CN-116704602-B - Gate channel spoofing prevention processing method, device, medium and equipment

CN116704602BCN 116704602 BCN116704602 BCN 116704602BCN-116704602-B

Abstract

The embodiment of the application provides a processing method, a device, a medium and equipment for preventing gate channel spoofing. The method comprises the steps of obtaining an image sequence obtained by shooting a target area, determining head characteristic information of pedestrians and corresponding head detection frames contained in images to be identified in the image sequence, predicting according to the head characteristic information and the head detection frames of the pedestrians, determining human body detection frames of the pedestrians, extracting corresponding human body appearance characteristics, carrying out target tracking according to timestamp information of the images to be identified and the head detection frames, the human body detection frames and the human body appearance characteristics of the pedestrians contained in the images to be identified, generating a motion track corresponding to the pedestrians, and triggering an impotent passing alarm if the pedestrians with face brushing requests pass do not finish passing after the gate is in a one-time switching state. The technical scheme of the embodiment of the application ensures the accuracy of the identification of the imposter and the passgate and improves the identification efficiency.

Inventors

CHEN MINGMU
WANG HANCHAO
XU SHAOKAI
JIA BAOZHI
HE YIFAN

Assignees

厦门瑞为信息技术有限公司

Dates

Publication Date: 20260505
Application Date: 20230601

Claims (10)

1. A gate channel spoofing prevention processing method is characterized by comprising the following steps: acquiring an image sequence obtained by shooting a target area, wherein the image sequence comprises a plurality of continuous images to be identified; performing target detection on each image to be identified in sequence to determine head characteristic information of pedestrians and corresponding head detection frames contained in the images to be identified; Predicting according to the head characteristic information and the head detection frames of each pedestrian, determining a human body detection frame of each pedestrian, and extracting human body appearance characteristics corresponding to the position information of the human body detection frame; Performing target tracking according to the timestamp information of the image to be identified and the head detection frame, the human body detection frame and the human body appearance characteristics of each pedestrian contained in the image to be identified, and generating a motion track corresponding to each pedestrian; According to the motion trail corresponding to each pedestrian, after the gate finishes a switching state once, triggering an impoverishment gate crossing alarm if the pedestrian who performs a face brushing request gate crossing does not finish gate crossing and other pedestrians exist to finish gate crossing; The human body prediction adopts a AnchorFree training mode, namely the target detection model can directly predict the distance between the center point of the head detection frame and the human body labeling frame, and the formula is as follows: Wherein, the Respectively represent the distance between the center point of the head detection frame predicted by the target detection model and the left, upper, right and lower boundaries of the human body detection frame Coordinate values representing the center point of the head detection frame; respectively representing the left, right, upper and lower coordinate values of the human body annotation frame; then, the human body frame is decoded again by the following formula in the prediction process: 。
2. The method according to claim 1, wherein performing object tracking according to the timestamp information of the image to be identified and the head detection frame, the human body detection frame, and the human body appearance feature of each pedestrian included in the image to be identified, generating a motion trail corresponding to each pedestrian, comprises: Determining a head motion incidence matrix and a human appearance incidence matrix between a current image to be identified and a previous image to be identified according to the timestamp information of the image to be identified and the head detection frame, the human detection frame and the human appearance characteristics of each pedestrian contained in the image to be identified; Calculating according to the head movement incidence matrix and the human appearance incidence matrix, and determining a corresponding cost matrix; and matching by adopting a Hungary algorithm according to the cost matrix, and carrying out Kalman filtering according to the head detection frames of the matched pedestrians in the two images to be identified so as to update the corresponding motion trail of the pedestrians.
3. The method of claim 2, wherein after performing kalman filtering according to the head detection frame of the pedestrian matched in the two images to be identified to update the motion trail corresponding to the pedestrian, the method further comprises: And updating the human body appearance characteristics corresponding to the pedestrian in the current image to be identified according to the human body appearance characteristics corresponding to the same pedestrian in the previous image to be identified.
4. A method according to claim 3, wherein the human appearance characteristics are updated according to the following formula: , Wherein, the In order to be a momentum of the light, Representing the human body appearance characteristics corresponding to the ith pedestrian, And representing the human appearance characteristics corresponding to the human detection frame of the ith pedestrian matched with the current image to be identified.
5. The method according to claim 1, wherein extracting human appearance features corresponding to position information of the human detection frame comprises: determining a center point of the human body detection frame according to the position information of the human body detection frame; And extracting the characteristics according to the central point of the human body detection frame, and determining the corresponding human body appearance characteristics.
6. The method of claim 5, wherein determining the corresponding human appearance feature based on feature extraction from the center point of the human detection frame comprises: according to the center point of the human body detection frame, learning is performed through multi-class cross entropy loss functions during training, and feature extraction is directly performed through a network during deployment, so that corresponding human body appearance features are determined.
7. A gate channel spoofing prevention processing apparatus, comprising: the image acquisition module is used for acquiring an image sequence obtained by shooting a target area, wherein the image sequence comprises a plurality of continuous images to be identified; The target detection module is used for sequentially carrying out target detection on each image to be identified so as to determine the head characteristic information of the pedestrian contained in the image to be identified and a corresponding head detection frame; The target detection module is further used for predicting according to the head characteristic information and the head detection frames of the pedestrians, determining human body detection frames of the pedestrians, and extracting human body appearance characteristics corresponding to the position information of the human body detection frames; the target tracking module is used for carrying out target tracking according to the timestamp information of the image to be identified and the head detection frame, the human body detection frame and the human body appearance characteristics of each pedestrian contained in the image to be identified, and generating a motion track corresponding to each pedestrian; The processing module is used for triggering an impoverishment gate crossing alarm if the passersby who performs a face brushing request gate crossing does not finish the gate crossing and other passersby exist to finish the gate crossing after the gate finishes a switching state once according to the motion trail corresponding to each passersby; The human body prediction adopts a AnchorFree training mode, namely the target detection model can directly predict the distance between the center point of the head detection frame and the human body labeling frame, and the formula is as follows: Wherein, the Respectively represent the distance between the center point of the head detection frame predicted by the target detection model and the left, upper, right and lower boundaries of the human body detection frame Coordinate values representing the center point of the head detection frame; respectively representing the left, right, upper and lower coordinate values of the human body annotation frame; then, the human body frame is decoded again by the following formula in the prediction process: 。
8. The apparatus of claim 7, wherein the target tracking module is configured to: Determining a head motion incidence matrix and a human appearance incidence matrix between a current image to be identified and a previous image to be identified according to the timestamp information of the image to be identified and the head detection frame, the human detection frame and the human appearance characteristics of each pedestrian contained in the image to be identified; Calculating according to the head movement incidence matrix and the human appearance incidence matrix, and determining a corresponding cost matrix; and matching by adopting a Hungary algorithm according to the cost matrix, and carrying out Kalman filtering according to the head detection frames of the matched pedestrians in the two images to be identified so as to update the corresponding motion trail of the pedestrians.
9. A computer-readable medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the gate channel spoofing prevention processing method as claimed in any one of claims 1 to 6.
10. An electronic device, comprising: One or more processors; storage means for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the gate channel spoofing prevention processing method of any of claims 1 to 6.

