CN113076880B - Dangerous article pattern recognition algorithm - Google Patents

Dangerous article pattern recognition algorithm Download PDF

Info

Publication number
CN113076880B
CN113076880B CN202110369262.5A CN202110369262A CN113076880B CN 113076880 B CN113076880 B CN 113076880B CN 202110369262 A CN202110369262 A CN 202110369262A CN 113076880 B CN113076880 B CN 113076880B
Authority
CN
China
Prior art keywords
image
detected
wavelet
dangerous
dangerous goods
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110369262.5A
Other languages
Chinese (zh)
Other versions
CN113076880A (en
Inventor
李元伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong Industry Technical College
Original Assignee
Guangdong Industry Technical College
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong Industry Technical College filed Critical Guangdong Industry Technical College
Priority to CN202110369262.5A priority Critical patent/CN113076880B/en
Publication of CN113076880A publication Critical patent/CN113076880A/en
Application granted granted Critical
Publication of CN113076880B publication Critical patent/CN113076880B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a dangerous article pattern recognition algorithm, which belongs to the technical field of electronics and information, and comprises the following specific steps: (1) X-ray detection; (2) image wavelet decomposition; (3) image wavelet reconstruction; (4) morphological filling; (5) image matching; (6) feature curve detection; the invention adopts X-ray image to detect and match the shape of the object with average effective atomic number more than 10, and matches the characteristic curve fitted by the high and low gray values of the object with average effective atomic number less than 10 with the characteristic curve of the standard dangerous goods; the dangerous goods can be detected more comprehensively by combining the two modes, so that the dangerous goods detection efficiency is improved, and the labor intensity of security check personnel is reduced.

