CN117930375A - Multi-dimensional detection technology fused channel type terahertz human body security inspection system - Google Patents
Multi-dimensional detection technology fused channel type terahertz human body security inspection system Download PDFInfo
- Publication number
- CN117930375A CN117930375A CN202410329967.8A CN202410329967A CN117930375A CN 117930375 A CN117930375 A CN 117930375A CN 202410329967 A CN202410329967 A CN 202410329967A CN 117930375 A CN117930375 A CN 117930375A
- Authority
- CN
- China
- Prior art keywords
- terahertz
- images
- image
- visible light
- security inspection
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 88
- 238000007689 inspection Methods 0.000 title claims abstract description 50
- 238000005516 engineering process Methods 0.000 title claims abstract description 15
- 238000003384 imaging method Methods 0.000 claims abstract description 36
- 238000013135 deep learning Methods 0.000 claims abstract description 12
- 230000003287 optical effect Effects 0.000 claims abstract description 7
- 239000013598 vector Substances 0.000 claims description 22
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 238000000034 method Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 239000011358 absorbing material Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 108010066057 cabin-1 Proteins 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000012634 optical imaging Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 241000287107 Passer Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000005057 refrigeration Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 239000012780 transparent material Substances 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a channel type terahertz human body security inspection system fused with a multi-dimensional detection technology, which relates to the technical field of security inspection equipment, wherein two terahertz imaging hosts are respectively placed in two equipment cabins; the security inspection channel comprises two groups of reflectors, the two groups of reflectors are respectively matched with two terahertz imaging hosts to carry out terahertz imaging on the side surfaces of the personnel to form two side surface terahertz images, and the two terahertz imaging hosts capture terahertz waves emitted by the front surface and the back surface of the personnel to form front surface terahertz images and back surface terahertz images; the optical camera collects visible light information of personnel in real time to form visible light images on the front side and the back side; the central control display system utilizes a deep learning target detection model to detect suspicious articles on terahertz images and visible light images of all sides, and matches the front terahertz images and the back terahertz images with the front visible light images and the back visible light images respectively, so that personnel targets are locked, and the comprehensiveness and the reliability of channel terahertz detection are improved.
Description
Technical Field
The invention relates to the technical field of terahertz detection, in particular to a channel type terahertz human body security inspection system fused with a multi-dimensional detection technology.
Background
With the development of modern transportation industry, rapid transportation means such as automobiles, trains, ships, high-speed rail cars, airplanes and the like gradually become the transportation mode for people to travel and select. The trips of large population and the security inspection of luggage carried along with the trips are followed. A baggage item security check machine is arranged in places such as railway stations, subway stations, airports, customs and the like to check whether suspicious objects are hidden in the baggage items. The conventional safety detection equipment is mostly single-channel or double-channel safety inspection machines, luggage articles sequentially enter the equipment to be scanned, and after image display, a safety inspector judges whether dangerous articles exist in the safety inspected luggage articles according to the images. The inspection equipment is low in inspection speed, and congestion is often caused when passenger flow is large, so that the requirements on all aspects of security inspection are higher and higher in the technical field of security inspection equipment, and particularly in a large passenger flow state, the security inspection speed is required, and the security inspection is comprehensive.
At present, the terahertz human body security inspection system has the problem that the side cannot be detected, if a person is required to rotate for one circle, the security inspection efficiency can be greatly reduced, the security inspection experience of the person is poor, and the person is easy to collide with the security inspection person. If a metal detection module is installed on the side of a channel, there is a problem that non-metallic objects on the side cannot be detected.
Disclosure of Invention
The invention aims to provide a channel terahertz human body security inspection system fused with a multi-dimensional detection technology. The invention solves the technical problems through the following technical scheme, and the multi-dimensional detection technology-fused channel terahertz human body security inspection system comprises: the device comprises two equipment cabins, two terahertz imaging hosts and a central control display system;
the security inspection channel is positioned between the two equipment cabins, and the two terahertz imaging hosts are respectively placed in the two equipment cabins;
the security inspection channel comprises two groups of reflectors, the two groups of reflectors are respectively matched with two terahertz imaging hosts, when a pedestrian passes through the security inspection channel, terahertz imaging is carried out on the side face of the person to form two side face terahertz images, and the two terahertz imaging hosts capture terahertz waves emitted by the front face and the back face of the person at the same time to form front face and back face terahertz images;
an optical camera is arranged in the equipment cabin, and visible light information of personnel is collected in real time to form front and back visible light images;
The central control display system detects suspicious articles by utilizing a deep-learning target detection model to the terahertz images and the visible light images of all sides, and matches the terahertz images of the front side and the back side with the visible light images of the front side and the back side respectively to lock personnel targets.
Further, fusing the front terahertz image and the back terahertz image, and fusing the two side terahertz images, and detecting targets:
Respectively comparing the front terahertz image and the back terahertz image with the two side terahertz images, calculating the characteristic value difference between the images, and selecting two terahertz images with the minimum characteristic value difference;
Comparing the pre-stored images of the suspicious object with two terahertz images with minimum difference values of characteristic values respectively, and calculating the matching degree S ij according to the pixel positions (i, j):
Where V ij is the feature vector at the location of coordinate (i, j) in the image of the suspicious item, And taking the coordinate position in the terahertz image corresponding to the highest matching degree point as the detected suspicious article center point for the feature vector at the coordinate (m, n) position in the kth image in the two terahertz images.
Further, comparing the front terahertz image and the back terahertz image with the two side terahertz images respectively, and calculating the characteristic value difference between the images, including:
Selecting one reference feature point in the front terahertz image or the back terahertz image as a preferred point, and extracting feature vectors of the reference feature point;
searching a plurality of matching feature points matched with the reference feature points in the left terahertz chart or the right terahertz chart;
extracting a plurality of feature vectors corresponding to the plurality of matching feature points, calculating similarity values of the plurality of feature vectors and the feature vectors of the extracted reference feature points, and screening out one matching feature point which is most matched with the reference feature points;
Repeating the steps until a plurality of reference feature points and a plurality of matched feature points matched with the reference feature points are selected, wherein the reference feature points are connected to form a reference feature unit, and the matched feature points are connected to form a matched feature unit;
and calculating the characteristic value difference between the front terahertz image, the back terahertz image and the two side terahertz images, and selecting the two terahertz images with the minimum characteristic value difference.
Further, matching the front and back terahertz images with front and back visible light images, respectively, includes:
The front or back visible light image and the front or back terahertz image are subjected to low-pass filtering to obtain images A and B;
the gray level layers of the two images A and B are overlapped, so that the distribution relation between gray level value sets of the two images is better compared and analyzed, and the difference between different data sets is more clearly observed.
The gray map layer stacking function H ab is:
Hab[a(i,j),b(i,j)]=Nab；i=1,…，M；j=1,…，N；
Wherein a (i, j), B (i, j) are gray value pairs at positions (i, j) in images a and B with pixels M x N, N ab representing the number of occurrences of the same gray value pair;
the superimposed normalized value P AB (a, B) for images a and B is:
Calculating a coincidence information value M (a, B) of the images a and B using the superimposed normalized value P AB (a, B):
wherein, P A (a) and P B (B) are the independent gray scale distribution values of images A and B, respectively; the larger the superposition information value M (A, B), the higher the matching degree is; otherwise, the lower the matching degree is.
Further, inputting the terahertz image into a terahertz target detection model for suspicious object detection, and outputting a terahertz image detection result; and synchronously inputting the visible light images into a visible light detection model, and outputting clothing and face detection results.
Further, the terahertz imaging host has a total detection module, which includes: a positive detection module unit and a side detection module; the positive detection module unit accounts for 3/4 of the total detection module and is responsible for collecting terahertz waves emitted by the front and back surfaces of the personnel, the side detection module unit accounts for 1/4 of the total detection module and is responsible for collecting terahertz waves emitted by the left and right side surfaces of the personnel.
Further, when the terahertz image detection result is judged to be dangerous, matching the terahertz image with the visible light image, and meanwhile marking the matched visible light image detection result as dangerous to lock the target; and when the terahertz image detection result is judged to be safe, the matched visible light image detection result is marked to be safe, whether the number of terahertz detection objects is larger than the set number is continuously judged, if so, the danger is output, and if not, the safety is output.
Compared with the prior art, the invention has the beneficial effects that:
1. By means of the design of the reflecting mirror group of the security inspection channel and matching with the terahertz imaging host, terahertz imaging detection of the side face, the front face and the back face of a person when the security inspection channel passes is achieved, the problem that the side face of an existing channel terahertz human body security inspection system cannot be detected is solved, and the comprehensiveness and reliability of the channel terahertz detection are improved;
2. The terahertz imaging technology and the optical imaging technology can be simultaneously utilized to realize multi-angle and omnibearing detection of the target, and the accuracy and reliability of security inspection are improved. Meanwhile, by combining a deep learning technology, the target can be rapidly and accurately identified, so that manual intervention is reduced, and the security inspection efficiency is improved;
3. By the design of the integrated display control system, the high integration effect of the system is realized, and the occupied area of the system is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort to a person skilled in the art.
FIG. 1 is a schematic diagram of a channel terahertz human body security inspection system fused by a multi-dimensional detection technology;
FIG. 2 is a schematic flow chart of analyzing terahertz images and visible light images by the central control display system of the invention;
FIG. 3 is a schematic diagram of the comprehensive analysis flow of the security inspection system of the present invention.
The figures represent the numbers:
1. an equipment compartment; 2. a terahertz imaging host; 3. a mirror group; 4. a mirror assembly housing; 5. a light shielding plate; 6. a security inspection channel; 7. a central control display system; 8. a refrigeration system; 9. and an optical acquisition module.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In the drawings of the specific embodiments of the present invention, in order to better and more clearly describe the working principle of each element in the system, the connection relationship of each part in the device is represented, but only the relative positional relationship between each element is clearly distinguished, and the limitations on the signal transmission direction, connection sequence and the structure size, dimension and shape of each part in the element or structure cannot be constructed.
The structure of the channel terahertz human body security inspection system for providing multi-dimensional detection technology integration in this embodiment refers to fig. 1, and includes two equipment cabins 1, two terahertz imaging hosts 2 and a security inspection channel, the security inspection channel is located between the two equipment cabins 1, and the two terahertz imaging hosts 2 are placed in the two equipment cabins 1.
The terahertz imaging host is provided with a total detection module, and the total detection module comprises: a positive detection module unit and a side detection module; the positive detection module unit accounts for 3/4 of the total detection module and is responsible for collecting terahertz waves emitted by the front and back surfaces of the personnel, the side detection module unit accounts for 1/4 of the total detection module and is responsible for collecting terahertz waves emitted by the left and right side surfaces of the personnel.
The two equipment cabins are the equipment cabin at the inlet and the equipment cabin at the outlet respectively, the refrigerating system 8 is arranged in the two equipment cabins 1, the environment temperature and the humidity in the equipment cabins are controlled, the terahertz imaging host is ensured not to be influenced by the external environment temperature, and the terahertz imaging host works at a proper environment temperature, so that the ideal imaging quality is achieved.
The security inspection channel 6 is internally provided with a reflecting mirror group 3, a reflecting mirror group housing 4 and a shading plate 5.
The reflecting mirror group 3 cooperates with the terahertz imaging host 2 to carry out terahertz imaging detection on the side surface of the passers. The equipment cabin 1 at the outlet is integrated with a central control display system 7, the working state of the whole system is controlled, the terahertz image of the terahertz imaging host is received, whether suspicious articles exist in the image is judged by combining with an artificial intelligent deep learning algorithm, the terahertz image is displayed on a display screen in real time, the positions of the suspicious articles are marked, and a security inspector rechecks in time according to the display conditions.
The working frequency band of the terahertz imaging host is 0.1-10 thz, and the imaging view field covers the width of the security inspection channel. The reflecting mirror group housing is made of terahertz wave band wave-transparent materials such as PVC, PP and the like, and the inside of the security inspection channel is made of wave-absorbing materials, so that the security inspection channel is not limited to natural wave-absorbing materials such as wood and the like.
Referring to fig. 1, when a person enters a security inspection channel at a normal walking speed and travels to an optimal imaging position between mirror groups 3, terahertz imaging hosts 2 in two equipment cabins 1 respectively perform terahertz imaging on the front and back sides of the person, and meanwhile, terahertz waves on the side surfaces of the person are received by an edge detection module of the terahertz imaging hosts through the mirror groups 3 and imaged in real time. The optical camera of the optical acquisition module 9 in the equipment cabin 1 simultaneously acquires visible light information of personnel in real time. When a person walks out of the security inspection channel, the terahertz real-time image and the visible light image are synchronously transmitted to the central control display system, the images are analyzed by combining a model based on deep learning, the visible light information and the terahertz information of the person are timely displayed, suspicious articles are marked, and the security inspector further rechecks the suspicious articles.
Fig. 2 is a schematic flow chart of a central control display system for analyzing terahertz images and visible light images, and the terahertz real-time images and the visible light images are synchronously transmitted to the central control display system, wherein the terahertz real-time images comprise front and back terahertz images and side terahertz images, the front and back terahertz images and the side terahertz images are respectively input into a front and back target detection model based on deep learning and a side target detection model based on deep learning for target detection, and a front and back detection result and a side detection result are output; the visible light images are synchronously input into a visible light detection model based on deep learning, clothing and face detection results are output, and finally the system comprehensively analyzes the front and back detection results, the side detection results and the clothing and face detection results to output security inspection results.
In a preferred embodiment, the front terahertz image, the back terahertz image and the side terahertz image can be fused and the target is detected, and the specific steps are as follows:
and calculating the characteristic value difference between the front terahertz image, the back terahertz image and the two side terahertz images, and selecting the two terahertz images with the minimum characteristic value difference.
And respectively comparing the front terahertz image and the back terahertz image with the left terahertz image and the right terahertz image, and calculating the characteristic value difference between the two images.
Selecting one reference feature point in the front terahertz image or the back terahertz image as a preferred point, and extracting feature vectors of the selected reference feature point, wherein the feature vectors comprise information of different dimensions such as color, texture, shape, size and the like.
Searching a plurality of matching feature points which are possibly matched with the reference feature points in the previous step in the left-side terahertz chart or the right-side terahertz chart.
Extracting a plurality of feature vectors corresponding to the plurality of matching feature points, calculating similarity values of the plurality of feature vectors and the feature vectors of the extracted reference feature points, and screening out the matching feature points which are most matched with the reference feature points.
Repeating the steps until a plurality of reference feature points and a plurality of matched feature points matched with the reference feature points are selected, wherein the reference feature points are connected to form a reference feature unit, and the matched feature points are connected to form a matched feature unit; and calculating the characteristic value difference between the front terahertz image and the back terahertz image and between the two side terahertz images, and selecting two terahertz images with the minimum characteristic value difference, for example, a front terahertz image and a left side terahertz image or a back terahertz image and a right side terahertz image.
Reading feature vectors of pre-stored images of suspicious articles, respectively calculating matching degree S ij according to pixel positions (i, j) and two terahertz images with minimum feature value difference calculated in the previous step, wherein the matching degree calculation formula is as follows:
Where V ij is the feature vector at the location of coordinate (i, j) in the image of the suspicious item, And taking the coordinates in the terahertz image corresponding to the point with the highest matching degree as the center point of the detected suspicious object as the feature vector at the position of the coordinates (m, n) in the kth image in the two images with the smallest difference values of the feature values.
As shown in fig. 3, which is a schematic diagram of a comprehensive analysis flow of the system, when the detection result of the terahertz image (i.e., the front and back detection result or the side detection result) is judged to be dangerous, matching the terahertz image with the visible light image, and meanwhile, marking the detection result of the matched visible light image as dangerous, thereby locking a target and outputting a dangerous signal; when the detection result of the terahertz image (namely the front and back detection result or the side detection result) is judged to be safe, the detection result of the matched visible light image is marked to be safe, whether the number of the terahertz detection objects is larger than the set number is continuously judged, if so, the danger is output, and if not, the safety is output.
In a preferred embodiment, the step of matching the terahertz image and the visible light image is:
The front or back visible light image and the front or back terahertz image are subjected to low-pass filtering to obtain images A and B;
the gray level layers of the two images A and B are overlapped, so that the distribution relation between gray level value sets of the two images is better compared and analyzed, and the difference between different data sets is more clearly observed.
The gray map layer stacking function H ab is:
Hab[a(i,j),b(i,j)]=Nab；i=1,…，M；j=1,…，N；
Wherein a (i, j), B (i, j) are gray value pairs at positions (i, j) in images a and B with pixels M x N, N ab representing the number of occurrences of the same gray value pair;
the superimposed normalized value P AB (a, B) for images a and B is:
Calculating a coincidence information value M (a, B) of the images a and B using the superimposed normalized value P AB (a, B):
wherein, P A (a) and P B (B) are the independent gray scale distribution values of images A and B, respectively; the larger the superposition information value M (A, B), the higher the matching degree is; otherwise, the lower the matching degree is.
The central control display system of the equipment cabin at the outlet judges whether the images have suspicious articles or not through an artificial intelligent deep learning algorithm, and displays terahertz images and visible light images on a display screen in real time to mark the positions of the suspicious articles.
The deep learning algorithm flow chart of the target detection model comprises the following steps:
s1, inputting an image, and converting terahertz original image data into single-channel gray image data.
The terahertz original image data are distributed in 10000-16000 in numerical value, and the image data are generally 0-255. The conversion of terahertz raw image data into grayscale image data can be represented by the following linear transformation function f:
s(x,y)=f(r(x,y))；
Where r (x, y) represents the data value at the terahertz original image position (x, y), and s (x, y) represents the pixel value at the position (x, y) in the converted grayscale image. The above formula is transformed into after the linear transformation:
s(x,y)=kr(x,y)+c；
where k is the slope and c is the intercept.
S2, interpolating the gray level image to a preset grid size.
In a general image interpolation algorithm, the aspect ratio of an object in an image is changed to cause image distortion, and the embodiment adopts LetterBox algorithm, and under the condition of maintaining the aspect ratio of the image, the whole image is interpolated to a preset grid size, and the specific steps are as follows:
Firstly, scaling the longest side of the gray image after linear transformation to a preset grid size, and then applying the scaling to the other side; after scaling, the size of one side is smaller than the preset grid size, and the two sides of the side are respectively zero-padded to the preset grid size.
And S3, extracting the characteristics of the interpolated image.
The interpolated image (size: 416×416) is input to Darkenet-53 network structure, and a series of convolution and residual operations are performed to obtain feature maps of 1/8 (size: 52×52), 1/16 (size: 26×26) and 1/32 (size: 13×13) of the original image respectively.
S4, judging whether the object and the position and the category of the object exist in the feature map.
For an input feature map, it is mapped to 3-scale output tensors, representing the probability that various objects exist for each grid of the feature map.
For example, a feature map of 416×416 is input, mapped to an output tensor of 3 scales, and 3 prior frames are set, so that there are 13×13×3+26×26×3+52×52×3= 10647 prediction targets in total.
Each prediction is a (4+1+num_class) =85-dimensional vector, and this 85-dimensional vector contains the frame coordinates (4 values), the frame confidence (1 value), and the probability of the object class (num_ classes).
S5, predicting a target boundary box through a loss functionAnd measuring the difference between the predicted target boundary box and the actual value, and judging whether the target is contained or not.
In the object detection model, rectangular boxes of different sizes and aspect ratios are predefined as a priori boxes, which are used to initialize the candidate regions.
In the training process, the model predicts not only the target boundary box position, but also whether each prior box contains a target, and the width and height of the preset prior box are as follows:
；
；
；
；
Using a rectangular box to represent a target boundary box, (c x,cy) representing the number of grids of which the upper left corner is far from the leftmost upper corner of the grid where the point is located; (p w,ph) represents the side length of the a priori frame;
(t x,ty): offset of the target center point relative to the upper left corner of the grid where the point is located;
(t w,th): predicting the width and height of the frame;
The resulting frame coordinate value is b x,by,bw,bh, and the net learning target is t x,ty,tw,th.
S6, using a non-maximum value to inhibit and eliminate overlapped boundary boxes, and selecting a target boundary box with highest confidence as a final detection result. In non-maximum suppression, the goal is to select the most likely bounding box from a set of overlapping target bounding boxes, and suppress other inaccurate bounding boxes.
And measuring the positions of the boundary frames, wherein an intersection ratio IoU index is adopted, and the intersection ratio is the basis of the target detection NMS and is the ratio of the intersection part area of the boundary frames to the intersection part area of the boundary frames.
The non-maximum suppression procedure for IoU is as follows:
1. sorting all the candidate frames according to the confidence scores;
2. Selecting a boundary box with highest confidence level, adding the boundary box to a final output list, and adding the boundary box to the final output list;
3. for the remaining candidate boxes, calculating IoU values of the candidate boxes and the current best box;
4. If IoU of a certain candidate box and the current best box is greater than a threshold (typically 0.5), deleting the candidate box;
5. the above process is repeated until there are no remaining candidate boxes.
The channel type terahertz human body security inspection system fused with the multi-dimensional detection technology can simultaneously utilize the terahertz imaging technology and the optical imaging technology to realize multi-angle and all-dimensional detection of a target, and improve the accuracy and reliability of security inspection. Meanwhile, by combining a deep learning technology, the target can be rapidly and accurately identified, manual intervention is reduced, and security inspection efficiency is improved.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present application, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted across a computer-readable storage medium. The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), etc.
While the application has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the application. Therefore, the protection scope of the application is subject to the protection scope of the claims.
Claims (7)
1. Channel-type terahertz human body security inspection system fused by multidimensional detection technology, which is characterized by comprising: the device comprises two equipment cabins, two terahertz imaging hosts and a central control display system;
the security inspection channel is positioned between the two equipment cabins, and the two terahertz imaging hosts are respectively placed in the two equipment cabins;
the security inspection channel comprises two groups of reflectors, the two groups of reflectors are respectively matched with two terahertz imaging hosts, when a pedestrian passes through the security inspection channel, terahertz imaging is carried out on the side face of the person to form two side face terahertz images, and the two terahertz imaging hosts capture terahertz waves emitted by the front face and the back face of the person at the same time to form front face and back face terahertz images;
an optical camera is arranged in the equipment cabin, and visible light information of personnel is collected in real time to form front and back visible light images;
The central control display system detects suspicious articles by utilizing a deep-learning target detection model to the terahertz images and the visible light images of all sides, and matches the terahertz images of the front side and the back side with the visible light images of the front side and the back side respectively to lock personnel targets.
2. The channel terahertz human body security system according to claim 1, wherein front and back terahertz images, two side terahertz images are fused and target detection is performed:
Respectively comparing the front terahertz image and the back terahertz image with the two side terahertz images, calculating the characteristic value difference between the images, and selecting two terahertz images with the minimum characteristic value difference;
Comparing the pre-stored images of the suspicious object with two terahertz images with minimum difference values of characteristic values respectively, and calculating the matching degree S ij according to the pixel positions (i, j):
Where V ij is the feature vector at the location of coordinate (i, j) in the image of the suspicious item, And taking the coordinate position in the terahertz image corresponding to the highest matching degree point as the detected suspicious article center point for the feature vector at the coordinate (m, n) position in the kth image in the two terahertz images.
3. The channel terahertz human body security system according to claim 2, wherein comparing the front and back terahertz images with the two side terahertz images, respectively, and calculating a characteristic value difference between the images, comprises:
Selecting one reference feature point in the front terahertz image or the back terahertz image as a preferred point, and extracting feature vectors of the reference feature point;
searching a plurality of matching feature points matched with the reference feature points in the left terahertz chart or the right terahertz chart;
extracting a plurality of feature vectors corresponding to the plurality of matching feature points, calculating similarity values of the plurality of feature vectors and the feature vectors of the extracted reference feature points, and screening out one matching feature point which is most matched with the reference feature points;
Repeating the steps until a plurality of reference feature points and a plurality of matched feature points matched with the reference feature points are selected, wherein the reference feature points are connected to form a reference feature unit, and the matched feature points are connected to form a matched feature unit;
and calculating the characteristic value difference between the front terahertz image, the back terahertz image and the two side terahertz images, and selecting the two terahertz images with the minimum characteristic value difference.
4. The channel terahertz human body security system of claim 1, wherein matching front and back terahertz images with front and back visible light images, respectively, comprises:
The front or back visible light image and the front or back terahertz image are subjected to low-pass filtering to obtain images A and B;
the gray map layers of the two images a and B are superimposed, and the gray map layer-stacking function H ab is:
Hab[a(i,j),b(i,j)]=Nab；i=1,…，M；j=1,…，N；
Wherein a (i, j), B (i, j) are gray value pairs at positions (i, j) in images a and B with pixels M x N, N ab representing the number of occurrences of the same gray value pair;
the superimposed normalized value P AB (a, B) for images a and B is:
；
Calculating a coincidence information value M (a, B) of the images a and B using the superimposed normalized value P AB (a, B):
；
wherein, P A (a) and P B (B) are the independent gray scale distribution values of images A and B, respectively; the larger the superposition information value M (A, B), the higher the matching degree is; otherwise, the lower the matching degree is.
5. The channel terahertz human body security inspection system according to claim 1, wherein terahertz images are input into a terahertz target detection model for suspicious object detection, and terahertz image detection results are output; and synchronously inputting the visible light images into a visible light detection model, and outputting clothing and face detection results.
6. The channel terahertz human body security system according to claim 1, wherein the terahertz imaging host has a total detection module comprising: a positive detection module unit and a side detection module; the positive detection module unit accounts for 3/4 of the total detection module and is responsible for collecting terahertz waves emitted by the front and back surfaces of the personnel, the side detection module unit accounts for 1/4 of the total detection module and is responsible for collecting terahertz waves emitted by the left and right side surfaces of the personnel.
7. The channel-type terahertz human body security inspection system according to claim 4, wherein when the terahertz image detection result is judged to be dangerous, the terahertz image and the visible light image are matched, and meanwhile, the matched visible light image detection result is also marked as dangerous, so that a target is locked; and when the terahertz image detection result is judged to be safe, the matched visible light image detection result is marked to be safe, whether the number of terahertz detection objects is larger than the set number is continuously judged, if so, the danger is output, and if not, the safety is output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410329967.8A CN117930375B (en) | 2024-03-22 | 2024-03-22 | Multi-dimensional detection technology fused channel type terahertz human body security inspection system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410329967.8A CN117930375B (en) | 2024-03-22 | 2024-03-22 | Multi-dimensional detection technology fused channel type terahertz human body security inspection system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117930375A true CN117930375A (en) | 2024-04-26 |
| CN117930375B CN117930375B (en) | 2024-06-25 |
Family
ID=90764976
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410329967.8A Active CN117930375B (en) | 2024-03-22 | 2024-03-22 | Multi-dimensional detection technology fused channel type terahertz human body security inspection system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117930375B (en) |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016035394A (en) * | 2014-08-01 | 2016-03-17 | パイオニア株式会社 | Terahertz wave imaging device and terahertz wave imaging method |
| CN108197649A (en) * | 2017-12-29 | 2018-06-22 | 厦门大学 | A kind of Terahertz image clustering analysis method and system |
| CN108491895A (en) * | 2018-05-21 | 2018-09-04 | 深圳市神飞电子科技有限公司 | A kind of Terahertz perspective imaging identifying system |
| CN108898132A (en) * | 2018-05-25 | 2018-11-27 | 广东工业大学 | A kind of terahertz image dangerous material recognition methods based on Shape context description |
| CN109101867A (en) * | 2018-06-11 | 2018-12-28 | 平安科技（深圳）有限公司 | A kind of image matching method, device, computer equipment and storage medium |
| CN110147807A (en) * | 2019-01-04 | 2019-08-20 | 上海海事大学 | A kind of ship intelligent recognition tracking |
| CN111257957A (en) * | 2020-02-25 | 2020-06-09 | 西安交通大学 | Identification tracking system and method based on passive terahertz imaging |
| EP3796051A1 (en) * | 2019-09-18 | 2021-03-24 | Nuctech Company Limited | Personnel security inspection method and personnel security inspection system |
| CN112633392A (en) * | 2020-12-29 | 2021-04-09 | 博微太赫兹信息科技有限公司 | Terahertz human body security inspection image target detection model training data augmentation method |
| EP3845886A1 (en) * | 2019-12-30 | 2021-07-07 | Tsinghua University | Terahertz wave security inspection system and method |
| CN113673548A (en) * | 2020-05-15 | 2021-11-19 | 同方威视技术股份有限公司 | Security inspection system and method integrating visible light image, depth image and terahertz image |
| CN114972458A (en) * | 2022-05-27 | 2022-08-30 | 广东亿嘉和科技有限公司 | Method and system for registering visible light and infrared thermal imaging images |
| CN115793083A (en) * | 2022-12-08 | 2023-03-14 | 上海亨临光电科技有限公司 | Passive millimeter wave/terahertz human body security inspection system |
| CN116299751A (en) * | 2023-03-22 | 2023-06-23 | 上海亨临光电科技有限公司 | Method and device for rapidly extracting passive terahertz human body security inspection information |
| KR102574103B1 (en) * | 2023-04-04 | 2023-09-04 | (주)팬옵틱스 | Apparatus and System for Detecting Hidden Object Based on Artificial Intelligence learning Model Using THz Scan Image |
| CN117132842A (en) * | 2023-10-26 | 2023-11-28 | 江苏鹰创科技有限公司 | Intelligent forbidden article detection method based on terahertz characteristics |
-
2024
- 2024-03-22 CN CN202410329967.8A patent/CN117930375B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016035394A (en) * | 2014-08-01 | 2016-03-17 | パイオニア株式会社 | Terahertz wave imaging device and terahertz wave imaging method |
| CN108197649A (en) * | 2017-12-29 | 2018-06-22 | 厦门大学 | A kind of Terahertz image clustering analysis method and system |
| CN108491895A (en) * | 2018-05-21 | 2018-09-04 | 深圳市神飞电子科技有限公司 | A kind of Terahertz perspective imaging identifying system |
| CN108898132A (en) * | 2018-05-25 | 2018-11-27 | 广东工业大学 | A kind of terahertz image dangerous material recognition methods based on Shape context description |
| CN109101867A (en) * | 2018-06-11 | 2018-12-28 | 平安科技（深圳）有限公司 | A kind of image matching method, device, computer equipment and storage medium |
| CN110147807A (en) * | 2019-01-04 | 2019-08-20 | 上海海事大学 | A kind of ship intelligent recognition tracking |
| EP3796051A1 (en) * | 2019-09-18 | 2021-03-24 | Nuctech Company Limited | Personnel security inspection method and personnel security inspection system |
| EP3845886A1 (en) * | 2019-12-30 | 2021-07-07 | Tsinghua University | Terahertz wave security inspection system and method |
| CN113126176A (en) * | 2019-12-30 | 2021-07-16 | 清华大学 | Terahertz wave security inspection system and method |
| CN111257957A (en) * | 2020-02-25 | 2020-06-09 | 西安交通大学 | Identification tracking system and method based on passive terahertz imaging |
| CN113673548A (en) * | 2020-05-15 | 2021-11-19 | 同方威视技术股份有限公司 | Security inspection system and method integrating visible light image, depth image and terahertz image |
| CN112633392A (en) * | 2020-12-29 | 2021-04-09 | 博微太赫兹信息科技有限公司 | Terahertz human body security inspection image target detection model training data augmentation method |
| CN114972458A (en) * | 2022-05-27 | 2022-08-30 | 广东亿嘉和科技有限公司 | Method and system for registering visible light and infrared thermal imaging images |
| CN115793083A (en) * | 2022-12-08 | 2023-03-14 | 上海亨临光电科技有限公司 | Passive millimeter wave/terahertz human body security inspection system |
| CN116299751A (en) * | 2023-03-22 | 2023-06-23 | 上海亨临光电科技有限公司 | Method and device for rapidly extracting passive terahertz human body security inspection information |
| KR102574103B1 (en) * | 2023-04-04 | 2023-09-04 | (주)팬옵틱스 | Apparatus and System for Detecting Hidden Object Based on Artificial Intelligence learning Model Using THz Scan Image |
| CN117132842A (en) * | 2023-10-26 | 2023-11-28 | 江苏鹰创科技有限公司 | Intelligent forbidden article detection method based on terahertz characteristics |
Non-Patent Citations (2)
| Title |
|---|
| 于勇: "《互联网电视概论》", 30 June 2020, 吉林人民出版社, pages: 111 * |
| 杨帆: "《数字图像处理与分析》", 31 May 2015, 北京航空航天大学出版社, pages: 280 - 281 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117930375B (en) | 2024-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109977877B (en) | Intelligent auxiliary image judging method, system and system control method for security inspection | |
| CN102819740B (en) | A kind of Single Infrared Image Frame Dim targets detection and localization method | |
| Izadi et al. | Automatic building detection in aerial images using a hierarchical feature based image segmentation | |
| CN105957054B (en) | A kind of image change detection method | |
| CN109948415A (en) | Remote sensing image object detection method based on filtering background and scale prediction | |
| US10186050B2 (en) | Method and system for detection of contraband narcotics in human digestive tract | |
| CN102073873B (en) | Method for selecting SAR (spaceborne synthetic aperture radar) scene matching area on basis of SVM (support vector machine) | |
| CN103226826B (en) | Based on the method for detecting change of remote sensing image of local entropy visual attention model | |
| Rogers et al. | A deep learning framework for the automated inspection of complex dual-energy x-ray cargo imagery | |
| CN105608691A (en) | High-resolution SAR image individual building extraction method | |
| CN113111727B (en) | Feature alignment-based method for detecting rotating target in remote sensing scene | |
| CN107767400A (en) | Remote sensing images sequence moving target detection method based on stratification significance analysis | |
| CN103729842A (en) | Fabric defect detection method based on local statistical characteristics and overall significance analysis | |
| CN112967289A (en) | Security check package matching method, system, equipment and storage medium | |
| CN115439693A (en) | Training method of target recognition network model, electronic device and program product | |
| Chen et al. | Space-based infrared aerial target detection method via interframe registration and spatial local contrast | |
| CN110046595A (en) | A kind of intensive method for detecting human face multiple dimensioned based on tandem type | |
| Wang et al. | Detection method of obstacles in the dangerous area of electric locomotive driving based on MSE-YOLOv4-Tiny | |
| CN113807231A (en) | X-ray contraband detection method based on UNET (UNET) downsampling convolutional neural network | |
| CN110992324B (en) | Intelligent dangerous goods detection method and system based on X-ray image | |
| CN117930375B (en) | Multi-dimensional detection technology fused channel type terahertz human body security inspection system | |
| CN114792300B (en) | X-ray broken needle detection method based on multi-scale attention | |
| CN113706815B (en) | Vehicle fire identification method combining YOLOv3 and optical flow method | |
| Sportouche et al. | Building detection and height retrieval in urban areas in the framework of high resolution optical and SAR data fusion | |
| CN114663711A (en) | X-ray security inspection scene-oriented dangerous goods detection method and device |
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 |