Description

Gate channel spoofing prevention processing method, device, medium and equipment Technical Field The application relates to the technical field of computer vision, in particular to a gate channel spoofing prevention processing method, a gate channel spoofing prevention processing device, a gate channel spoofing prevention medium and gate channel spoofing prevention equipment. Background In modern society, in order to improve the passing efficiency, public places such as airports and subways select gates to check tickets. However, in actual use, some pedestrians often adopt means such as impersonation and displacement, so that fraudulent ticket checking and gate checking evasion supervision bring great difficulty to safety management. In the current technical scheme, identity card checking or card swiping is often adopted for identity verification, but the method has lower checking efficiency and higher cost. Disclosure of Invention The embodiment of the application provides a processing method, a device, a medium and equipment for preventing cheating of a gate channel, which can further ensure the accuracy of identifying the false gate behavior at least to a certain extent and improve the identification efficiency. Other features and advantages of the application will be apparent from the following detailed description, or may be learned by the practice of the application. According to an aspect of the embodiment of the present application, there is provided a method for processing gate channel spoofing prevention, including: acquiring an image sequence obtained by shooting a target area, wherein the image sequence comprises a plurality of continuous images to be identified; performing target detection on each image to be identified in sequence to determine head characteristic information of pedestrians and corresponding head detection frames contained in the images to be identified; Predicting according to the head characteristic information and the head detection frames of each pedestrian, determining a human body detection frame of each pedestrian, and extracting human body appearance characteristics corresponding to the position information of the human body detection frame; Performing target tracking according to the timestamp information of the image to be identified and the head detection frame, the human body detection frame and the human body appearance characteristics of each pedestrian contained in the image to be identified, and generating a motion track corresponding to each pedestrian; And according to the motion trail corresponding to each pedestrian, after the gate finishes a switching state once, triggering an impoverishment gate crossing alarm if the pedestrian who performs a face brushing request gate crossing does not finish gate crossing and other pedestrians exist to finish gate crossing. According to an aspect of an embodiment of the present application, there is provided a processing apparatus for preventing fraud in a gate channel, the apparatus including: the image acquisition module is used for acquiring an image sequence obtained by shooting a target area, wherein the image sequence comprises a plurality of continuous images to be identified; The target detection module is used for sequentially carrying out target detection on each image to be identified so as to determine the head characteristic information of the pedestrian contained in the image to be identified and a corresponding head detection frame; The target detection module is further used for predicting according to the head characteristic information and the head detection frames of the pedestrians, determining human body detection frames of the pedestrians, and extracting human body appearance characteristics corresponding to the position information of the human body detection frames; the target tracking module is used for carrying out target tracking according to the timestamp information of the image to be identified and the head detection frame, the human body detection frame and the human body appearance characteristics of each pedestrian contained in the image to be identified, and generating a motion track corresponding to each pedestrian; And the processing module is used for triggering an impoverishment gate crossing alarm if the passersby who performs a face brushing request gate crossing does not finish the gate crossing and other passersby exist to finish the gate crossing after the gate finishes a switching state once according to the motion trail corresponding to each passersby. According to an aspect of the embodiments of the present application, there is provided a computer readable medium having stored thereon a computer program which, when executed by a processor, implements a gate channel spoofing prevention processing method as described in the above embodiments. According to one aspect of the embodiment of the application, the electronic device comprises one or more processors and a storage devic