Description

Dangerous article pattern recognition algorithm
Technical Field
The invention relates to the technical field of electronics and information, in particular to a dangerous goods pattern recognition algorithm.
Background
Through retrieval, chinese patent number CN105427712B discloses an automatic dangerous article identification device and method based on three-dimensional X-ray imaging, and the detection method of the X-ray detection system is single in detection means and low in detection accuracy; at present, security inspection devices such as magnetic needles, metal weapon detection doors, X-ray detectors and the like are mostly arranged in places such as airports, stations, government buildings, prisons and the like needing security inspection, and can find dangerous goods such as weapons, common explosives and the like, thereby playing an important role in security inspection work; however, the devices are limited by the prior technical conditions, are not satisfactory in the actual use process, and have high false alarm rate and omission rate; moreover, with the development of scientific technology, criminals and terrorists have gradually utilized high and new technologies to manufacture new weapons, explosives and the like; such as high-precision bombs, plastic explosives, and many drugs manufactured by using integrated circuit technology, the conventional detection means cannot be adopted; therefore, it becomes important to invent a dangerous article pattern recognition algorithm;
most of the existing dangerous goods identification methods adopt traditional safety inspection equipment to be matched with security inspection personnel for dangerous goods identification, although the traditional safety inspection equipment can find dangerous goods such as weapons and common explosives, the labor intensity of the security inspection personnel is easy to increase, the traditional safety inspection equipment is limited by the prior technical conditions, the high-precision bomb, plastic explosive, a plurality of drugs and the like manufactured by the integrated circuit technology cannot be identified rapidly, and the identification effect is not ideal for articles with low density such as kerosene, ethanol and the like; therefore, we propose a dangerous goods pattern recognition algorithm.
Disclosure of Invention
The invention aims to solve the defects in the prior art, and provides a dangerous article pattern recognition algorithm.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the dangerous article pattern recognition algorithm comprises the following specific steps:
(1) X-ray detection: obtaining an average effective atomic sequence value of an object image to be detected by an X-ray detector, if the value is more than or equal to 10, jumping to the step (2) to carry out image wavelet decomposition on the object image to be detected, and if the value is less than or equal to 10, jumping to the step (6) to carry out characteristic curve detection on the object image to be detected;
(2) Image wavelet decomposition: carrying out one-layer wavelet decomposition on an image of an object to be detected by adopting a wavelet decomposition method, wherein each layer of image is decomposed into a low-frequency wavelet coefficient matrix and a high-frequency wavelet coefficient matrix, the low-frequency coefficients represent approximate parts of the image, and the high-frequency wavelet coefficients comprise coefficients in the horizontal direction, the vertical direction and the diagonal direction;
(3) Image wavelet reconstruction: performing edge detection on an image approximate part, namely a low-frequency part by using a Canny algorithm to obtain a sub-image of the low-frequency part; simultaneously, the Roberts and Sobel algorithm is utilized to carry out edge detection on the horizontal direction, the vertical direction and the diagonal direction coefficients of the image, namely the high-frequency part, so as to obtain a sub-image of the high-frequency part; carrying out wavelet reconstruction on the sub-image of the low-frequency part and the sub-image of the high-frequency part to obtain an edge image;
(4) Morphological filling: morphological filling is carried out on the edge image by using a hole filling mode, so that an enhanced edge image to be detected is obtained;
(5) Image matching: constructing a dangerous article standard image, and matching the enhanced edge image to be detected with the dangerous article standard image by using a curvature contour corner method to obtain a dangerous article judgment result of the curvature corner;
(6) And (3) feature curve detection: and constructing a common dangerous goods characteristic curve, and simultaneously adopting an inorganic matter characteristic curve method to perform dangerous goods detection matching on the image of the object to be detected to obtain a dangerous goods judgment result of the curve.
Preferably, the image wavelet decomposition of step (2) requires a scale function in two dimensions
Figure BDA0003008657670000031
And a two-dimensional wavelet ψ H (x,y)、ψ V (x, y) and ψ D (x, y), which can be expressed as a one-dimensional scale function +.>
Figure BDA0003008657670000032
And the product of the corresponding wavelet functions of ψ, the following formula:
Figure BDA0003008657670000033
its two-dimensional separable direction-sensitive wavelet can be expressed as:
Figure BDA0003008657670000034
Figure BDA0003008657670000035
ψ D (x,y)＝ψ(x)ψ(y) (4)
wherein: psi phi type H 、ψ V Sum phi D Representing the image intensity variations in different directions, respectively.
Preferably, given the separable two-dimensional scale and wavelet function, the scale and wavelet basis function are defined as follows:
Figure BDA0003008657670000036
Figure BDA0003008657670000037
wherein: i= { H, V, D } represents wavelets in different directions;
the f (x, y) discrete wavelet transform of size mxn is then:
Figure BDA0003008657670000041
Figure BDA0003008657670000042
wherein: j (j) 0 For the initial dimensions of the device,
Figure BDA0003008657670000043
define the scale j 0 An approximation image of f (x, y); w (W) ψ (j, m, n) contains detail images in horizontal, vertical and diagonal directions; let j 0 =0, and select n=m=2 j J=0, 1,2, …, j-1 and m, n=0, 1,2, … 2 j -1。
Preferably, the wavelet reconstruction formula in step (3) is as follows:
Figure BDA0003008657670000044
preferably, the hole filling formula in the step (4) is as follows:
Figure BDA0003008657670000045
wherein: k=1, 2,3.
Preferably, the mathematical definition of the curvature in step (5) is as follows: taking an arc section from a point M on the smooth arc, wherein the length of the arc section is deltas, the corresponding tangential angle is deltaalpha, and defining the arc section;
wherein the average curvature over Δs is:
Figure BDA0003008657670000046
the curvature at point M is:
Figure BDA0003008657670000047
the calculation formula of the discrete curvature is as follows:
Figure BDA0003008657670000048
preferably, the curvature in the step (5) includes global and local curvatures of the image, and the specific calculation process is as follows:
s1: calculating the average curvature of the contour of each enhanced edge image to be detected;
s2: then taking the average curvature of the contour as a threshold value, judging the current contour point, and if the current curvature is smaller than the threshold value, judging the current contour point as a real contour angular point;
s3: after extracting the corner points of all the enhanced edge images to be detected, performing corner point matching on the enhanced edge images to be detected and the dangerous goods standard images;
when the corner points are matched, corner point feature descriptors are constructed, and the corner point feature descriptors are defined as follows: let P be 1 、P 2 And P 3 Is the adjacent three corner points on the contour curve S, the curvature of which is K respectively 1 、K 2 And K 3 ，P 1 To P 2 Is L 1 ，P 2 To P 3 Is L 2 ) P is then 2 The descriptor B of (2) is:
B＝(L 1 +L 2 )(K 3 -K 1 ) (14)。
preferably, the specific steps of image matching in the step (5) are as follows:
SS1: extracting a complete edge image with clear outline characteristics and real corner points of a dangerous standard image;
SS2: respectively solving descriptors B and B' of each enhanced edge image corner to be detected and dangerous standard image corner by using the method (14);
SS3: performing sequential difference operation on each corner descriptor in the enhanced edge image to be detected and all corner descriptors in the dangerous goods standard image, if the ratio of the first term to the second term of the difference vector is smaller than a threshold value, the threshold value is 0.5, matching the diagonal points of the two images, otherwise, not matching;
SS4: counting the matching similarity percentage of the corner points of the same target contour, and if the similarity is larger than a preset value, determining the object as a dangerous object.
Preferably, the specific process of detecting the characteristic curve in the step (6) is as follows:
SSS1: obtaining high-low energy gray values of an image of an object to be detected through an X-ray security inspection machine, and drawing a characteristic curve of the image of the object to be detected according to the high-low energy gray values;
SSS2: constructing a common dangerous goods characteristic curve and taking the characteristic curve as a standard for detecting dangerous goods;
SSS3: comparing the characteristic curve of the image of the object to be detected with the characteristic curve of the common dangerous goods; if the characteristic curve of the image of the object to be detected is close to the characteristic curve of the dangerous goods, judging that the object is the dangerous goods, otherwise, judging that the object is normal.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the invention, the average effective atomic number value of an object is obtained by an X-ray detector, the detection object is divided into two types, the shape detection and the matching are carried out on the object with the average effective atomic number larger than 10 by adopting an X-ray image, and the matching is carried out on the object with the average effective atomic number smaller than 10 by utilizing the characteristic curve fitted by the high and low gray values of the object and the characteristic curve of a standard dangerous object; the dangerous goods can be detected more comprehensively by combining the two modes, so that the dangerous goods detection efficiency is improved, and the labor intensity of security check personnel is reduced.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
FIG. 1 is a general flow chart of a dangerous goods pattern recognition algorithm according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention and to simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention.
Example 1
The dangerous article pattern recognition algorithm comprises the following specific steps:
(1) X-ray detection: obtaining an average effective atomic sequence value of an object image to be detected by an X-ray detector, if the value is more than or equal to 10, jumping to the step (2) to carry out image wavelet decomposition on the object image to be detected, and if the value is less than or equal to 10, jumping to the step (6) to carry out characteristic curve detection on the object image to be detected;
(2) Image wavelet decomposition: carrying out one-layer wavelet decomposition on an image of an object to be detected by adopting a wavelet decomposition method, wherein each layer of image is decomposed into a low-frequency wavelet coefficient matrix and a high-frequency wavelet coefficient matrix, the low-frequency coefficients represent approximate parts of the image, and the high-frequency wavelet coefficients comprise coefficients in the horizontal direction, the vertical direction and the diagonal direction;
(3) Image wavelet reconstruction: performing edge detection on an image approximate part, namely a low-frequency part by using a Canny algorithm to obtain a sub-image of the low-frequency part; simultaneously, the Roberts and Sobel algorithm is utilized to carry out edge detection on the horizontal direction, the vertical direction and the diagonal direction coefficients of the image, namely the high-frequency part, so as to obtain a sub-image of the high-frequency part; carrying out wavelet reconstruction on the sub-image of the low-frequency part and the sub-image of the high-frequency part to obtain an edge image;
(4) Morphological filling: morphological filling is carried out on the edge image by using a hole filling mode, so that an enhanced edge image to be detected is obtained;
(5) Image matching: constructing a dangerous article standard image, and matching the enhanced edge image to be detected with the dangerous article standard image by using a curvature contour corner method to obtain a dangerous article judgment result of the curvature corner;
(6) And (3) feature curve detection: and constructing a common dangerous goods characteristic curve, and simultaneously adopting an inorganic matter characteristic curve method to perform dangerous goods detection matching on the image of the object to be detected to obtain a dangerous goods judgment result of the curve.
Step (2) image wavelet decomposition in two-dimensional case, a scale function is required
Figure BDA0003008657670000081
And a two-dimensional wavelet ψ H (x,y)、ψ V (x, y) and ψ D (x, y), which can be expressed as a one-dimensional scale function +.>
Figure BDA0003008657670000082
And the product of the corresponding wavelet functions of ψ, the following formula:
Figure BDA0003008657670000083
its two-dimensional separable direction-sensitive wavelet can be expressed as:
Figure BDA0003008657670000084
/>
Figure BDA0003008657670000085
ψ D (x,y)＝ψ(x)ψ(y) (4)
wherein: psi phi type H 、ψ V Sum phi D Representing the image intensity variations in different directions, respectively.
Given the separable two-dimensional scale and wavelet function, the scale and wavelet basis functions are defined as follows:
Figure BDA0003008657670000086
Figure BDA0003008657670000087
wherein: i= { H, V, D } represents wavelets in different directions;
the f (x, y) discrete wavelet transform of size mxn is then:
Figure BDA0003008657670000091
Figure BDA0003008657670000092
wherein: j (j) 0 For the initial dimensions of the device,
Figure BDA0003008657670000093
define the scale j 0 An approximation image of f (x, y); w (W) ψ (j, m, n) contains detail images in horizontal, vertical and diagonal directions; let j 0 =0, and select n=m=2 j J=0, 1,2, …, j-1 and m, n=0, 1,2, … 2 j -1。
The wavelet reconstruction formula in the step (3) is as follows:
Figure BDA0003008657670000094
the hole filling formula in the step (4) is as follows:
Figure BDA0003008657670000095
wherein: k=1, 2,3.
The mathematical definition of the curvature of step (5) is as follows: taking an arc section from a point M on the smooth arc, wherein the length of the arc section is deltas, the corresponding tangential angle is deltaalpha, and defining the arc section;
wherein the average curvature over Δs is:
Figure BDA0003008657670000096
the curvature at point M is:
Figure BDA0003008657670000097
the calculation formula of the discrete curvature is as follows:
Figure BDA0003008657670000098
the curvature in the step (5) comprises the global curvature and the local curvature of the image, and the specific calculation process is as follows:
s1: calculating the average curvature of the contour of each enhanced edge image to be detected;
s2: then taking the average curvature of the contour as a threshold value, judging the current contour point, and if the current curvature is smaller than the threshold value, judging the current contour point as a real contour angular point;
s3: after extracting the corner points of all the enhanced edge images to be detected, performing corner point matching on the enhanced edge images to be detected and the dangerous goods standard images;
constructing corner feature descriptors during corner matching, wherein the corner feature descriptors are defined as: let P be 1 、P 2 And P 3 Is the adjacent three corner points on the contour curve S, the curvature of which is K respectively 1 、K 2 And K 3 ，P 1 To P 2 Is L 1 ，P 2 To P 3 Is L 2 ) P is then 2 The descriptor B of (2) is:
B＝(L 1 +L 2 )(K 3 -K 1 ) (14)。
the specific steps of the image matching in the step (5) are as follows:
SS1: extracting a complete edge image with clear outline characteristics and real corner points of a dangerous standard image;
SS2: respectively solving descriptors B and B' of each enhanced edge image corner to be detected and dangerous standard image corner by using the method (14);
SS3: performing sequential difference operation on each corner descriptor in the enhanced edge image to be detected and all corner descriptors in the dangerous goods standard image, if the ratio of the first term to the second term of the difference vector is smaller than a threshold value, the threshold value is 0.5, matching the diagonal points of the two images, otherwise, not matching;
SS4: counting the matching similarity percentage of the corner points of the same target contour, and if the similarity is larger than a preset value, determining the object as a dangerous object.
The specific process of the characteristic curve detection in the step (6) is as follows:
SSS1: obtaining high-low energy gray values of an image of an object to be detected through an X-ray security inspection machine, and drawing a characteristic curve of the image of the object to be detected according to the high-low energy gray values;
SSS2: constructing a common dangerous goods characteristic curve and taking the characteristic curve as a standard for detecting dangerous goods;
SSS3: comparing the characteristic curve of the image of the object to be detected with the characteristic curve of the common dangerous goods; if the characteristic curve of the image of the object to be detected is close to the characteristic curve of the dangerous goods, judging that the object is the dangerous goods, otherwise, judging that the object is normal.
In order to verify the image matching effect, an X-ray dangerous article image with the size of 256 multiplied by 256 is selected as a sample (the sample comprises a pistol, handcuffs and a grenade), matlab7.1 is adopted for simulation, and the specific data of the test are shown in the following table:
Figure BDA0003008657670000111
example 2
The dangerous article pattern recognition algorithm comprises the following specific steps:
(1) X-ray detection: obtaining an average effective atomic sequence value of an object image to be detected by an X-ray detector, if the value is more than or equal to 10, jumping to the step (2) to carry out image wavelet decomposition on the object image to be detected, and if the value is less than or equal to 10, jumping to the step (6) to carry out characteristic curve detection on the object image to be detected;
(2) Image wavelet decomposition: carrying out one-layer wavelet decomposition on an image of an object to be detected by adopting a wavelet decomposition method, wherein each layer of image is decomposed into a low-frequency wavelet coefficient matrix and a high-frequency wavelet coefficient matrix, the low-frequency coefficients represent approximate parts of the image, and the high-frequency wavelet coefficients comprise coefficients in the horizontal direction, the vertical direction and the diagonal direction;
(3) Image wavelet reconstruction: performing edge detection on an image approximate part, namely a low-frequency part by using a Canny algorithm to obtain a sub-image of the low-frequency part; simultaneously, the Roberts and Sobel algorithm is utilized to carry out edge detection on the horizontal direction, the vertical direction and the diagonal direction coefficients of the image, namely the high-frequency part, so as to obtain a sub-image of the high-frequency part; carrying out wavelet reconstruction on the sub-image of the low-frequency part and the sub-image of the high-frequency part to obtain an edge image;
(4) Morphological filling: morphological filling is carried out on the edge image by using a hole filling mode, so that an enhanced edge image to be detected is obtained;
(5) Image matching: constructing a dangerous article standard image, and matching the enhanced edge image to be detected with the dangerous article standard image by using a curvature contour corner method to obtain a dangerous article judgment result of the curvature corner;
(6) And (3) feature curve detection: constructing a common dangerous goods characteristic curve, and simultaneously adopting an inorganic substance characteristic curve method to perform dangerous goods detection matching on an image of an object to be detected to obtain a dangerous goods judgment result of the curve;
the procedure is as in example 1.
In order to utilize the verification characteristic curve to detect the effect of inorganic dangerous goods, firecrackers with different thicknesses are selected in a test, the firecrackers are detected through the established firecracker characteristic curve, and test detection data are specifically as follows:
Figure BDA0003008657670000121
Figure BDA0003008657670000131
example 3
The dangerous article pattern recognition algorithm comprises the following specific steps:
(1) X-ray detection: obtaining an average effective atomic sequence value of an object image to be detected by an X-ray detector, if the value is more than or equal to 10, jumping to the step (2) to carry out image wavelet decomposition on the object image to be detected, and if the value is less than or equal to 10, jumping to the step (6) to carry out characteristic curve detection on the object image to be detected;
(2) Image wavelet decomposition: carrying out one-layer wavelet decomposition on an image of an object to be detected by adopting a wavelet decomposition method, wherein each layer of image is decomposed into a low-frequency wavelet coefficient matrix and a high-frequency wavelet coefficient matrix, the low-frequency coefficients represent approximate parts of the image, and the high-frequency wavelet coefficients comprise coefficients in the horizontal direction, the vertical direction and the diagonal direction;
(3) Image wavelet reconstruction: performing edge detection on an image approximate part, namely a low-frequency part by using a Canny algorithm to obtain a sub-image of the low-frequency part; simultaneously, the Roberts and Sobel algorithm is utilized to carry out edge detection on the horizontal direction, the vertical direction and the diagonal direction coefficients of the image, namely the high-frequency part, so as to obtain a sub-image of the high-frequency part; carrying out wavelet reconstruction on the sub-image of the low-frequency part and the sub-image of the high-frequency part to obtain an edge image;
(4) Morphological filling: morphological filling is carried out on the edge image by using a hole filling mode, so that an enhanced edge image to be detected is obtained;
(5) Image matching: constructing a dangerous article standard image, and matching the enhanced edge image to be detected with the dangerous article standard image by using a curvature contour corner method to obtain a dangerous article judgment result of the curvature corner;
(6) And (3) feature curve detection: constructing a common dangerous goods characteristic curve, and simultaneously adopting an inorganic substance characteristic curve method to perform dangerous goods detection matching on an image of an object to be detected to obtain a dangerous goods judgment result of the curve;
the procedure is as in example 1.
In order to test the effectiveness of the invention, several articles including cutters, toy guns, bottled alcohol, gasoline and the like are selected as test samples, and are respectively and independently mixed with other articles, and a dangerous article shape image database (28) and common dangerous article characteristic curves (15) are established as dangerous article judgment standards, and a plurality of tests are carried out in different directions and positions, wherein the detection accuracy of different articles is specifically shown in the following table:
article and method for manufacturing the same Number of samples Accuracy rate of False detection rate Leak rate
Cutting tool 80 91.25％ 7.50％ 8.13％
Toy gun 80 87.50％ 5.20％ 7.50％
Gasoline 80 92.00％ 6.25％ 5.00％
Bottled alcohol 80 90.00％ 5.00％ 7.13％
As can be seen from the data tables of the embodiment 1, the embodiment 2 and the embodiment 3, the detection accuracy of the invention on dangerous goods is higher, and the false detection rate and the omission ratio are both below 10%, so that the invention can more comprehensively detect dangerous goods carried in the plums, improves the detection efficiency of the dangerous goods, and is beneficial to reducing the labor intensity of security check personnel.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.

Claims (7)

1. The dangerous article pattern recognition algorithm is characterized by comprising the following specific steps of:
(1) X-ray detection: obtaining an average effective atomic sequence value of an object image to be detected by an X-ray detector, if the value is more than or equal to 10, jumping to the step (2) to carry out image wavelet decomposition on the object image to be detected, and if the value is less than or equal to 10, jumping to the step (6) to carry out characteristic curve detection on the object image to be detected;
(2) Image wavelet decomposition: carrying out one-layer wavelet decomposition on an image of an object to be detected by adopting a wavelet decomposition method, wherein each layer of image is decomposed into a low-frequency wavelet coefficient matrix and a high-frequency wavelet coefficient matrix, the low-frequency coefficients represent approximate parts of the image, and the high-frequency wavelet coefficients comprise coefficients in the horizontal direction, the vertical direction and the diagonal direction;
(3) Image wavelet reconstruction: performing edge detection on an image approximate part, namely a low-frequency part by using a Canny algorithm to obtain a sub-image of the low-frequency part; simultaneously, the Roberts and Sobel algorithm is utilized to carry out edge detection on the horizontal direction, the vertical direction and the diagonal direction coefficients of the image, namely the high-frequency part, so as to obtain a sub-image of the high-frequency part; carrying out wavelet reconstruction on the sub-image of the low-frequency part and the sub-image of the high-frequency part to obtain an edge image;
(4) Morphological filling: morphological filling is carried out on the edge image by using a hole filling mode, so that an enhanced edge image to be detected is obtained;
(5) Image matching: constructing a dangerous article standard image, and matching the enhanced edge image to be detected with the dangerous article standard image by using a curvature contour corner method to obtain a dangerous article judgment result of the curvature corner;
(6) And (3) feature curve detection: constructing a common dangerous goods characteristic curve, and simultaneously adopting an inorganic substance characteristic curve method to perform dangerous goods detection matching on an image of an object to be detected to obtain a dangerous goods judgment result of the curve;
wherein the image wavelet decomposition of step (2) requires a scale function in two dimensions
Figure QLYQS_1
And a two-dimensional wavelet ψ H (x,y)、ψ V (x, y) and ψ D (x, y), which can be expressed as a one-dimensional scale function +.>
Figure QLYQS_2
And the product of the corresponding wavelet functions of ψ, the following formula:
Figure QLYQS_3
its two-dimensional separable direction-sensitive wavelet can be expressed as:
Figure QLYQS_4
Figure QLYQS_5
ψ D (x,y)＝ψ(x)ψ(y) (4)
wherein: psi phi type H 、ψ V Sum phi D Representing the image intensity variations in different directions, respectively;
after the separable two-dimensional scale and the wavelet function are given, defining a scale and a wavelet basis function, wherein the scale and the wavelet basis function are as follows:
Figure QLYQS_6
Figure QLYQS_7
wherein: i= { H, V, D } represents wavelets in different directions;
the f (x, y) discrete wavelet transform of size mxn is then:
Figure QLYQS_8
Figure QLYQS_9
wherein: j (j) 0 For the initial dimensions of the device,
Figure QLYQS_10
define the scale j 0 An approximation image of f (x, y); w (W) ψ (j, m, n) contains detail images in horizontal, vertical and diagonal directions; let j 0 =0, and select n=m=2 j J=0, 1,2, …, j-1 and m, n=0, 1,2, … 2 j -1。
2. The dangerous goods pattern recognition algorithm of claim 1, wherein the wavelet reconstruction formula of step (3) is as follows:
Figure QLYQS_11
3. the dangerous goods pattern recognition algorithm of claim 1, wherein the hole filling formula of step (4) is as follows:
Figure QLYQS_12
wherein: k=1, 2,3.
4. The dangerous object pattern recognition algorithm of claim 1, wherein the mathematical definition of the curvature of step (5) is as follows: taking an arc section from a point M on the smooth arc, wherein the length of the arc section is deltas, the corresponding tangential angle is deltaalpha, and defining the arc section;
wherein the average curvature over Δs is:
Figure QLYQS_13
the curvature at point M is:
Figure QLYQS_14
the calculation formula of the discrete curvature is as follows:
Figure QLYQS_15
5. the dangerous object pattern recognition algorithm according to claim 1, wherein the curvature in the step (5) comprises global and local curvatures of the image, and the specific calculation process is as follows:
s1: calculating the average curvature of the contour of each enhanced edge image to be detected;
s2: then taking the average curvature of the contour as a threshold value, judging the current contour point, and if the current curvature is smaller than the threshold value, judging the current contour point as a real contour angular point;
s3: after extracting the corner points of all the enhanced edge images to be detected, performing corner point matching on the enhanced edge images to be detected and the dangerous goods standard images;
when the corner points are matched, corner point feature descriptors are constructed, and the corner point feature descriptors are defined as follows: let P be 1 、P 2 And P 3 Is the adjacent three corner points on the contour curve S, the curvature of which is K respectively 1 、K 2 And K 3 ，P 1 To P 2 Is L 1 ，P 2 To P 3 Is L 2 ) P is then 2 The descriptor B of (2) is:
B＝(L 1 +L 2 )(K 3 -K 1 ) (14)。
6. the dangerous goods pattern recognition algorithm according to claim 1, wherein the specific steps of the image matching in the step (5) are as follows:
SS1: extracting a complete edge image with clear outline characteristics and real corner points of a dangerous standard image;
SS2: respectively solving descriptors B and B' of each enhanced edge image corner to be detected and dangerous standard image corner by using the method (14);
SS3: performing sequential difference operation on each corner descriptor in the enhanced edge image to be detected and all corner descriptors in the dangerous goods standard image, if the ratio of the first term to the second term of the difference vector is smaller than a threshold value, the threshold value is 0.5, matching the diagonal points of the two images, otherwise, not matching;
SS4: counting the matching similarity percentage of the corner points of the same target contour, and if the similarity is larger than a preset value, taking the object as a dangerous object.
7. The dangerous goods pattern recognition algorithm according to claim 1, wherein the specific process of feature curve detection in the step (6) is as follows:
SSS1: obtaining high-low energy gray values of an image of an object to be detected through an X-ray security inspection machine, and drawing a characteristic curve of the image of the object to be detected according to the high-low energy gray values;
SSS2: constructing a common dangerous goods characteristic curve and taking the characteristic curve as a standard for detecting dangerous goods;
SSS3: comparing the characteristic curve of the image of the object to be detected with the characteristic curve of the common dangerous goods; if the characteristic curve of the image of the object to be detected is close to the characteristic curve of the dangerous goods, judging that the object is the dangerous goods, otherwise, judging that the object is normal.
CN202110369262.5A 2021-04-06 2021-04-06 Dangerous article pattern recognition algorithm Active CN113076880B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110369262.5A CN113076880B (en) 2021-04-06 2021-04-06 Dangerous article pattern recognition algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110369262.5A CN113076880B (en) 2021-04-06 2021-04-06 Dangerous article pattern recognition algorithm

Publications (2)

Publication Number Publication Date
CN113076880A CN113076880A (en) 2021-07-06
CN113076880B true CN113076880B (en) 2023-06-06

Family

ID=76615151

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110369262.5A Active CN113076880B (en) 2021-04-06 2021-04-06 Dangerous article pattern recognition algorithm

Country Status (1)

Country Link
CN (1) CN113076880B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114332512B (en) * 2022-03-14 2022-05-31 南京正驰科技发展有限公司 Security check card port detection method and system based on big data

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103198483A (en) * 2013-04-07 2013-07-10 西安电子科技大学 Multiple time phase remote sensing image registration method based on edge and spectral reflectivity curve
CN110516531A (en) * 2019-07-11 2019-11-29 广东工业大学 A kind of recognition methods of the dangerous mark based on template matching
CN111091111A (en) * 2019-12-25 2020-05-01 河北工业大学 Vehicle bottom dangerous target identification method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103198483A (en) * 2013-04-07 2013-07-10 西安电子科技大学 Multiple time phase remote sensing image registration method based on edge and spectral reflectivity curve
CN110516531A (en) * 2019-07-11 2019-11-29 广东工业大学 A kind of recognition methods of the dangerous mark based on template matching
CN111091111A (en) * 2019-12-25 2020-05-01 河北工业大学 Vehicle bottom dangerous target identification method

Also Published As

Publication number Publication date
CN113076880A (en) 2021-07-06

Similar Documents

Publication Publication Date Title
CN108510467B (en) SAR image target identification method based on depth deformable convolution neural network
CN110516561B (en) SAR image target identification method based on DCGAN and CNN
CN103761748B (en) Anomaly detection method and device
CN106778595B (en) Method for detecting abnormal behaviors in crowd based on Gaussian mixture model
CN110533051B (en) Automatic detection method for contraband in X-ray security inspection image based on convolutional neural network
CN111325748A (en) Infrared thermal image nondestructive testing method based on convolutional neural network
CN108717069B (en) High-pressure container thermal imaging defect detection method based on line variable step length segmentation
CN112712093B (en) Security check image identification method and device, electronic equipment and storage medium
CN110619352A (en) Typical infrared target classification method based on deep convolutional neural network
CN113076880B (en) Dangerous article pattern recognition algorithm
CN114299011A (en) Remote sensing target quadrilateral frame rapid detection method based on deep learning
JP5023238B2 (en) How to select the best trace area for shell case-based automatic region segmentation and shell case comparison
Chumuang et al. Analysis of X-ray for locating the weapon in the vehicle by using scale-invariant features transform
WO2008142675A1 (en) A method and a system for organizing an image database
Chu et al. Selecting valid correlation areas for automated bullet identification system based on striation detection
Mirdanies et al. Object recognition system in remote controlled weapon station using SIFT and SURF methods
Hsu et al. Comparative analysis of five neural network models
CN104809444B (en) The half ridge anisotropic filter Method of IR Small Target Fusion Recognition based on tensor extension
CN108931572B (en) Pressure container thermal imaging defect detection method based on variable line segmentation and region growing method
Hassaïne et al. An online signature verification system for forgery and disguise detection
CN106228122A (en) planetary surface feature matching method based on set similarity
Kong et al. A firearm identification system based on neural network
CN111914910A (en) Airplane formation identification method based on multi-observation-point shape context
CN111738288A (en) Bullet warhead fast matching method based on single-point laser detection
Dai et al. Smoke detection in infrared images based on superpixel segmentation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant