CN102565794B - Microwave security inspection system for automatically detecting dangerous object hidden in human body - Google Patents

Abstract

The invention relates to a microwave security inspection system for automatic detection of hidden dangerous objects in the human body, which includes a microwave transmitting and receiving device, a data acquisition device, an imaging processing device and an automatic detection device for hidden dangerous objects in the human body, wherein the microwave transmitting and receiving device, data The acquisition device, imaging processing device and automatic detection device for hidden dangerous objects in the human body are connected in sequence; the system has a high detection accuracy, meets the requirements of use, and solves the problem of automatic detection of hidden dangerous objects in the human body during the security inspection process.

Description

A microwave security inspection system for automatic detection of hidden dangerous objects in the human body

technical field

The invention relates to a security inspection system, in particular to a microwave security inspection system for automatically detecting hidden dangerous objects in a human body, and belongs to the technical field of image processing and security inspection.

Background technique

Microwaves have certain penetration in the process of propagation. Through microwave imaging technology, it is possible to obtain imaging result images of the scanned human body that are not affected by clothing, etc.; and then automatically complete the detection of dangerous objects hidden under the detected person's clothing based on the microwave results, such as metal knives, unknown liquids, sharp objects, etc. Objects, etc., is a fast, safe and effective security technology means. If in the detection process, the detection of suspicious and dangerous objects is completed directly through the naked eye observation of the security inspection operator, it will take a lot of manpower, financial resources and time. Therefore, it is of great significance to design a microwave security inspection system that can automatically detect hidden dangerous objects of different attributes and types that may exist on the human body.

Existing dangerous object detection systems: On the one hand, the automatic detection method of dangerous objects in the system is mainly based on visible light image data; The detection feature used by the CFAR method is a real-valued amplitude feature, that is, a gray feature. However, in the application of the applicant's microwave security inspection system for automatic detection of hidden dangerous objects, the above method is not applicable. The main reasons are as follows: 1) The imaging mechanism of microwave images and visible light images is essentially different, and the gray level of microwave images is low and the definition is low. , and is affected by coherent speckle noise. Object detection methods in visible light images cannot be directly applied in microwave images; 2) There are many forms of hidden dangerous objects involved in this application, and grayscale is only one of the possible detection features. Therefore, based on grayscale The CFAR method of detecting features cannot meet the needs of automatic detection of dangerous objects in microwave security inspection systems.

To sum up, in order to realize the automatic detection of hidden dangerous objects in the human body, it is necessary to construct a system that can distinguish dangerous objects in different situations according to the performance characteristics of dangerous objects in different materials, shapes, and sizes of interest in microwave imaging result images. The detection characteristics of the object, the design of an effective microwave security inspection system for the automatic detection of hidden dangerous objects in the human body.

Contents of the invention

The purpose of the present invention is to provide a microwave security inspection system with the function of automatic detection of hidden dangerous objects in the human body, which mainly includes the following components:

A microwave transmitting and receiving device, a data acquisition device, an imaging processing device, and an automatic detection device for hidden dangerous objects in the human body, wherein the microwave transmitting and receiving device, a data acquisition device, an imaging processing device, and an automatic detection device for hidden dangerous objects in the human body are connected in sequence , and the automatic detection device for hidden dangerous objects in the human body includes the following devices: an image input device, a local prominent brightness area feature extraction device, a local grayscale variance feature analysis device, a vertical edge feature extraction device, a detection feature fusion device, and a dangerous object Area detection device, detection result filtering device, and dangerous object area identification device; the image input device is respectively connected with the local prominent brightness area feature extraction device, the local gray variance feature analysis device, and the vertical edge feature extraction device, and the local The significant brightness area feature extraction device, the local gray variance feature analysis device, and the vertical direction edge feature extraction device are respectively connected to the detection feature fusion device, the detection feature fusion device, the dangerous object area detection device, the detection result filtering device, the dangerous object The area identification devices are connected sequentially.

