection device according to claim 8 is configured as described above, it is possible to easily and surely detect a moving object despite a small increase in manufacturing cost. An excellent effect of being able to provide a body detection device is obtained. For example, if this moving object detection device is used as a monitoring device, the load required for the monitoring operation can be reduced and the moving object can be reliably detected. Moreover, since the imaging means can be configured by adding a slight configuration to the conventional CCD camera or by attaching an adapter, the configuration can be made easily and the manufacturing cost is not increased. Therefore, the effect as a monitoring device is great. Further, in addition to the effects of the moving object detection device according to the eighth aspect, the moving object detection device according to the ninth to eleventh embodiments provides quick notification (alarm), reliable and efficient recording. Is possible, and the practical effect is great. That is, the conventional monitoring device records video using a recording medium such as an endless tape regardless of the presence or absence of a moving object. Therefore, recording media such as tapes were often wasted, and there was a strong demand for improvement. On the other hand, in the case of the structure of the present invention, since the recording means can be separately activated only when the detecting means detects the moving object, the recording medium can be used efficiently.

As the recording means, a recording medium such as the above-mentioned endless tape is used to record a signal input to the imaging means, or a combined image created by an imaging means to extract the moving object. Record the frame data on a recording medium. Further, such image information can be compressed and stored by a conventionally known compression means.

Furthermore, the moving object detection device according to claim 12 or claim 13 is to improve the detection performance and processing speed under low illuminance conditions in addition to the effect of the moving object detection device according to claim 8. And can quickly detect moving objects.

Furthermore, in addition to the effects described in claim 8, the moving object detection device described in claims 14 and 15 is caused by shortening the time required for moving object detection, improving robustness, and disturbance. Thus, it is possible to prevent recognition as a detection target despite being not a detection target. In other words, the detection of moving objects can be reliably performed and the speed thereof can be increased, and the practical effect is great.

Claims

The scope of the claims

. By alternately inverting the quantization direction of the input frame obtained by the imaging means, a positive frame and a negative frame are alternately output, and a stationary portion common to the forwardly adjacent positive and negative frames. A moving object detection method characterized by extracting a moving object portion by equalizing the density value of the moving object.

Contrast correction for correcting the luminance of a first frame on the time axis of the input frame group consisting of a desired number of input frames to a predetermined value for those whose luminance is higher than the desired value And requantization of the remaining frames in the input frame group using the contrast correction information obtained in the first frame. The moving object detection method according to claim 1.

Contrast correction for correcting the luminance to a predetermined value for the first frame on the time axis of the input frame group consisting of the desired number of input frames for those whose luminance is higher than the desired value And requantizing the frame subsequent to this frame by using the contrast correction information in the frame preceding the frame. Moving object detection method.

At least one of the input frames is set as a search target frame, and a maximum density value and a minimum density value are calculated from all pixels constituting the search target frame, and the first density value and the second density value are respectively determined. And the difference between the maximum density value and the minimum density value is calculated as the third 閟 value, and the remaining frames excluding the search target frame from the input frame 0 are set as template frames, The maximum density value and the minimum density value are calculated from all the pixels constituting the template frame, and the maximum density value and the minimum density value obtained from the template frame are calculated as the first threshold value and the second density value, respectively. It is determined whether the maximum density value and the minimum temperature value do not exceed the first and second minimum values. It is determined that no moving object exists in the template frame, and if the maximum density value and the minimum density value obtained from the template frame both exceed the first and second thresholds, The difference between the maximum density value and the minimum density value obtained from the template frame is compared with the third 闞 value. If the difference is smaller than the third 闼 value, the difference is set to the template frame. Is It is determined that there is no moving object, and when the maximum density value and the minimum density value obtained from the template frame both exceed the first and second 閱 values, the difference is the third value. The moving object detection method according to any one of claims 1 to 3, wherein it is determined that a moving object is present in the template frame when the value is equal to or greater than the threshold value.

Changes in coordinate data in each input frame of pixels estimated to form a moving object portion among pixels constituting the input frame and pixels in the vicinity thereof, and changes in these images and images in the vicinity thereof. The moving direction and> in the input frame are detected, and an object consisting of all pixels in which the change in the coordinate data coincides with the moving direction and an object composed of pixels in the vicinity thereof are regarded as one moving object. The moving object detection method according to any one of claims 1 to 4.

For the first frame on the time axis of the input frame group consisting of the desired number of input frames, a contrast correction that corrects the luminance to a predetermined value for those whose luminance is higher than the desired value is performed. And then requantized.Then, the above-mentioned contrast correction of the input frame group is performed, and the re-quantization is performed.The remaining frames of the input frame group are obtained in the first frame. Requantization is performed using the contrast correction information, a first contrast correction is performed, and then at least one of the moving object detection frames obtained by performing the first contrast correction is performed. The maximum number of pixels and the minimum density value are calculated from all the pixels constituting the search target frame, and the first and second K values are set as the first 闼 value and the second K value, respectively. The difference between the large-rarity value and the minimum-rarity value is calculated and used as the third value. The remaining frames excluding the search target frame among the moving object detection frames are used as template frames. The maximum density value and the minimum density value are calculated from all the pixels constituting the frame, and the maximum and minimum intensity values obtained from these template frames are calculated as the first K value and the second density value, respectively. If the maximum density value and the minimum density value do not exceed the above 1st and 2nd 1 比較 values, a moving object does not exist in the template frame. The maximum density value and the minimum density value obtained from the template frame both exceed the first and second threshold values. In this case, the difference between the maximum density value and the minimum density value obtained from the template frame is compared with the third 閱 value, and the difference is smaller than the third 閟 value. For example, it is determined that no moving object exists in the template frame, and when the maximum density value and the minimum density value obtained from the template frame both exceed the first and second 閎 values, If the difference is equal to or greater than the third value, the skip processing is performed to determine that a moving object exists in the template frame, and the moving object detection is performed for each of the pixels constituting the moving object detection frame. The change in the coordinate data in the frame and the moving direction of each of these pixels in the moving object detection frame are detected, and all the pixels in which the change in the coordinate data and the moving direction almost match are detected. Moving object detection method according to claim 1, wherein, wherein the performing base-vector linking process La configured object regarded as one moving object.

