CN105069805A - Passive terahertz image segmentation method - Google Patents

Abstract

The invention discloses a passive terahertz image segmentation method, which comprises the following steps: step 1, obtaining a passive terahertz two-dimensional image of a human body; step two, performing denoising preprocessing on the two-dimensional image; step three, converting the The preprocessed image after denoising is divided into n regions equally, and the overall average gray level of the passive terahertz two-dimensional image is set to aver, wherein the average gray value of each area is aver _k , and the coordinates in a certain area k are (x , y) point’s gray value f _k (x, y), the area threshold of area k is T _k , where k∈[1,n], the validity threshold of this area is T _y , then according to the following formula (3 ), (4) and (5) to obtain the image, and step 4, determine whether there is a metal object outline in the obtained image, if there is a metal object outline, then determine that the subject is carrying a metal object, if there is no metal object outline, it is determined that the subject is not carrying metal objects.

Description

Passive Terahertz Image Segmentation Method

technical field

The invention belongs to the field of image processing and is used for target segmentation of passive terahertz images, in particular to a passive terahertz image segmentation method.

Background technique

Terahertz radiation generally refers to far-infrared electromagnetic radiation with a wavelength in the range of 3mm-3μm. In recent years, the application of passive terahertz imaging technology in the field of security inspection has shown impressive development prospects. Compared with traditional security inspection methods, passive hertz imaging technology has the following three advantages: First, in the terahertz band, many molecules There is a special absorption frequency above 1THz, so the use of terahertz spectroscopy can identify the type and composition of substances, and can be used for the detection of explosives, drugs, biochemicals and other dangerous goods in the security inspection process; secondly, terahertz radiation is sensitive to many nonmetallic objects , Non-polar substances, such as: clothing, cartons, plastics and other common shelters and packaging materials have strong penetrating power, and can directly detect dangerous goods wrapped or covered; again, the photon energy of terahertz radiation is low, Without the ionizing properties of X-rays, it will not cause damage to objects and personnel, and is very suitable for human body security inspection. Although passive terahertz imaging technology has shown great potential in the field of security inspection, due to the limitations of detector sensitivity, low-pass effect of antenna, and diffraction effect of imaging system, the resolution and signal-to-noise ratio of passive terahertz images are generally low. , which is not conducive to the rapid segmentation and recognition of the inspected items in the image, and also limits the development of this technology. Therefore, the development of a fast and high-precision passive terahertz image segmentation method is of great significance for improving the detection performance of terahertz imaging security devices.

Contents of the invention

An object of the present invention is to solve at least the above problems and/or disadvantages and to provide at least the advantages as will be described hereinafter.

Another object of the present invention is to provide a passive terahertz image segmentation method. In the present invention, the image is divided into multiple regions for processing separately, and combined with steps such as denoising and preprocessing, it solves the problem of the detection of objects in passive terahertz images. The problem of fast segmentation and recognition of human body regions under examination.

The technical scheme provided by the invention is:

A passive terahertz image segmentation method, comprising the following steps:

Step 1. Obtain a passive terahertz two-dimensional image of the subject;

Step 2, preprocessing: performing denoising preprocessing on the passive terahertz two-dimensional image obtained in step 1;

Step 3. Region segmentation: Divide the passive terahertz two-dimensional image after denoising and preprocessing in step two into n regions, and set the overall gray value of the passive terahertz two-dimensional image to be aver, where each region The average gray value of the area k is aver _k , the gray value of a point with coordinates (x, y) in a certain area k is f _k (x, y), and the area threshold of area k is T _k , where k∈[1,n ], the validity threshold of this region is T _y , and then according to the following formulas (3), (4) and (5), the points whose gray value is less than the threshold T _k are removed to obtain the image,

T _y =aver-r(3)

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ \mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Wherein, n is a positive integer, r=20; and, enter step 4 afterwards,

Step 4. Determine whether there is a metal object outline in the obtained image. If there is a metal object outline, it is determined that the subject is carrying a metal object. If there is no metal object outline, it is determined that the subject is not carrying a metal object. .

Preferably, in the passive terahertz image segmentation method, after the third step and before the fourth step, it also includes:

Step A, area correction: each information point in the image obtained in step 3 is calculated as follows: count the number N of information points in the area where the information point is located and in all areas adjacent to the area where the information point is located,

If N<4, remove all information points in the area where the information point is located and in all areas adjacent to the area described by the information point as noise points,

If N=5, it is determined that the 5 information points are all valid information points,

If N≥6, it is determined that there is a hole in the image in the area, and it is determined that all the information points in the area where the information point is located and in all areas adjacent to the area described by the information point are valid information points, and the hole is filled;

to obtain a region-corrected image.

Preferably, in the passive terahertz image segmentation method, after the step A and before the step four, it also includes:

Step B, performing filtering processing on the image obtained in step A.

Preferably, in the passive terahertz image segmentation method, in the first step, the specific process of obtaining the passive terahertz two-dimensional image of the subject includes: first scanning to obtain the passive terahertz one-dimensional image of the subject data, and then convert the one-dimensional data into a two-dimensional image.

Preferably, in the passive terahertz image segmentation method, in step 2, the specific process of performing denoising preprocessing on the passive terahertz two-dimensional image obtained in step 1 is as follows:

2.1) Set the width of the two-dimensional image as width, the height as height, the gray value of the point (x, y) is f(x, y), and the average gray value of the overall image is aver, then there is,

$\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; v</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

2.2) set the two-dimensional image threshold to be T, make T=θ·aver, where 0<θ<1, and remove points less than the two-dimensional image threshold T as noise points, as shown in formula (2):

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Preferably, in the passive terahertz image segmentation method, in the second step, θ=0.6.

Preferably, in the passive terahertz image segmentation method, in the step B, the filtering process is a two-way filtering process.

Preferably, in the passive terahertz image segmentation method, in the step B, the bidirectional filtering process is performed three times.

Preferably, in the passive terahertz image segmentation method, n=16 or 36

The present invention at least includes the following beneficial effects:

The method of the present invention first performs denoising preprocessing on the original terahertz image, and then performs region segmentation on it, and uses a bidirectional filtering method to perform region correction on the segmented image in view of its characteristics of gathering useful information points and dispersing noise points. The experimental results prove that the method of the present invention can quickly and effectively segment the human body area and the metal objects hidden on the human body from the passive terahertz image. to see the presence of metal objects. The method of the invention is fast and accurate, and is suitable for security inspection machines.

Other advantages, objectives and features of the present invention will partly be embodied through the following descriptions, and partly will be understood by those skilled in the art through the study and practice of the present invention.

Description of drawings

Fig. 1 is an original image of a passive terahertz two-dimensional image of a subject obtained in one embodiment of the present invention.

Fig. 2 is a result diagram of an image after denoising preprocessing in one embodiment of the present invention.

Fig. 3(a) is an effect diagram of single threshold segmentation when T=75 in one embodiment of the present invention.

Fig. 3(b) is an effect diagram of single threshold segmentation when T=130 in one embodiment of the present invention.

FIG. 4 is a result diagram of an image after region segmentation in one embodiment of the present invention.

FIG. 5 is a schematic diagram of a target correction 3×3 area in one embodiment of the present invention.

FIG. 6 is a schematic diagram of determining that noise points are eliminated when the number of information points in an area is less than 4 in one embodiment of the present invention.

Fig. 7 is a schematic diagram of determining valid information points when the number of information points in the area is 5 in one embodiment of the present invention.

Fig. 8 is a schematic diagram of filling holes when it is determined that there is a hole in the human body area when the information points in the area are greater than 6 in one embodiment of the present invention.

Fig. 9(a) is an original image of a passive terahertz image obtained after being processed by the method of the present invention in one embodiment of the present invention.

Fig. 9(b) is a display form of an image segmentation result map obtained after being processed by the method of the present invention in one embodiment of the present invention.

Fig. 10 is an original image of a passive terahertz two-dimensional image of a subject obtained in still another embodiment of the present invention.

Fig. 11 is a display form of the image segmentation results obtained after being processed by the method of the present invention in another embodiment of the present invention.

Fig. 12 is an original image of a passive terahertz two-dimensional image of a subject obtained in another embodiment of the present invention.

Fig. 13 is a display form of the image segmentation results obtained after being processed by the method of the present invention in another embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings, so that those skilled in the art can implement it with reference to the description.

It should be understood that terms such as "having", "comprising" and "including" as used herein do not entail the presence or addition of one or more other elements or combinations thereof.

The invention provides a passive terahertz image segmentation method, comprising the following steps:

Step 1. Obtain a passive terahertz two-dimensional image of the subject;

Step 2, preprocessing: performing denoising preprocessing on the passive terahertz two-dimensional image obtained in step 1;

Step 3. Region segmentation: Divide the passive terahertz two-dimensional image after denoising and preprocessing in step two into n regions, and set the overall gray value of the passive terahertz two-dimensional image to be aver, where each region The average gray value of the area k is aver _k , the gray value of a point with coordinates (x, y) in a certain area k is f _k (x, y), and the area threshold of area k is T _k , where k∈[1,n ], the validity threshold of this region is T _y , and then according to the following formulas (3), (4) and (5), the points whose gray value is less than the threshold T _k are removed to obtain the image,

T _y =aver-r(3)

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ \mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Wherein, n is a positive integer, r=20; and, enter step 4 afterwards,

Step 4. Determine whether there is a metal object outline in the obtained image. If there is a metal object outline, it is determined that the subject is carrying a metal object. If there is no metal object outline, it is determined that the subject is not carrying a metal object. .

Because there are still noise points in the image background after region segmentation, and there are holes caused by mis-segmentation in the human body region, usually the noise points in the segmented image are isolated and scattered in the image, while the useful information points are often aggregated. Aiming at this characteristic, the present invention proposes an area correction method. For example: in one of the embodiments of the present invention, as a preference, after the step three and before the step four, it also includes:

Step A, area correction: each information point in the image obtained in step 3 is calculated as follows: count the number N of information points in the area where the information point is located and in all areas adjacent to the area where the information point is located,

If N<4, remove all information points in the area where the information point is located and in all areas adjacent to the area described by the information point as noise points,

If N=5, it is determined that the 5 information points are all valid information points,

If N≥6, it is determined that there is a hole in the image in the area, and it is determined that all the information points in the area where the information point is located and in all areas adjacent to the area described by the information point are valid information points, and the hole is filled;

to obtain a region-corrected image.

In order to avoid the image offset caused by the above method, in one embodiment of the present invention, as a preference, after the step A and before the step 4, it also includes:

Step B, performing filtering processing on the image obtained in step A.

In the above solution, preferably, in the step B, the filtering process is a bidirectional filtering process.

In the above scheme, preferably, in the step B, the bidirectional filtering process is performed three times.

In one embodiment of the present invention, as a preference, in the first step, the specific process of obtaining the passive terahertz two-dimensional image of the subject includes: first scanning to obtain the passive terahertz one-dimensional data of the subject, Then the one-dimensional data is converted into a two-dimensional image.

In one embodiment of the present invention, as a preference, in the second step, the specific process of performing denoising preprocessing on the passive terahertz two-dimensional image obtained in the first step is as follows:

2.1) Set the width of the two-dimensional image as width, the height as height, the gray value of the point (x, y) is f(x, y), and the average gray value of the overall image is aver, then there is,

$\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; v</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

2.2) set the two-dimensional image threshold to be T, make T=θ·aver, where 0<θ<1, and remove points less than the two-dimensional image threshold T as noise points, as shown in the following formula (2):

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

In the above solution, according to empirical values, as a preference, in the second step, θ=0.6. In this way, the effect of the image obtained by denoising preprocessing is the best.

In some embodiments of the present invention, preferably, n=16 or 36.

In one embodiment of the present invention, a terahertz two-dimensional image of a human body is obtained as shown in FIG. 1 . Since the unprocessed terahertz image has a low signal-to-noise ratio, preprocessing of the image is required. In the passive terahertz image, the brightness of the human body area is relatively high, while the background and metal objects hidden in the human body area have low brightness. Therefore, the method of threshold segmentation can be used to preprocess the image to improve the signal-to-noise ratio of the image and increase the distance between the target and the background. Contrast:

Let the width of the image be width, the height be height, the gray value of the point (x, y) be f(x, y), and the average gray value of the image be aver, as shown in formula (1):

$\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; v</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Set the threshold as T, let T=θ·aver, where θ=0.6, and remove the points smaller than the threshold as noise points, as shown in formula (2):

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Figure 1 after preprocessing is shown in Figure 2. It can be seen that some noise points in the background have been removed.

In the process of segmenting passive terahertz images, the gray value of metal objects hidden in the human body area may be higher than the gray value of the edge of the human body. If the commonly used single-threshold segmentation method is used, it may be Because the threshold is too small, the metal object cannot be segmented or the threshold is too large, resulting in incomplete segmentation of the human body area, as shown in Figure 3(a) and Figure 3(b), Figure 3(a), when T=75, the calculation Finally, it can be seen that the metal objects in the human body area cannot be segmented, and Fig. 3(b) shows that when T=130, the human body area is incomplete after calculation.

In order to avoid this situation, the present invention adopts a region segmentation method, divides the image into 36 regions, and performs image segmentation processing on each region. Let the average gray value of the image as a whole be aver, the average gray value of each region be aver _k , the gray value of a point whose coordinates are (x, y) in region k is f _k (x, y), and the region threshold in region k be T _k , where k∈[1,36], set the validity threshold of the region as T _y , where T _y is shown in formula (3), if the average gray value of the region is less than the validity threshold, it is determined that there are fewer target points in the region, In order to avoid the introduction of noise, the threshold value of the area should be replaced, as shown in formula (4), and then the gray value of each point is judged. If the gray value of the point is less than the threshold value of the area where the point is located, it is determined that the point is noise removal. As shown in formula (5), where r=20:

T _y =aver-r(3)

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ \mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; ,</span>& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Figure 2 shows the result of area segmentation in Figure 4. It can be seen from the figure that using the area segmentation method can not only segment the outline of metal objects, but also ensure the integrity of the human body area to a large extent.

It can be seen from Figure 4 that there are still noise points in the image background after region segmentation, and there are holes caused by mis-segmentation in the human body region. Generally, the noise points in the segmented image are isolated and scattered in the image, while the useful information points are often aggregated. Aiming at this characteristic, the present invention proposes a region correction method.

Area correction method: firstly find the segmented information points in the image, after searching the information points, count the number of information points in the 3×3 area where the information point is located, if the area where the information point is located is located at the edge of the image, then take Information points in all adjacent areas are sufficient. Let the area where it is located be Tr={tr(x,y),x∈[0,2],y∈[0,2]}, where tr(x,y) is the coordinate (x,y) in the area The gray value of , as shown in Figure 5, white is the current information point, and the coordinates of this point are (x, y):

Assuming that the number of target points in the region Tr is N, there are the following three situations:

1. When N<4, remove all points in the 3×3 area as noise points, namely:

Tr={tr(x,y)=0, x∈[0,2],y∈[0,2]} (6)

As shown in Figure 6.

2. When N=5, the points in the 3×3 area are not processed, and they are all judged as valid information points, as shown in Figure 7:

3. When N≥6, all points in the 3×3 area are identified as valid information points, namely:

Tr={tr(x,y)=255, x∈[0,2],y∈[0,2]} (7)

As shown in Figure 8.

In order to avoid the image shift caused by this method, the image should be filtered from left to right first, and then from right to left to complete the filtering process in each direction, which is called bidirectional filtering. The two-way filtering algorithm can not only effectively remove isolated and scattered noise points, but also fill holes in the human body area caused by mis-segmentation. After performing two-way filtering on Figure 4 for three times, the results are shown in Figure 9(a) and Figure 9(b), where Figure 9(a) is the original image of the passive terahertz image obtained after processing, and Figure 9(b) is An easier-to-observe image obtained after image conversion of Figure 9(a).

In yet another embodiment of the present invention, the obtained terahertz two-dimensional image of the human body is shown in Figure 10, and the method steps adopted in the above embodiment are still used, except that the image is divided into 16 regions, and the rest Do the same processing, and finally get the result graph shown in Figure 11.

In yet another embodiment of the present invention, the obtained terahertz two-dimensional image of the human body is shown in Figure 12, and the method steps adopted in the above embodiment are still used, except that the image is divided into 16 regions, and the rest Do the same processing, and finally get the result graph shown in Figure 13.

It should be known that n is not limited to 16 or 36 listed here, and the user can divide the image into an appropriate number of regions according to the actual two-dimensional image.

The number of modules and processing scales described here are used to simplify the description of the present invention. Applications, modifications and changes to the two-dimensional image thresholds and area thresholds, validity thresholds, etc. of the present invention will be apparent to those skilled in the art.

As mentioned above, according to the present invention, since the passive terahertz two-dimensional image is obtained through methods such as area segmentation, it is possible to obtain clear outlines of the inspected human body and metal objects, improving the accuracy and speed of security inspection.

Although the embodiment of the present invention has been disclosed as above, it is not limited to the use listed in the specification and implementation, it can be applied to various fields suitable for the present invention, and it can be easily understood by those skilled in the art Therefore, the invention is not limited to the specific details and examples shown and described herein without departing from the general concept defined by the claims and their equivalents.

Claims (9)

1. a passive type Terahertz image partition method, is characterized in that, comprises the following steps:

Step one, obtain the passive type Terahertz two dimensional image of tested human body;

Step 2, pre-service: the passive type Terahertz two dimensional image obtained in step one is carried out noise suppression preprocessing;

Step 3, region segmentation: the passive type Terahertz two dimensional image after step 2 noise suppression preprocessing is divided into n region, setting described passive type Terahertz two dimensional image overall intensity average is aver, and wherein the gray average in each region is aver _k, in a certain region k, coordinate is the gray-scale value that (x, y) puts is f _k(x, y), the region threshold of region k is T _k, wherein k ∈ [1, n], this region validity threshold value is T _y, be less than threshold value T according to following formula (3), (4) and (5) computing removing gray-scale value afterwards _kpoint obtain image,

T _y＝aver-r(3)

Wherein, n is positive integer, r=20; And, enter step 4 afterwards,

Whether there is metal profile in the image that step 4, judgement obtain, if there is metal profile, then judging that described tested human body carries metal, if there is not metal profile, then judging that described tested human body does not carry metal.

2. passive type Terahertz image partition method as claimed in claim 1, is characterized in that, also comprise after described step 3 He before described step 4:

Steps A, region are corrected: calculate as follows each information point in the image obtained in step 3: the number N adding up the information point in this information point region and all regions adjacent with this information point region,

If N < 4, then all information points in this information point region and all regions adjacent with region described in this information point are removed as noise spot,

If N=5, then judge that these 5 information points are effective information point,

If N >=6, then judge that in this region, hole appears in image, judge that all information points in this information point region and all regions adjacent with region described in this information point are effective information point, fill described hole;

To obtain the image after region is corrected.

3. passive type Terahertz image partition method as claimed in claim 2, is characterized in that, also comprise after described steps A He before described step 4:

Step B, filtering process is carried out to the image obtained in steps A.

4. passive type Terahertz image partition method as claimed in claim 1, it is characterized in that, in described step one, the detailed process obtaining the passive type Terahertz two dimensional image of tested human body comprises: first scanning obtains the one-dimensional data of the passive type Terahertz of tested human body, more described one-dimensional data is converted into two dimensional image afterwards.

5. passive type Terahertz image partition method as claimed in claim 1, is characterized in that, in described step 2, by the detailed process that the passive type Terahertz two dimensional image obtained in step one carries out noise suppression preprocessing be:

2.1) setting X-Y scheme image width is width, and height is height, and the gray-scale value of point (x, y) is f (x, y), and general image gray average is aver, then have,

2.2) setting described two dimensional image threshold value is T, makes T=θ aver, and wherein 0 < θ < 1, removes the point being less than two dimensional image threshold value T as noise spot, shown in following formula (2):

。

6. passive type Terahertz image partition method as claimed in claim 5, is characterized in that, in described step 2, and θ=0.6.

7. passive type Terahertz image partition method as claimed in claim 3, it is characterized in that, in described step B, described filtering is treated to bidirectional filtering process.

8. passive type Terahertz image partition method as claimed in claim 7, it is characterized in that, in described step B, described bidirectional filtering process carries out 3 times.

9. passive type Terahertz image partition method as claimed in claim 1, is characterized in that, n=16 or 36.