Preferably, the local salient brightness region feature extraction device executes the following process:

1) Set the number of segmentation iterations used to extract local prominent brightness regions as 2, the first segmentation will separate the human body area from the background in the image, and the grayscale segmentation threshold of the human body area is T _b ; the second iteration segmentation will correspond to The local significant brightness area of the dangerous object is segmented from the human body area, and the gray-scale segmentation threshold of the dangerous object area is T _d ;

2) Combining with the gray-scale segmentation threshold T _d of the dangerous object area, construct the characteristic matrix G of the local prominent brightness area: extract the pixel area whose gray value is higher than T _d in the input microwave gray-scale image I, and set the rest positions to 0:

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

The number of rows and columns of the matrix G is equal to the number of rows and columns of the microwave grayscale image I, respectively.

Preferably, the local salient brightness region feature extraction device performs the following sub-processes in process 1):

1-1) The first iterative segmentation: first construct a global grayscale histogram H _g for the input microwave grayscale image I, and the grayscale range is [0, 255];

1-2) The first iterative segmentation: using the automatic threshold segmentation algorithm to calculate the gray-scale segmentation threshold T _b of the human body region in the global gray-scale histogram H _g ;

1-3) The second iterative segmentation: Combining the obtained human body region grayscale segmentation threshold T _b , construct the human body region grayscale histogram H _b , the grayscale range [T _b , 255];

1-4) The second iterative segmentation: an automatic threshold segmentation algorithm is used to calculate the gray level segmentation threshold T _d of the dangerous object area in the gray level histogram H _b of the human body area.

Preferably, the local gray variance feature analysis device performs the following process:

1) Set the window scale of the local gray variance analysis to s, where s is a natural number;

2) Calculate any point (x, y) in the microwave grayscale image I, the grayscale variance value in the square area L _{(x, y)} whose neighborhood size is s × s:

${\mathrm{<span\; class="notranslate">\; var</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>\frac{\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; L</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">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; L</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>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

E(L _{(x, y)} ) represents the gray value of the local neighborhood L _{(x, y) at the point (x, y} );

3) Construct the local gray variance feature matrix V corresponding to the microwave gray image I by combining the image local gray variance analysis results:

V(x,y)=var _(x,y)

The number of rows and columns of matrix V is equal to the number of rows and columns of microwave grayscale image I respectively;

4) Normalize the obtained local gray variance feature matrix V to obtain the local gray variance normalized feature matrix That is, the eigenvalues in V are linearly mapped to the range [0, 255]:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; \&times;</span>\frac{\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span>}$

Among them, the functions max(V) and min(V) represent the calculation of the maximum and minimum values in the local gray variance feature matrix V, respectively.

Preferably, the device for extracting edge features in the vertical direction performs the following process:

1) Construct the horizontal edge extraction operator h _h , the diagonal edge extraction operator h _s , the anti-diagonal edge extraction operator h _as , and the vertical edge extraction operator h _v :

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$ ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right]$

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\end{array}\right]$ ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

2) For any point (x, y) in the microwave grayscale image I, calculate the edge detection response in the horizontal direction e _h , the edge detection response in the diagonal direction e _s , the edge detection response in the anti-diagonal direction e _as , Vertical direction edge detection response e _v :

${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}$ ${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$

${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}$ ${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}$

3) Combining with the edge detection response settlement results, calculate the edge feature e (x, y) in the vertical direction at the point _{(x, y)} :

e _{(x, y)} = |e _h |+|e _s |+|e _as |-|e _v |

Among them, |.| is the absolute value calculation function;

4) Combining the calculation results of removing the edge features in the vertical direction at each point, construct the feature matrix E for removing the edge features in the vertical direction corresponding to the microwave grayscale image I:

E(x,y)=e _(x,y)

The number of rows and columns of matrix E is equal to the number of rows and columns of microwave grayscale image I respectively;

5) Normalize the obtained edge feature matrix E in the vertical direction to obtain a normalized matrix for edge feature removal in the vertical direction That is, the eigenvalues in E are linearly mapped to the range [0, 255]:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; E.</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; \&times;</span>\frac{\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span>}$

Among them, the functions max(E) and min(E) represent the calculation of the maximum and minimum values in the vertical edge feature matrix E, respectively.

通过加权归一化融合，获得二维融合特征矩阵F(x，y)：Preferably, the detection feature fusion device combines the obtained local prominent brightness region feature matrix G, the local gray level variance normalized feature matrix Normalize matrix to remove vertical edge features

Through weighted and normalized fusion, the two-dimensional fusion feature matrix F(x, y) is obtained:

$f(x,\mathrm{the\; y})=\mathrm{\&alpha;}\&times;G(x,\mathrm{the\; y})+\mathrm{\&beta;}\&times;\stackrel{\&OverBar;}{V}(x,\mathrm{the\; y})+\mathrm{\&gamma;}\&times;\stackrel{\&OverBar;}{\mathrm{E.}}(x,\mathrm{the\; y}),$ Satisfy α+β+γ=1

Among them, the weighting coefficients α, β, and γ are real numbers.

Preferably, the dangerous object area detection device performs the following procedures:

1) Combining with the two-dimensional fusion feature matrix F(x, y), set the fusion feature segmentation threshold T _f , where T _f is a real number;

2) Set the region where the feature value in the two-dimensional fusion feature matrix F(x, y) is greater than the fusion feature segmentation threshold T _f is the region of the rough dangerous object detection result. The value of this region is set to 255, and the values at other positions are set to 0.

Preferably, the detection result filtering device executes the following process:

1) Perform morphological closing operation on the obtained rough dangerous object detection result area;

2) Filter out the sub-regions whose area is smaller than T _area in the rough dangerous object detection result region, and the remaining sub-regions are the dangerous object detection result region, and T _area is a natural number.

Preferably, the hazardous object area identification device performs the following procedures:

1) Combine the obtained dangerous object detection result area to extract the edge contour of the dangerous object detection result area;

2) Mark the extracted edge contour in the input microwave grayscale image I.

The beneficial effects of the invention are: the system can be effectively used in real security inspection applications, has high detection accuracy, meets the requirements of use, and solves the problem of automatic detection of hidden dangerous objects in the human body during the security inspection process.

Description of drawings

Fig. 1 is a schematic structural diagram of a microwave security inspection system for automatic detection of hidden dangerous objects in the human body according to the present invention;

Fig. 2 is a schematic structural view of the automatic detection device for hidden dangerous objects in the human body in Fig. 1;

Fig. 3 is a comparison diagram before and after using a microwave security inspection system for automatic detection of hidden dangerous objects in the human body according to the present invention.

Detailed ways

Principle of invention

In a microwave security inspection system (such as a near-field millimeter-wave imaging security inspection system), after microwave signal transmission, reception, and imaging of the received signal, a microwave grayscale image is formed. The automatic detection process of hidden dangerous objects under occlusion presents the automatic detection results to the system operator. The invention is used for the problem of automatic detection of hidden dangerous objects in the human body during the security inspection process.

In the microwave security inspection system, after the imaging processing of the microwave echo signal is completed, it is necessary to automatically detect the hidden dangerous objects covered by the detected human clothes in the image results, so as to assist the system operator to make a judgment and complete the security inspection process. Usually, the dangerous objects of interest have a variety of attributes: in terms of materials, they mainly include liquid, metal, plastic, powder, etc.; in terms of shapes, they mainly include: long strips, acute angles, short widths, etc. On the one hand, the visible light image object detection system is not applicable to the microwave imaging result data; on the other hand, the existing security inspection system is difficult to meet the object detection requirements in different situations, as introduced in the background technology section.

In order to solve the above problems, the present invention provides an accurate and effective microwave security inspection system for automatic detection of hidden dangerous objects in the human body.

The inventors of the present invention have found the following technical characteristics through a large number of experiments: interested dangerous objects made of metal, liquid and other materials appear as local areas with higher gray values in microwave grayscale images; The dangerous object of interest at the boundary does not necessarily have a high gray value in the microwave grayscale image, but there is a more obvious grayscale change within its own area or relative to the background area, which is manifested as a larger local grayscale Variance; and for objects with special contour shapes, such as long strips and sharp angles, they appear as more obvious edge structure features in the imaging results.

The present invention utilizes this technical feature to extract three different detection features of dangerous objects from the input microwave grayscale image in the automatic detection device for hidden dangerous objects in the human body. Different detection features are used to distinguish and detect the dangerous objects of interest in different situations object. Among them, the local significant brightness area feature is used to detect the dangerous objects of interest that are clearly distinguished in the gray scale brightness, such as metal, liquid material objects, etc.; the local gray variance feature is used to detect the presence of obvious gray areas in the gray scale Dangerous objects of interest, such as plastic objects, powder objects, mixed material objects, human body area boundary objects, etc.; removing edge features in the vertical direction has two functions. On the one hand, edge features can detect typical rigid structural objects or Dangerous objects of interest with obvious shapes and structures, such as bricks, knives, etc., on the other hand, can remove a large amount of vertical edge information caused by human bones in the imaging results, reducing false alarms that may be caused during the detection process. After reasonable scaling, three different dangerous object detection features are weighted and normalized for fusion. Finally, the automatic detection of human hidden dangerous objects in different situations is completed through fusion features.

A microwave security inspection system for automatic detection of hidden dangerous objects in the human body according to the present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

As shown in Figure 1, the microwave security inspection system for automatic detection of hidden dangerous objects in the human body of the present invention mainly includes: a microwave transmitting and receiving device, a data acquisition device, an imaging processing device and an automatic detection device for hidden dangerous objects in the human body, wherein the microwave transmitting and receiving device , a data acquisition device, an imaging processing device and an automatic detection device for hidden dangerous objects in the human body are connected in sequence.

Microwave transmitting and receiving devices are used to generate, transmit and receive electromagnetic waves; data acquisition devices are used to convert analog signals into digital signals; imaging processing devices are used to convert echo digital signals into image signals; In the image signal, the hidden dangerous objects carried under the cover of human clothing are automatically detected and marked.

The microwave security inspection system transmits microwaves to the detected human body through the microwave transmitting device; the microwave receiving device receives the echo signal reflected by the human body, and converts the signal from analog form to digital form through the data acquisition device for processing; next, the digital signal is transmitted to the imaging The processing device, the imaging processing device obtains a high-resolution two-dimensional image signal through synthetic aperture processing; then, the automatic detection device for hidden dangerous objects on the human body uses the corresponding processing module to automatically detect the hidden dangerous objects carried under the cover of human clothing, and mark it; finally, transmit the result image of the automatic detection of hidden dangerous objects on the human body to the computer display device outside the system, and present it to the system operator and the detected person.

The core of the present invention-the automatic detection device for hidden dangerous objects in the human body will be described in detail below.

As shown in Figure 2, the automatic detection device for hidden dangerous objects in the human body mainly includes the following components: image input device, local prominent brightness area feature extraction device, local gray variance feature analysis device, vertical edge feature extraction device, Detection feature fusion device, dangerous object area detection device, detection result filtering device, dangerous object area identification device; the image input device is respectively connected with the local significant brightness area feature extraction device, the local gray variance feature analysis device, and the vertical direction edge feature extraction device Connection, the local significant brightness area feature extraction device, the local gray variance feature analysis device, and the vertical direction edge feature extraction device are respectively connected with the detection feature fusion device, the detection feature fusion device, the dangerous object area detection device, the detection result filtering device, the dangerous The object area identification devices are connected sequentially.

The automatic detection device for hidden dangerous objects in the human body can be realized in the form of hardware (such as DSP, PC) combined with software (such as C/C++ programming) as a whole, and can also be realized independently by the molecular device in the form of hardware combined with software, and then through Combination between sub-devices is complete.

The local prominent brightness area feature extraction device is used to extract local prominent brightness area features from the microwave grayscale image I, and it performs the following functional process:

1) Set the number of segmentation iterations used to extract local prominent brightness regions as 2, the first segmentation will separate the human body area from the background in the image, and the grayscale segmentation threshold of the human body area is T _b ; the second iteration segmentation will correspond to The local significant brightness area of the dangerous object is segmented from the human body area, and the gray-scale segmentation threshold of the dangerous object area is T _d ;

1-1) The first iterative segmentation: first construct a global grayscale histogram _Hg for the input microwave grayscale image I (as shown in the sub-image (a) in Figure 3), and the grayscale range is [0, 255];

1-2) The first iterative segmentation: using the automatic threshold segmentation algorithm to calculate the gray-scale segmentation threshold T _b of the human body region in the global gray-scale histogram H _g ;

1-3) The second iterative segmentation: Combining the obtained human body region grayscale segmentation threshold T _b , construct the human body region grayscale histogram H _b , the grayscale range [T _b , 255];

1-4) The second iterative segmentation: an automatic threshold segmentation algorithm is used to calculate the gray level segmentation threshold T _d of the dangerous object area in the gray level histogram H _b of the human body area.

2) Combining with the gray-scale segmentation threshold T _d of the dangerous object area, construct the characteristic matrix G of the local prominent brightness area: extract the pixel area whose gray value is higher than T _d in the input microwave gray-scale image I, and set the rest positions to 0:

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

The number of rows and columns of the matrix G is equal to the number of rows and columns of the microwave grayscale image I, respectively.

The local gray-scale variance feature analysis device is used to extract the local gray-scale variance feature matrix from the microwave gray-scale image I, and it performs the following functional process:

1) Set the window scale of the local gray variance analysis to s, where s is a natural number;

2) Calculate any point (x, y) in the microwave grayscale image I, the grayscale variance value in the square area L _{(x, y)} whose neighborhood size is s × s:

${\mathrm{<span\; class="notranslate">\; var</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>\frac{\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; L</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">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; L</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>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

E(L _{(x, y)} ) represents the gray value of the local neighborhood L _{(x, y) at the point (x, y} );

3) Construct the local gray variance feature matrix V corresponding to the microwave gray image I by combining the image local gray variance analysis results:

V(x,y)=var _(x,y)

The number of rows and columns of matrix V is equal to the number of rows and columns of microwave grayscale image I respectively;

即把V中的特征值按线性方式映射到值域[0，255]范围内：4) Normalize the obtained local gray variance feature matrix V to obtain the local gray variance normalized feature matrix

That is, the eigenvalues in V are linearly mapped to the range [0, 255]:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; \&times;</span>\frac{\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; )</span>}$

Among them, the functions max(V) and min(V) represent the calculation of the maximum and minimum values in the local gray variance feature matrix V, respectively.

The vertical direction edge feature extraction device is used to extract the vertical direction edge feature matrix to the microwave grayscale image I, and it performs the following process:

1) Construct the horizontal edge extraction operator h _h , the diagonal edge extraction operator h _s , the anti-diagonal edge extraction operator h _as , and the vertical edge extraction operator h _v :

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$ ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right]$

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\end{array}\right]$ ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 2</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

2) For any point (x, y) in the microwave grayscale image I, calculate the edge detection response in the horizontal direction e _h , the edge detection response in the diagonal direction e _s , the edge detection response in the anti-diagonal direction e _as , Vertical direction edge detection response e _v :

${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}}$ ${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$

${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; as</span>}}$ ${\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}$

3) Combining with the edge detection response settlement results, calculate the edge feature e (x, y) in the vertical direction at the point _{(x, y)} :

e _{(x, y)} = |e _h |+|e _s |+|e _as |-|e _v |

Among them, |.| is the absolute value calculation function;

4) Combining the calculation results of removing the edge features in the vertical direction at each point, construct the feature matrix E for removing the edge features in the vertical direction corresponding to the microwave grayscale image I:

E(x,y)=e _(x,y)

The number of rows and columns of matrix E is equal to the number of rows and columns of microwave grayscale image I respectively;

即把E中的特征值按线性方式映射到值域[0，255]范围内：5) Normalize the obtained edge feature matrix E in the vertical direction to obtain a normalized matrix for edge feature removal in the vertical direction

That is, the eigenvalues in E are linearly mapped to the range [0, 255]:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; E.</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; \&times;</span>\frac{\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span>}$

Among them, the functions max(E) and min(E) represent the calculation of the maximum and minimum values in the vertical edge feature matrix E, respectively.

去垂直方向边缘特征归一化矩阵

通过加权归一化融合，获得二维融合特征矩阵F(x，y)：The detection feature fusion device combines the obtained local prominent brightness region feature matrix G and the local gray level variance normalized feature matrix

Normalize matrix to remove vertical edge features

Through weighted and normalized fusion, the two-dimensional fusion feature matrix F(x, y) is obtained:

$f(x,\mathrm{the\; y})=\mathrm{\&alpha;}\&times;G(x,\mathrm{the\; y})+\mathrm{\&beta;}\&times;\stackrel{\&OverBar;}{V}(x,\mathrm{the\; y})+\mathrm{\&gamma;}\&times;\stackrel{\&OverBar;}{\mathrm{E.}}(x,\mathrm{the\; y}),$ Satisfy α+β+γ=1

Among them, the weighting coefficients α, β, and γ are real numbers.

The dangerous object area detection device is used to segment the fusion feature matrix F(x, y) and detect the dangerous object area, which performs the following functional process:

1) Combining with the two-dimensional fusion feature matrix F(x, y), set the fusion feature segmentation threshold T _f , where T _f is a real number;

2) Set the region where the feature value in the two-dimensional fusion feature matrix F(x, y) is greater than the fusion feature segmentation threshold T _f is the region of the rough dangerous object detection result. The value of this region is set to 255, and the values at other positions are set to 0.

The detection result filtering device is used to filter the detection results of the dangerous object area, and it performs the following functional process:

1) Perform morphological closing operation on the obtained rough dangerous object detection result area;

2) Filter out the sub-regions whose area is smaller than T _area in the rough dangerous object detection result region, and the remaining sub-regions are the dangerous object detection result region, and T _area is a natural number.

The dangerous object area identification device is used to identify the hazardous object area in combination with the filtered detection results, and it performs the following functional process:

1) Combine the obtained dangerous object detection result area to extract the edge contour of the dangerous object detection result area;

2) Mark the extracted edge contour in the input microwave grayscale image I, as shown in sub-image (b) in Fig. 3 .

Through experimental verification, a microwave security inspection system for automatic detection of hidden dangerous objects in the human body can be effectively used in real security inspection applications, has a high detection accuracy, meets the requirements of use, and solves the problem of hidden dangerous objects in the human body during the security inspection process. auto-detection issues.

The above embodiments are only for the purpose of clarifying the limitations of the present invention, and the actual protection scope of the present invention is not limited thereto, and all variations or changes made based on the idea of the present invention are within the protection scope of the present invention.

Claims (6)

1. a dangerous object hidden in human body detects microwave security inspection system automatically, it is characterized in that, comprise: the microwave transmitting and receiving device, data collector, imaging processing device and dangerous object hidden in human body automatic detection device, wherein, described microwave transmitting and receiving device, data collector, imaging processing device and dangerous object hidden in human body automatic detection device connect in turn, and described dangerous object hidden in human body automatic detection device comprises such as lower device: image-input device, local significantly luminance area feature deriving means, local gray level Variance feature analytical equipment, remove vertical direction Edge Gradient Feature device, the detected characteristics fusing device, danger body region pick-up unit, the testing result filtration unit, danger body region identity device; Described image-input device respectively with the remarkable luminance area feature deriving means in part, local gray level Variance feature analytical equipment, go vertical direction Edge Gradient Feature device to be connected, the remarkable luminance area feature deriving means in described part, local gray level Variance feature analytical equipment, go vertical direction Edge Gradient Feature device to be connected with the detected characteristics fusing device respectively, described detected characteristics fusing device, danger body region pick-up unit, testing result filtration unit, danger body region identity device connect in turn; The microwave transmitting and receiving device for generation of, emission, receive electromagnetic wave; It is digital signal that data collector is used for analog-signal transitions; The imaging processing device is used for changing the echo digital signal into picture signal; Dangerous object hidden in human body automatic detection device automatic human body clothing in picture signal blocks lower entrained hiding dangerous object, the line identifier of going forward side by side;

The remarkable luminance area feature deriving means in described part is carried out following flow process:

1) setting the iterations of cutting apart that is used for the local remarkable luminance area of extraction is 2, cuts apart for the first time human region in the image is split from background, and human region gray scale segmentation threshold is T _bThe local significantly luminance area that the second time, iteration was cut apart the dangerous object of correspondence splits from human region, and danger body region gray scale segmentation threshold is T _d

2) in conjunction with danger body region gray scale segmentation threshold T _d, make up local significantly luminance area eigenmatrix G: will input that gray-scale value is higher than T among the microwave gray level image I _dPixel region extract, all the other positions are set to 0:

$G(x,y)=\left\{\begin{array}{cc}I(x,y)& I(x,y)\&GreaterEqual;T_d\\ 0& I(x,y)<T_d\end{array}\right.$

The line number of matrix G, columns equate with the row, column number of microwave gray level image I respectively;

Described local gray level Variance feature analytical equipment is carried out following flow process:

1) setting local gray level variance analysis window size is s, and s is natural number;

2) any point (x, y) of calculating in microwave gray level image I faces the square area L that the territory range size is s * s _{(x, y}) interior gray variance value:

${\mathrm{var}}_{(x,y)}=\frac{\underset{1\&le;i\&le;s}{\mathrm{\&Sigma;}}\underset{1\&le;j\&le;s}{\mathrm{\&Sigma;}}{[L_{(x,y)}(i,j)-E\left(L_{(x,y)}\right)]}^2}{s^2}$

E (L _{(x, y)}) part located of representative point (x, y) faces territory L _{(x, y)}Gray average;

3) combining image local gray level the results of analysis of variance makes up local gray level Variance feature matrix V corresponding to microwave gray level image I:

V(x,y)=var _(x,y)

The line number of matrix V, columns equate with the row, column number of microwave gray level image I respectively;

4) the local gray level Variance feature matrix V of gained is carried out normalized, obtain local gray level variance normalization eigenmatrix

, namely the eigenwert among the V is mapped in codomain [0, the 255] scope by linear mode:

$\stackrel{\&OverBar;}{V}(x,y)=255\&times;\frac{V(x,y)-\min \left(V\right)}{\max \left(V\right)-\min \left(V\right)}$

Wherein, function max (V), min (V) represent respectively to calculate maximal value and the minimum value in the local gray level Variance feature matrix V;

The described vertical direction Edge Gradient Feature device that goes is carried out following flow process:

1) makes up horizontal direction arithmetic operators h _h, diagonal arithmetic operators h _s, back-diagonal direction arithmetic operators h _As, vertical direction arithmetic operators h _v:

$h_h=\left[\begin{array}{ccc}1& 2& 1\\ 0& 0& 0\\ -1& -2& -1\end{array}\right],$ $h_s=\left[\begin{array}{ccc}0& 1& 2\\ -1& 0& 1\\ -2& -1& 0\end{array}\right],$

$h_{\mathrm{as}}=\left[\begin{array}{ccc}2& 1& 0\\ 1& 0& -1\\ 0& -1& -2\end{array}\right],$ $h_v=\left[\begin{array}{ccc}-1& 0& 1\\ -2& 0& 2\\ -1& 0& 1\end{array}\right];$

2) to any point (x, y) among the microwave gray level image I, calculate this some place horizontal direction rim detection response e _h, diagonal rim detection response e _s, back-diagonal direction rim detection response e _As, vertical direction rim detection response e _v:

$e_h=I(x,y)\&CircleTimes;h_h,$ $e_s=I(x,y)\&CircleTimes;h_s,$

$e_{\mathrm{as}}=I(x,y)\&CircleTimes;h_{\mathrm{as}},$ $e_v=I(x,y)\&CircleTimes;h_v;$

3) jointing edge detects the response checkout result, and calculation level (x, y) is located removes vertical direction edge feature e _{(x, y)}:

e _(x,y)=|e _h|+|e _s|+|e _as|-|e _v|

Wherein, | .| is the computing function that takes absolute value;

4) in conjunction with the each point place go vertical direction edge feature result of calculation, make up microwave gray level image I corresponding remove vertical direction edge feature matrix E:

E(x,y)=e _(x,y)

The line number of matrix E, columns equate with the row, column number of microwave gray level image I respectively;

5) the vertical direction edge feature matrix E that goes of gained carried out normalized, obtain to go vertical direction edge feature normalization matrix

, namely the eigenwert among the E is mapped in codomain [0, the 255] scope by linear mode:

$\stackrel{\&OverBar;}{E}(x,y)=255\&times;\frac{E(x,y)-\min \left(E\right)}{\max \left(E\right)-\min \left(E\right)}$

Wherein, function max (E), min (E) represent respectively to calculate maximal value and the minimum value of going among the vertical direction edge feature matrix E.

2. a kind of dangerous object hidden in human body as claimed in claim 1 detects microwave security inspection system automatically, it is characterized in that, the remarkable luminance area feature deriving means in described part is in flow process 1) in carry out following sub-process:

1-1) iteration is cut apart for the first time: at first the microwave gray level image I of input made up overall grey level histogram H _g, tonal range [0,255];

1-2) iteration is cut apart for the first time: adopt the automatic threshold segmentation algorithm at overall grey level histogram H _gMiddle calculating human region gray scale segmentation threshold T _b

1-3) iteration is cut apart for the second time: in conjunction with acquired human region gray scale segmentation threshold T _b, make up human region grey level histogram H _b, tonal range [T _b, 255];

1-4) iteration is cut apart for the second time: adopt the automatic threshold segmentation algorithm at human region grey level histogram H _bMiddle calculating danger body region gray scale segmentation threshold T _d

3. a kind of dangerous object hidden in human body as claimed in claim 1 detects microwave security inspection system automatically, it is characterized in that, described detected characteristics fusing device is in conjunction with local significantly luminance area eigenmatrix G, the local gray level variance normalization eigenmatrix of gained

, remove vertical direction edge feature normalizing

Change matrix

Merge by weighting normalization, obtain two-dimentional fusion feature matrix F (x, y):

$F(x,y)=\mathrm{\&alpha;}\&times;G(x,y)+\mathrm{\&beta;}\stackrel{\&OverBar;}{V}(x,y)+\mathrm{\&gamma;}\&times;\stackrel{\&OverBar;}{E}(x,y),$ Satisfy alpha+beta+γ=1

Wherein, weighting coefficient α, β, γ are real number.

4. a kind of dangerous object hidden in human body as claimed in claim 1 detects microwave security inspection system automatically, it is characterized in that, described danger body region pick-up unit is carried out following flow process:

1) in conjunction with two-dimentional fusion feature matrix F (x, y), sets fusion feature segmentation threshold T _f, T _fBe real number;

2) set the middle eigenwert of two-dimentional fusion feature matrix F (x, y) greater than fusion feature segmentation threshold T _fThe zone be rough dangerous object detection results area, this regional value is set to 255, the value of all the other positions is set to 0.

5. a kind of dangerous object hidden in human body as claimed in claim 1 detects microwave security inspection system automatically, it is characterized in that, described testing result filtration unit is carried out following flow process:

1) the rough dangerous object detection results area that obtains is carried out closing operation of mathematical morphology;

2) in the rough dangerous object detection results area of filtering area less than T _AreaSubregion, remaining subregion is dangerous object detection results area, T _AreaBe natural number.

6. a kind of dangerous object hidden in human body as claimed in claim 1 detects microwave security inspection system automatically, it is characterized in that, described danger body region identity device is carried out following flow process:

1) in conjunction with the dangerous object detection results area that obtains, extracts the edge contour of dangerous object detection results area;

2) the extract edge contour that obtains is identified out in input microwave gray level image I.

Images