Instead of the first contrast correction, for the first frame on the time axis of the input frame group consisting of a desired number of input frames, the brightness is set to a predetermined value for a frame whose brightness is equal to or more than a desired value. Re-quantize by applying a contrast correction to correct the value, re-quantize by performing the above-mentioned contrast correction of the input frame group, and re-quantize the frame following this frame. 7. The moving object detection method according to claim 6, wherein a second contrast correction for performing requantization using the contrast correction information in a previous frame is performed.

. A photoelectric conversion means for converting light input into a charge signal and storing the charge signal. The photoelectric conversion means converts input optical information having a luminance of a predetermined value or more to a high level and a signal of a luminance lower than a predetermined value. The first quantization, which quantizes the luminance as a low level, and the second quantization, which quantizes the luminance whose luminance is equal to or higher than a predetermined value as a low level and the luminance lower than the predetermined value as a high level, are alternately performed. The first quantized image frame is compared with the second quantized image frame following the first quantization, and the pixels having the same luminance are canceled by the two. Thus, a moving object detection device is characterized in that pixels having different luminances can be extracted freely in both cases, which are regarded as moving objects.

9. The moving object detection device according to claim 8, wherein the photoelectric conversion unit is an imaging unit, and the detection unit detects a moving object extracted by the imaging unit.

10. The moving object detection device according to claim 9, further comprising a notification unit that operates when the presence of the moving object is detected by the detection unit and notifies the presence of the moving object.

11. A moving object detection device according to claim 9 or claim 10, comprising a recording device that operates when the presence of the moving object is detected by the detection means and records the moving object. apparatus.

12. Contrast correction means for performing contrast correction for correcting a luminance of a pixel having a desired value or more to a predetermined value among pixels in a state where the light input is quantized, and The photoelectric conversion means is capable of converting the light input corrected by the contrast correction means into a charge signal, and the contrast 0 correction means is capable of converting the input frame group comprising a desired number of input frames. The contrast correction is performed on the first frame on the time axis and requantized, and the contrast correction obtained on the first frame is obtained for the remaining frames of the input frame group. The moving object 5 detection device according to any one of claims 8 to 11, wherein the device has a function of performing requantization using information.

13 3. Among the pixels in a state where the light input is quantized, a contrast correction means for performing a contrast correction for correcting a luminance of a pixel having a luminance equal to or more than a desired value to a predetermined value. And the photoelectric conversion means is capable of converting the light input corrected by the contrast correction means into a charge signal, and the contrast 0 correction means is capable of converting a desired number of input frames from a desired number of input frames. The first frame on the time axis in the group of input frames is requantized by performing the above-mentioned contrast correction, and the contrast of the frame following this frame in the frame preceding that frame is performed. 12. The moving object detecting device according to claim 8, wherein the moving object detecting device has a function of performing requantization using the correction information.

14. First contrast correction means, skip processing means, and vector connection processing means, wherein the first contrast correction means includes an input frame group composed of a desired number of input frames. Of the first frame on the time axis is requantized by performing the above-described contrast correction. The second contrast processing means has a function of requantizing the image using the contrast correction information obtained in the first frame, and the skip processing means is provided by the first contrast correction means. At least one of the moving object detection frames obtained by performing the contrast correction is set as a search target frame, and a maximum value and a minimum density value are calculated from all pixels constituting the search target frame. The first and second threshold values are used as the first and second threshold values, respectively, and the difference between the maximum density value and the minimum density value is calculated as a third threshold value. The remaining frames excluding the target frame are set as template frames, and the maximum and minimum density values are calculated from all the pixels constituting the template frames. It is determined whether the maximum density value and the minimum density value obtained from the plate frame exceed the first threshold value and the second 第二 value, respectively, and the maximum density value and the minimum density value are determined. If none of the above-mentioned first and second values is exceeded, it is determined that no moving object exists in the template frame, and the maximum density value and the minimum density value obtained from the template frame are In any case, when the values exceed the first and second values, the difference between the maximum density value and the minimum density value obtained from the template frame is compared with the third threshold value. If the difference is smaller than the third threshold value, it is determined that no moving object exists in the template frame, and both the maximum density value and the minimum intensity value obtained from the template frame are higher. If the difference exceeds the first 1¾2 value and the difference is equal to or greater than the 3¾1¾ value, it has a function of determining that a moving object exists in the template frame. The vector connection processing means detects a change in coordinate data of the pixels constituting the moving object detection frame in each moving object detection frame, and a moving direction of these pixels in each moving object detection frame. Claims 8 to 11 are characterized in that they have a function of regarding an object composed of all the pixels whose change in the coordinate data and the moving direction substantially coincide with each other as one moving object. The moving object detection device according to any one of the above.

In place of the first contrast correction means, the contrast is applied to the first frame on the time axis of the input frame group composed of a desired number of input frames. A second contrast correction means having a function of performing re-quantization by performing a contrast correction and re-quantizing a frame subsequent to this frame by using contrast correction information in a frame preceding the frame. 15. The moving object detection device according to claim 14, further comprising: