WO2005010840A1

WO2005010840A1 - Method for automatic protection

Info

Publication number: WO2005010840A1
Application number: PCT/CH2004/000465
Authority: WO
Inventors: Phillip GEISSBÜHLER; Anton Gunzinger; Markus Emil Schenkel; Maurice Boonen
Original assignee: Supercomputing Systems Ag
Priority date: 2003-07-25
Filing date: 2004-07-23
Publication date: 2005-02-03
Also published as: CH697014A5

Abstract

The method for protection of systems can evaluate a recorded image of a monitored region with imaging processing means such that predictions as to the dimensions of the object are possible. A light source (1) in an illumination direction is thus used to project at least one line of light onto a surface (4) for monitoring, generating a light pattern (2). The pattern of light is at least partly recorded from a direction different to that of the illumination direction by means of an image surface sensor (3). The camera signal thus generated is then fed to an image processing device (6), which determines the deviation of the light pattern (2) from a reference standard. By quantitative evaluation of the separation, using the image processing means, the height of the object can also be approximately determined.

Description

METHOD FOR AUTOMATIC SECURING

The invention relates to a method and a device for automatic protection of systems.

Securing machines, doors, vehicles and other objects is important in modern society in both the productive and leisure sectors. This safety function is mainly implemented today with the help of optical systems:

According to the state of the art, one possibility for protection is based on the light barrier principle: a light source emits one or more light beams that strike one or more light catchers. If one or more light beams are interrupted, an alarm is triggered. These methods have the advantage that all components involved are always checked during normal operation.

On the one hand, the disadvantage of these methods is the high adjustment effort

(Light transmitter optics, monitoring path optics must be aligned) and, on the other hand, the complex structure when surfaces or even rooms are to be monitored. Since light rays are always interrupted in these processes, Room surveillance (e.g. for robots or on autonomous vehicles) requires complex and expensive structures with mirrors or light guides.

A description of an alternative method for monitoring systems based on the known triangulation principle can be found in EP-A-1 300 691. According to this document, a pattern of light spots is projected onto a monitoring area. A receiving device determines whether light spots have moved away from a certain area, which indicates the presence of an object. Disadvantages are the great expenditure of optical elements for the generation of a pattern of light spots and the lack of possibility to make quantitative statements about the object.

It is therefore an object of the invention to provide a method and a device which overcomes the disadvantages of existing methods and devices and which makes it possible to provide reliable protection, even in the case of small, disruptive objects, using means that are not visually complex. The method and the device should preferably also enable the quantitative determination of properties of the monitored or interfering objects.

This object is achieved by the invention as defined in the patent claims.

The method according to the invention uses the triangulation principle which is known per se, which, to put it simply, is based on the fact that the position of a light reflex on an object depends on its distance from the viewer, if the direction from which the light source shines depends on the Direction of view is different. The triangulation principle is used here in such a way that it can be implemented cost-effectively with the use of components such as, for example, a laser light source and area image sensor and image processing techniques that run on corresponding image processing computers.

In the context of this document, “light” is always understood to mean both visible light and electromagnetic radiation in the non-visible area. “Image” and “image processing” are likewise not limited to images taken in the visible, they can also be infrared images, etc.

The invention is also based on the idea that when monitoring with the aid of image processing means, quantitative statements about the object and its

3D properties can be made. This is based on the fact that in the case of known - or determined by one-time, camera type-specific calibration - optical

Characteristics of the camera an image position corresponds to a certain angle of incidence. With trigonometric formulas known per se from conventional triangulation methods, a height of the object or the distance to the

Object can be determined from the angle of incidence. According to one embodiment, the displacement of the

Light pattern therefore evaluated quantitatively, according to a special

Embodiment, the limitation of accuracy is overcome by a pixel resolution of the camera by performing a sub-pixel approximation.

The method according to the invention avoids disadvantages of existing systems. It is optically very simple. All that is required are means for generating at least one light line - for example a “laser light meter” - and an image area sensor, for example a camera. There is no need for complex optical adjustment. The image processing that follows the image acquisition requires a certain amount Electronics and programming effort. However, this is not significant for a large number of pieces, since electronic components - in contrast to precise optical instruments - can be manufactured inexpensively in large quantities and, in particular, since no complex setup of each individual location is necessary.

Furthermore, the approach according to the invention enables not only the detection of objects but also the detection of their properties. For this purpose, according to a preferred embodiment, the deviation of the light pattern from a reference (for example, corresponding to the image of the light pattern, if there is no object in the detection area of the image area sensor) is recorded quantitatively and thus height information or distance information is obtained. Together with speed information - such information may be available if the monitored space is above a conveyor belt, for example - the volume of an object can thus be determined. This opens up additional options for monitoring machines. The invention is therefore not limited to systems in which only disturbing objects are detected, the detection of which causes an alarm signal or a machine to be switched off. Rather, it also includes the possibility of performing complex monitoring or even sorting functions.

An embodiment of the invention and its mode of operation is shown schematically below with reference to a drawing.

Figure 1 of the drawing shows a schematic of the process.

Figure 2 shows very schematically the basic principle of triangulation. FIG. 3 shows an illustration of the principle of sub-pixel approximation.

Figure 4 shows an illustration of the principle of the difference image calculation with 2 cameras.

1 schematically shows an area 4 to be monitored. In the following description it is assumed that an object 0 is located in the area to be monitored, that is to say on the area 4 to be monitored. A 3-dimensional image is to be generated from this object, which in the example described then triggers an alarm. The space to be monitored is illuminated with the aid of a “laser light meter” 1 as a straight line of light per se, the light direction having to be shifted relative to the camera axis that records the scene. The light meter essentially cuts through the entire area to be monitored and is, for example, . not interrupted or at most only interrupted by short sections, so that the total interruptions make up at most a small fraction (eg a tenth) of the total width. This ensures that narrow and marginal objects are not overlooked. The "light meter" is now reflected from the empty space as a line. If there is an object 0 in the room, this line is changed.

Instead of a single light line, it is also possible to use a plurality of light lines which are parallel or in any arrangement not parallel to one another and which complement one another or to form a pattern, for example by crossing, continuing with one another, etc. The light line (s) Of course, must not necessarily be straight. Different light patterns can be projected at different times and / or from independent light sources. Lasers, for example diode lasers, but also other preferably monochrome light sources (ie Light sources with a limited bandwidth compared to the spectrum of visible light), for example light-emitting diodes (LEDs).

According to a special embodiment, as long as there is no object in the monitoring area, no light is thrown back from the surface 4, since there is no object on it. Then the light pattern is projected almost to infinity. The light pattern can therefore only be detected when there is an object in the camera area. In this embodiment as well, the light pattern on the surface 4 serves as a reference, ie the pattern that would arise if there were a flat object on the surface 4. In this embodiment, it is also possible to detect objects that are located behind the surface.

The procedure works as follows:

An image area sensor 3 records the scene. The light pattern 2 arising from the light line is also visible on this recording.

The signal from the image area sensor 3 is converted to a digital signal in the analog / digital conversion unit 5 and is thus available to the subsequent evaluation units as a digital signal.

The image evaluation 6 detects the course of the light line 2 in the image and creates a 3-D model from it in the manner set out above.

The control computer 7 recognizes the situation. On the basis of this change, the control computer 7 can generate an alarm signal 8. The image recorded during the alarm situation - recorded by the image area sensor 3 or possibly by another camera - can possibly be saved for documentation purposes. In addition or as an alternative to the alarm signal, the height or distance of the object can be determined quantitatively depending on the lateral position. The principle of triangulation is shown very schematically in FIG. Due to the presence of object 0, the point of impact of the light, which is perceived by an image area sensor 3, is perceived as shifted in comparison to the state without object 0 (dashed line). Using trigonometric formulas, the height or distance of the object is calculated from the angle of incidence α and the angle of incidence β as well as the known distance of the camera from the monitored plane. The specific formulas to be used (even in the case of arrangements that differ from the drawn configuration) are known from classic triangulation and are not listed again here. The present invention makes use of the knowledge that the angle α is expressed in a positional shift of the light pattern on the image due to the camera optics. The deviation in the image of the light pattern from the reference can be determined by means of image processing means, and from this the angle α or, as explained, the height of the object can be inferred.

The positional shift of the light pattern on the image recorded by the camera (ie the image area sensor) can be determined, for example, by evaluating which image points (pixels) the light pattern is mapped to. In general, the image of a thin line will also not only trigger signals in a single row of pixels, but also in neighboring pixels. This can be used to determine the position with a resolution that is finer than the "pitch", ie the size given by the pixel spacing. The position of all pixels in which a signal is measured is averaged. The averaging is weighted according to signal strength. This is done, for example, via all pixels lying next to one another in one direction (for example, x-direction in an xy-Cartesian coordinate system), specifically for all rows of pixels (in the example, for all y values). The method can be used to determine precise position information with sub-pixel accuracy, even if the pattern is not exactly mapped to the center of a row of pixels. This is illustrated in FIG. 3, where the intensities 22 are plotted as a function of the x position of the pixels 21. The actual position 23 determined with the sub-pixel approximation is not in the center of the pixel with the highest intensity value.

A lack of the light line or parts of it in the image or, depending on the chosen structure, a shift of the entire light line compared to the reference indicates a lack of functionality of the system. A change in the brightness of the light line in the image can also indicate a lack of functionality. In this case, a suitable warning signal and / or a check can be triggered. In contrast to known monitoring systems, the presence of an object - as an alarm case - can be distinguished from a system failure - as a malfunction - according to the approach according to the invention.

If there is enough light from the light pattern, the line 2 at the viewing angle 4 of the image area sensor is much brighter than all other objects; in this case, the image evaluation 6 can be carried out by simply setting a threshold value: if the signal is greater than this threshold value, it is the light line 2, otherwise it is not part of the light line.

Now this protection device should also be in different

Lighting conditions can be used. It also has to work in direct, competitive solar radiation. So that is the case , the light output of the light source 1 would have to be increased so much that the eye of a person could be damaged by direct radiation. Various measures can be taken to reduce the light output so much that it becomes harmless to humans without endangering the way the system works. These measures are to be explained in more detail below:

The sun shines over a large frequency spectrum. Laser light sources, such as are usefully used as light source 1, are monochrome. By using an optical filter 9, for example a band-pass or low-pass filter in front of the camera, which is tuned to the frequency range of the laser, most of the sunlight or other lighting can be suppressed. The optical filter thus favors the recognizability of the monochrome light source in relation to the ambient light. (Note: Lasers can be used both in the visible and in the invisible area. The selection depends on the application; should the endangered area be visible (e.g. machine security) or should it be invisible (e.g. access security))

The light source 1 can usually be modulated or pulsed; for a short period of time it can be 100 or 1000 times the continuous output. At the same time, modern cameras have an electronic shutter. This can be synchronized with the light pulse. In Fig. 1, the control computer 7 controls the light source via line 10 and the shutter of the camera via line 11.

Despite the above measure, bright areas of light can still appear in the image, for example due to direct headlight radiation in the recording area of the camera 4. In order to suppress these interference signals, an image can be suppressed without the light source 1 is switched on, recorded and temporarily stored in an image memory 12. The picture is then taken with the light source switched on. This is compared by the comparator 13 with the stored image (for example by forming a difference). As a result, slowly changing strong lights in the recording area 4 can be suppressed comparatively, for example in comparison with the recording frequency of the camera. The picture can be taken with the light source switched off once at the beginning of a work cycle or continuously - alternating with lighting periods.

If there is a highly dynamic situation in terms of lighting, be it that the monitoring sensor is mounted on a vehicle or that it has to carry out object monitoring with fast objects, the recording of a reference image can result in an additional intolerable dead time. In this case, a method according to FIG. 4 is suitable. In this arrangement, two sensors 3, 3 'are available. One sensor 3 triggers the image acquisition at the same time as the pulsation of the light source, while the other sensor 3 'immediately takes an image, the image acquisition time being approximately identical for both sensors. Both sensor signals are converted into digital signals in the analog / digital converters 5, 5 '. The transformation unit 20 'transforms the signal of the second sensor 3' geometrically and in terms of light intensity to the geometric position of the first sensor 3. In the comparator 13, the two images are compared (e.g. by forming a difference) and then further evaluated analogously as described above.

It can happen that the lighting conditions in the surveillance area change very quickly if, for example, suddenly there is a pattern on the surveillance area with very different reflection properties: one sub-area reflects very strongly, while another sub-area hardly reflects. If it is the allow time requirements, the gain of the exposure time of the sensor at the next image acquisition can be adjusted so that a meaningful evaluation is then possible. However, if the time requirements for the system are very high, in this case an additional sensor system (i.e. an image area sensor including A / D converter, possibly its own optics, and possibly the means for reference formation) or more additional sensor systems can help: While one sensor system is set to be very light-sensitive with a first image area sensor and overdrives for large input signals, a second sensor system with a second image area sensor can take over the area which is not very light-sensitive. The sensor system which provides better signal statistics is used for further processing. As an alternative or in addition to this, the images recorded by both sensor systems can also be combined, for example after correction or geometric transformation, for example by addition. The acquisition of the images exposed to different intensities can, but does not have to take place simultaneously.

For example, more than one sensor system can also be used in situations in which an object's shadow can lead to inaccessible areas, or in order to simultaneously detect the height and another geometric dimension of the object. These, for example, equivalent at least two sensor systems are preferably used simultaneously.

In the simplest version, the device described here can be used to detect a change and thus for simple machine protection. The system can be calibrated; the height or distance of object 0 can thus be calculated directly from the displacement of light line 2 in the image. If the geometric arrangement is known, the height or distance of the recording system can also be calculated. If object 0 is moving at a known speed (eg on a conveyor belt for parts or people), the volume of object 0 can also be estimated with this method. Under certain circumstances, the image processing device can also be provided with means for specifying an object speed in addition to the above. In the arrangement shown in FIG. 1, the object can be conveyed, for example, in a plane containing the surface 4 and in a direction approximately perpendicular to the direction of the light meter. The volume can be determined in the image processing means itself or in a control and / or evaluation computer.

Typical areas of application of the presented method are:

Protection of machines

Control of access systems (e.g. doors)

Measurement of flows of people, counts (department stores, traffic facilities, public transport (tram, train, bus))

- Electronic vestibule monitoring of (autonomous) vehicles, with the advantage over mechanical systems that the vehicle is stopped before it mechanically touches the object (e.g. human).

Protection of objects from being approached by a human or a machine, or the detection of unauthorized access, e.g. the protection of a painting in a museum. Monitor an area that is divided into several contiguous or non-contiguous zones. A different alarm signal can be triggered for each zone.

Claims

PATENT CLAIMS

1. Procedure for securing systems, whereby

- A light source (1) projects at least one light line onto a surface to be monitored (4) from an illumination direction, whereby a light pattern (2) is created

- The light pattern (2) is detected at least partially from a direction different from the direction of illumination by an image area sensor (3), whereby a camera signal is generated,

- The camera signal is then fed to an image processing device (6),

- And the image processing device (6) determines the deviation of the light pattern (2) from a reference.

2. The method according to claim 1, characterized in that a monochrome light source is used as the light source and an optical filter (9) for the frequency of the monochrome light source is used in front of the camera (3).

3. The method according to claim 1 or 2, characterized in that the light source is pulsed or modulated and the time of triggering the pulsation or the control of the modulation by a control computer (7) via a controller (10) of the light source (1) and that the image recording of the camera (3) is switched on only during this time.

Method according to one of the preceding claims, characterized in that an image is recorded and temporarily stored without switching on the light source (1) and this image is compared with an image illuminated by the light source (1) and this signal is passed on to the image evaluation (6) becomes.

5. The method according to any one of the preceding claims, characterized in that two image area sensors (3, 3 ') are used to record the image, of which a first sensor (3) records an image simultaneously with the pulsation of the light source and a second image area sensor (3 ') takes a reference image immediately before or immediately afterwards, that the image of at least one of the two image area sensors (3') is geometrically transformed, and that those of the two images are then compared, for example by forming a difference, and the result is used for further image evaluation.

6. The method according to any one of the preceding claims, characterized in that two or more different sensor systems are present, each with an image area sensor, which have different light sensitivity due to optical measures, for example a diaphragm, or electrical measures, for example electronic amplification

Method according to claim 6, characterized in that, depending on the situation, that sensor system with the more suitable signal statistics is used for further processing.

A method according to claim 6 or 7, characterized in that the two images are combined with different light sensitivities, for example by addition, and the result is used for further processing.

9. The method according to any one of the preceding claims, characterized in that the deviation of the light pattern (2) from the reference is quantitatively calculated, whereby conclusions can be drawn about the height of at least one object (0) located on the area to be monitored.

10. The method according to claim 9, characterized in that the movement of an object located in the monitored area (0) is also detected and the volume of the object is estimated on the basis of the height and the movement of the object (0).

11. The method according to any one of the preceding claims, characterized in that the image area sensor (3) detects data in a pixel format, that an average weighted over at least two pixels is formed for determining the deviation and that the position deviation of the light pattern (2) from the reference is determined in a sub-pixel approximation.

12. The method according to any one of the preceding claims, characterized in that a result determined by the image processing device (6) is fed to a control computer (7) and this generates a danger signal in the event of deviations from a predetermined standard.

13. The method according to claim 12, characterized in that depending on the type of deviation from the norm, different hazard signals are generated with different meanings.

14. The method according to claim 12 or 13, characterized in that in the event of an alarm, an image, for example captured by the image area sensor, is stored for documentation purposes.

15. The method according to any one of the preceding claims, characterized in that when the entire light pattern or parts thereof are not visible, a check of the system is triggered.

16. The method according to any one of the preceding claims, characterized in that several independent light sources are used to generate the light pattern.

17. The method according to any one of the preceding claims, characterized in that the light pattern (2) can only be detected in the presence of an object to be detected.

18. The method according to any one of the preceding claims, characterized in that the system is used to secure machines.

19. The method according to any one of claims 1 to 17, characterized in that the system for monitoring / controlling any access systems, such as doors, is used.

20. The method according to any one of claims 1 to 17, characterized in that it flows with people, e.g. in department stores or when entering trains or train stations.

21. The method according to any one of claims 1 to 17, characterized in that the anteroom of (autonomous) vehicles can be monitored.

22. The method according to any one of claims 1 to 17, characterized in that the access to objects is monitored.

23. Device for securing systems, comprising a light source with means for projecting at least one light line onto a surface to be monitored, an image surface sensor for monitoring the surface, an image processing device (6) connected to the image surface sensor (3) and with means for evaluating a the image area sensor and camera signal processed by the image processing device such that the deviation of a light pattern (2) formed by the at least one light line from a reference can be determined quantitatively.

24. The device according to claim 23, characterized by a further image area sensor (3 ') for forming a reference signal.

25. The apparatus of claim 23 or 24, characterized by a further sensor system with an image area sensor and with a different light sensitivity.

26. Device according to one of claims 23-25, comprising a control computer (7) connected to the image processing device (6), which is programmable in such a way that the device for carrying out the method according to one of claims 1 to 16 can be activated.

27. The device according to any one of claims 23-26, characterized by a further, simultaneously usable sensor system, for example. For the simultaneous detection of geometric dimensions in height and distance or for avoiding non-detectable areas caused by shadows.