Double-door channel light curtain pedestrian dynamic detection method and risk interception system

By setting up multi-layer light curtains and recognition units in the dual-door channel, establishing a human body sensing curve feature map and using a human body posture detection model, the problems of inaccurate detection of the dual-door channel light curtain and insufficient passage efficiency are solved, and more accurate Pedestrian status monitoring and traffic efficiency have been improved.

CN118553043BActive Publication Date: 2024-11-01SHENZHEN MAXVISION TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411008327.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-01
Estimated Expiration
2044-07-26

AI Technical Summary

Technical Problem

The double-door channel light curtain detection is inaccurate and the traffic efficiency is insufficient. It is affected by interference from swing arms, luggage, ambient light, etc., resulting in inaccurate pedestrian status and anti-trailing detection.

Method used

The front bucket horizontal light curtain, front bucket gate, middle bucket horizontal light curtain and rear bucket gate are used to set up the front bucket horizontal light curtain, and the human body sensing curve feature map is established through the ID identification unit and the biometric recognition unit to match the human body sensing curve features. Track pedestrian targets in real time, improve detection accuracy and anti-trailing ability, and monitor pedestrian status through human posture detection model.

Benefits of technology

It improves the accuracy of human detection and anti-trailing ability, improves the accuracy and traffic efficiency of synchronous detection of multi-peer targets, standardizes the traffic status of pedestrians, cracks down on violations, and enhances the reliability of inspection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118553043B_ABST
    Figure CN118553043B_ABST
Patent Text Reader

Abstract

The present application provides a double-door channel light curtain pedestrian dynamic detection method and a risk interception system, the double-door channel light curtain pedestrian dynamic detection method is applied to the double-door channel, the certificate information is identified by the certificate recognition unit, and the front barrel gate is opened after successful identification; the front barrel light curtain data is obtained through the front barrel horizontal light curtain and the first vertical light curtain, and a curve graph of the trigger point number y of the front barrel light curtain data with respect to the time node x is established, and the front barrel gate is closed after matching the human body sensing curve characteristics from the curve graph; the pedestrian target is detected from the front segment data of the middle barrel light curtain data obtained from the middle barrel horizontal light curtain, the pedestrian target is associated with the certificate information, and the status of the associated pedestrian target is tracked in real time and detected; the identity information of all pedestrians in the middle barrel horizontal light curtain is identified by the biometric recognition unit according to the passage order, and the rear barrel gate is opened after the identity information and the certificate information are successfully matched. The method of the present application improves the detection accuracy and passage efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the technical field of measuring light curtains, and more specifically, to a double-door channel light curtain pedestrian dynamic detection method and a risk interception system. Background Art

[0002] The double-door channel is a security inspection device with double gates, which is often used to conduct security checks on pedestrians in important security inspection areas. When an abnormal pedestrian is found after inspection, the abnormal pedestrian is located between the double gates, which is conducive to the rapid implementation of risk interception.

[0003] The existing double-door channel adopts a one-by-one access method, that is, the entrance gate will not be opened until the pedestrian passes through the exit gate, which leads to insufficient traffic efficiency. A light curtain will be set up inside the double-door channel for pedestrian position detection and anti-tailgating detection. The status of pedestrians, such as staying or running, is monitored by the pedestrian position. However, the light curtain in the channel is interfered by swing arms, luggage, ambient light, etc., resulting in inaccurate pedestrian status and anti-tailgating detection. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a double-door channel light curtain pedestrian dynamic detection method and a risk interception system to solve the technical problems of inaccurate double-door channel light curtain detection and insufficient traffic efficiency in the prior art.

[0005] To achieve the above-mentioned purpose, the technical solution adopted in the present application is: to provide a double-door channel light curtain pedestrian dynamic detection method, which is applied to the double-door channel, wherein the double-door channel is sequentially provided with a front barrel transverse light curtain, a front barrel gate, a middle barrel transverse light curtain and a rear barrel gate, wherein the front barrel gate is provided with a first vertical light curtain and a document recognition unit, and the rear barrel gate is provided with a biometric feature recognition unit; the double-door channel light curtain pedestrian dynamic detection method comprises the steps of:

[0006] The document information is recognized by the document recognition unit, and the front barrel gate is opened after successful recognition;

[0007] The front bucket light curtain data is obtained through the front bucket horizontal light curtain and the first vertical light curtain, a curve graph of the trigger point number y of the front bucket light curtain data with respect to the time node x is established, and the front bucket gate is closed after matching the human body sensing curve characteristics from the curve graph;

[0008] Detect pedestrian targets from the front-end data of the middle barrel light curtain obtained from the middle barrel transverse light curtain, associate the pedestrian targets with the certificate information, and track and detect the status of the associated pedestrian targets in real time;

[0009] The biometric recognition unit identifies the identity information of all pedestrians in the horizontal light curtain of the center barrel according to the pass order, and opens the rear barrel gate after the identity information matches the certificate information successfully.

[0010] In another embodiment of the present application, a method for matching a human body sensing curve feature from a curve graph comprises the steps of:

[0011] Collect the pre-collected data of the light curtain of the human body passing through the front barrel gate at different speeds;

[0012] Establish a curve graph of the trigger point number y of the light curtain pre-collected data with respect to the time node x, and save it to the human body sensing curve feature database;

[0013] Match the curve graph established based on the front bucket light curtain data with the human body sensing curve feature database;

[0014] If there is a curve graph whose average cosine distance is less than the distance threshold, the human body sensing curve feature is matched from the curve graph.

[0015] In another embodiment of the present application, a method for matching a curve graph established based on front bucket light curtain data with a human body sensing curve feature database comprises the steps of:

[0016] Detect the current speed of pedestrians based on the front barrel transverse light curtain;

[0017] Filter out a curve graph matching the current passing speed from a human body sensing curve feature database;

[0018] Then, the curve graph established based on the front bucket light curtain data is matched with the screened human body sensing curve feature database.

[0019] In another embodiment of the present application, a method for detecting the current speed of a pedestrian based on a front bucket transverse light curtain comprises the steps of:

[0020] Based on the front bucket transverse light curtain, obtain the front bucket transverse light curtain data within the time range from when the front bucket gate is opened to when all light spots in the front bucket transverse light curtain are blocked at least once;

[0021] Define the total number of light spots on the front barrel transverse light curtain as n, where n is a positive even number, T i is the time point when the i-th light spot is blocked for the first time, L is the maximum measured length of the front barrel transverse light curtain, and the calculation formula for the current passing speed V is: .

[0022] In another embodiment of the present application, a method for matching a curve graph established based on front bucket light curtain data with a screened human body sensing curve feature database comprises the steps of:

[0023] In addition, y>c, c is a positive integer, and a peak feature F1 in a curve graph established based on the front bucket light curtain data is obtained, and a peak feature set F2 based on a screened human body sensing curve feature database is obtained;

[0024] The peak feature F1 is matched with the peak feature set F2.

[0025] In another embodiment of the present application, the middle barrel horizontal light curtain includes an upper row of light curtains and a lower row of light curtains, the upper row of light curtains is used to sense the upper body of a human body, and the lower row of light curtains is used to sense the lower body of a human body.

[0026] In another embodiment of the present application, a method for detecting a pedestrian target from the front segment data of the middle barrel light curtain data obtained by the middle barrel transverse light curtain comprises the steps of:

[0027] Train a human posture detection model based on light curtain detection data;

[0028] Intercept the front-end data of the middle barrel light curtain data;

[0029] Inputting the preceding data of the continuous multiple frames into a human posture detection model to detect pedestrian targets;

[0030] The pedestrian targets detected in each frame are converted into pedestrian target feature values ​​and saved in the pedestrian feature library.

[0031] In another embodiment of the present application, the training method of the human posture detection model comprises the steps of:

[0032] Acquire a top view image of the double-door passage based on a top-down perspective in a real environment, and acquire the middle barrel light curtain data to be marked through the middle barrel transverse light curtain at the same time node;

[0033] Detect all human targets in the overhead image, and map the positions of all human targets to the middle barrel light curtain data to be labeled, so as to obtain the labeled middle barrel light curtain data;

[0034] The human posture detection model is trained through several labeled mid-barrel light curtain data.

[0035] In another embodiment of the present application, a method for real-time tracking and detecting the status of an associated pedestrian target includes the steps of:

[0036] Pedestrian detection is performed on the middle barrel light curtain data through the human posture detection model, and the characteristic values ​​of all suspicious pedestrian targets are output;

[0037] Match all suspicious pedestrian target feature values ​​with pedestrian target feature values ​​in the pedestrian feature database, and filter out unmatched suspicious pedestrian target feature values;

[0038] Update the tracking position of the pedestrian target according to the matching results;

[0039] The state detection result of the corresponding pedestrian target is output according to the updated tracking position change.

[0040] A risk interception system includes a monitoring platform and a double-door channel as described above. Multiple pedestrians are checked simultaneously through the double-door channel. When identity inconsistency is detected by a document recognition unit and a biometric recognition unit, the identity detection result is sent to the monitoring platform.

[0041] Compared with the prior art, the double-door channel light curtain pedestrian dynamic detection method provided by the present application is characterized by providing a first vertical light curtain and a certificate recognition unit on the front barrel gate, establishing a curve diagram of the trigger point number y of the front barrel light curtain data with respect to the time node x to match the human body sensing curve characteristics, thereby improving the accuracy of human body detection and anti-tailgating capability; then, pedestrian targets are detected from the front segment data of the middle barrel light curtain data obtained from the middle barrel horizontal light curtain, and the pedestrian targets are associated with the certificate information and the detection status is tracked in real time. On the one hand, the detected non-tracked interference targets can be filtered to improve the accuracy of the synchronous detection of multiple pedestrian targets; on the other hand, the pedestrian status can be detected in real time to supervise the passers-by throughout the process; the identity information is identified according to the passage sequence, the passage status of pedestrians is standardized, the customs clearance behavior of using certificates in violation of regulations is cracked down, and the reliability of multi-person synchronous inspection and the passage efficiency are improved.

[0042] Compared with the prior art, the risk interception system provided by the present application includes a monitoring platform and a double-door channel as described above. After the monitoring platform receives an identity detection result with an inconsistent identity, the abnormal pedestrian is intercepted between the front barrel gate and the rear barrel gate, making it convenient for the staff to implement interception at the first time. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0044] Figure 1 A schematic diagram of the structure of a double-door channel provided in an embodiment of the present application;

[0045] Figure 2 Based on Figure 1 A schematic flow chart of a double-door channel light curtain pedestrian dynamic detection method in a double-door channel;

[0046] Figure 3 A graph of the trigger point number y of the front bucket light curtain data provided in an embodiment of the present application with respect to the time node x;

[0047] Figure 4 Schematic diagram of a method for detecting the current speed of pedestrians based on a front barrel transverse light curtain provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0049] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.

[0050] It should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0051] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0052] Please also read Figures 1 to 3 , the double-door channel light curtain pedestrian dynamic detection method provided in the embodiment of the present application is now described. The double-door channel light curtain pedestrian dynamic detection method is applied to a double-door channel, and the double-door channel is sequentially provided with a front barrel transverse light curtain 10, a front barrel gate 20, a middle barrel transverse light curtain 30, a rear barrel gate 40, and a rear barrel transverse light curtain 50. Specifically, the front barrel gate 20 is provided with a first vertical light curtain 60 and a document recognition unit 70, and the rear barrel gate 40 is provided with a second vertical light curtain 80 and a biometric recognition unit 90.

[0053] It is worth mentioning that in the field of double-door channels, the double-door channels can be divided into front barrel, middle barrel and rear barrel according to the positional relationship. The front barrel transverse light curtain 10 and the front barrel gate 20 belong to the front barrel, the middle barrel transverse light curtain 30 belongs to the middle barrel, and the rear barrel gate 40 and the rear barrel transverse light curtain 50 belong to the rear barrel.

[0054] Specifically, the double-door channel light curtain pedestrian dynamic detection method comprises the steps of:

[0055] Step S1, identifying the certificate information through the certificate identification unit 70, and opening the front barrel gate 20 after successful identification.

[0056] It is understandable that in step S1, the pedestrian enters the double-door passage by swiping a document, which includes but is not limited to a passport, an ID card, a regional pass, etc., and the document information includes but is not limited to name, ID card number, validity period, etc.

[0057] Step S2, obtain the front barrel light curtain data through the front barrel horizontal light curtain 10 and the first vertical light curtain 60, establish a curve graph of the trigger point number y of the front barrel light curtain data with respect to the time node x, and close the front barrel gate 20 after matching the human body sensing curve characteristics from the curve graph.

[0058] In step S2, the front barrel light curtain data also includes the detection results of the front barrel horizontal light curtain 10 and the first vertical light curtain 60, that is, the trigger point number y of the front barrel light curtain data is the sum of the blocked points of the front barrel horizontal light curtain 10 and the first vertical light curtain 60 at the same time node x.

[0059] Because the front barrel transverse light curtain 10 is in front and the first vertical light curtain 60 is in the back, when a pedestrian enters the front and rear, the front barrel transverse light curtain 10 is triggered first and then the first vertical light curtain 60 is triggered. Therefore, in the curve diagram of the triggering point number y of the front barrel light curtain data with respect to the time node x, when a pedestrian passes by, a human body sensing curve characteristic with a sharp increase and a sharp decrease in the slope of the curve will be presented, and interferences such as swing arms, luggage, and ambient light cannot completely block the entire first vertical light curtain 60, thereby improving the accuracy of human body detection, and improving the interference filtering capability and anti-tailgating capability.

[0060] Therefore, after matching the human body sensing curve characteristics from the curve graph, the front barrel gate 20 is closed immediately, thereby improving the anti-tailgating detection effect.

[0061] Step S3, detecting a pedestrian target from the front-end data of the middle barrel light curtain data obtained by the middle barrel transverse light curtain 30, associating the pedestrian target with the certificate information, and tracking and detecting the status of the associated pedestrian target in real time.

[0062] In step S3, after the pedestrian passes through the front barrel, he must be in the front section of the middle barrel transverse light curtain 30. The pedestrian target can be detected based on the front section data of the middle barrel light curtain data. Based on the ID information recognized in step S1, the pedestrian target is associated with the ID information, and the pedestrian target is tracked in real time based on the middle barrel light curtain data.

[0063] In this way, when there are multiple pedestrians in the barrel, by real-time tracking of all pedestrians, on the one hand, the detected non-tracked interference targets can be filtered out to improve the accuracy of simultaneous detection of multiple pedestrian targets; on the other hand, the status of pedestrians can be detected in real time, such as staying, running, parallel, cutting in line, etc., to supervise the passers-by.

[0064] In step S4, the biometric identification unit 90 identifies the identity information of all pedestrians in the center barrel transverse light curtain 30 according to the pass order, and opens the rear barrel gate 40 after the identity information matches the certificate information successfully.

[0065] In step S4, since the pedestrian target and the certificate information are associated in step S3 and the detection status is tracked in real time, the barrel gate 40 recognizes the identity information according to the pass order after exiting, which not only requires the verification of the unity of person and certificate, but also detects the pass order, standardizes the pass status of pedestrians, cracks down on illegal clearance behaviors such as borrowing, stealing, and mixing certificates, and improves the reliability of multi-person simultaneous inspection.

[0066] Step S5, obtain the rear barrel light curtain data through the rear barrel horizontal light curtain 50 and the second vertical light curtain 80, establish a curve graph of the trigger point number y of the rear barrel light curtain data with respect to the time node x, and close the rear barrel gate 40 after matching the human body sensing curve characteristics from the curve graph.

[0067] In step S5, the principle is the same as that of step S2, and its function is to detect pedestrians passing through the rear barrel gate 40, improve interference filtering capability and prevent the gate from pinching pedestrians.

[0068] Compared with the prior art, the double-door channel light curtain pedestrian dynamic detection method provided by the present application is characterized by providing a first vertical light curtain 60 and a certificate recognition unit 70 on the front barrel gate 20, establishing a curve diagram of the trigger point number y of the front barrel light curtain data with respect to the time node x to match the human body sensing curve characteristics, thereby improving the accuracy of human body detection and anti-tailgating capability; then, pedestrian targets are detected from the front segment data of the middle barrel light curtain data obtained from the middle barrel horizontal light curtain 30, and the pedestrian targets are associated with the certificate information and the detection status is tracked in real time. On the one hand, the detected non-tracked interference targets can be filtered to improve the accuracy of the synchronous detection of multiple pedestrian targets; on the other hand, the pedestrian status can be detected in real time to supervise the passers-by throughout the process; the identity information is identified according to the passage sequence, the passage status of pedestrians is standardized, the customs clearance behavior of using certificates in violation of regulations is cracked down, and the reliability of multi-person synchronous inspection and the passage efficiency are improved.

[0069] In another embodiment of the present application, in step S2, the method for matching the human body sensing curve feature from the curve graph comprises the steps of:

[0070] Collect the light curtain pre-collected data of the human body passing through the front barrel gate 20 at different speeds;

[0071] Establish a curve graph of the trigger point number y of the light curtain pre-collected data with respect to the time node x, and save it to the human body sensing curve feature database;

[0072] Match the curve graph established based on the front bucket light curtain data with the human body sensing curve feature database;

[0073] If there is a curve graph whose average cosine distance is less than the distance threshold, the human body sensing curve feature is matched from the curve graph.

[0074] It can be understood that in this embodiment, by establishing a human body sensing curve feature database at different speeds, the electronic data is converted into image features, and the average cosine distance is calculated based on the image detection level, thereby improving the accuracy of light curtain human body detection.

[0075] Furthermore, in step S2, the method of matching the curve graph established based on the front bucket light curtain data with the human body sensing curve feature database comprises the steps of:

[0076] Detect the current speed of pedestrians based on the front barrel transverse light curtain 10;

[0077] Filter out a curve graph matching the current passing speed from a human body sensing curve feature database;

[0078] Then, the curve graph established based on the front bucket light curtain data is matched with the screened human body sensing curve feature database.

[0079] It is understandable that since the speed of travel affects the slope of the curve, it is matched with the human body sensing curve feature database after speed screening, and the speed dimension and image dimension are combined to improve the accuracy of the matching. For example, if the current speed is 0.8m / s, a curve with a speed of 0.7m / s-0.9m / s is screened from the human body sensing curve feature database. Even if there is a curve with an average cosine distance less than the distance threshold, the interference detection result can be filtered due to the large deviation of the speed of travel.

[0080] For further information, please also refer to Figure 4 In step S2, the method for detecting the current speed of pedestrians based on the front barrel transverse light curtain 10 comprises the steps of:

[0081] Based on the front bucket transverse light curtain 10, the front bucket transverse light curtain data is obtained within the time range from when the front bucket gate 20 is opened to when all light spots in the front bucket transverse light curtain 10 are blocked at least once;

[0082] Define the total number of light spots of the front barrel transverse light curtain 10 as n, n is a positive even number, T iis the time point when the i-th light spot is blocked for the first time, L is the maximum measured length of the front bucket transverse light curtain 10, and the calculation formula of the current passing speed V is: .

[0083] It can be understood that in this embodiment, The light spot and The time interval between the first occlusion of the light points, that is, the total groups of light points, and the distance between each group of light points is , and each group of light spots can get a speed, and the average speed of all groups of light spots can be calculated by the above formula. This method can effectively solve the interference of the swing arm action on the speed detection, because the swing arm is a cyclic action and enters the front barrel before the human body. This embodiment uses T i As the time point when the i-th light point is first blocked, that is, T i is the time when the swing arm reaches the light spot for the first time, so each group of light spots can offset the interference of the swing arm. In this way, the average speed of the swing arm passing through all groups of light spots can be obtained to improve the detection accuracy of the passing speed.

[0084] For further information, please also refer to Figure 3 In step S2, a method for matching a curve graph established based on the front bucket light curtain data with a screened human body sensing curve feature database comprises the steps of:

[0085] In addition, y>c, c is a positive integer, and a peak feature F1 in a curve graph established based on the front bucket light curtain data is obtained, and a peak feature set F2 based on a screened human body sensing curve feature database is obtained;

[0086] The peak feature F1 is matched with the peak feature set F2.

[0087] It can be understood that in this embodiment, the front barrel horizontal light curtain 10 is used to detect the speed, and the first vertical light curtain 60 is used to amplify the peak characteristics. In this way, the same peak characteristics at similar speeds can be matched based on the front barrel light curtain data, thereby further improving the human body detection accuracy and interference filtering ability of the front barrel.

[0088] In another embodiment of the present application, please refer to Figure 1 In step S3, the middle barrel transverse light curtain 30 includes an upper row of light curtains and a lower row of light curtains, the upper row of light curtains is used to sense the upper body of the human body, and the lower row of light curtains is used to sense the lower body of the human body.

[0089] It can be understood that by detecting pedestrians simultaneously with the upper and lower light curtains, the dimension of the front barrel light curtain data is increased, and the accuracy of pedestrian targets and tracking detection status is improved.

[0090] Further, in step S3, the method for detecting a pedestrian target from the front segment data of the middle barrel light curtain data obtained by the middle barrel transverse light curtain 30 comprises the steps of:

[0091] Train a human posture detection model based on light curtain detection data;

[0092] Intercept the front-end data of the middle barrel light curtain data;

[0093] Inputting the preceding data of the continuous multiple frames into a human posture detection model to detect pedestrian targets;

[0094] The pedestrian targets detected in each frame are converted into pedestrian target feature values ​​and saved in the pedestrian feature library.

[0095] It is understandable that, since the pedestrian must be in the front section of the middle barrel transverse light curtain 30 after passing through the front barrel, the pedestrian target can be detected based on the front section data of the middle barrel light curtain data. In this embodiment, the pedestrian target is detected by inputting the front section data of multiple consecutive frames into the human body posture detection model, so that the pedestrian target feature value of the pedestrian can be obtained.

[0096] Further, in step S3, the training method of the human posture detection model comprises the steps of:

[0097] Acquire a top view image of the double-door passage based on a top view perspective in a real environment, and acquire the middle barrel light curtain data to be marked through the middle barrel transverse light curtain 30 at the same time node;

[0098] Detect all human targets in the overhead image, and map the positions of all human targets to the middle barrel light curtain data to be labeled, so as to obtain the labeled middle barrel light curtain data;

[0099] The human posture detection model is trained through several labeled mid-barrel light curtain data.

[0100] It is understandable that since the human eye cannot directly see the position of a person in the barrel light curtain data to be annotated, it is impossible to annotate. In this embodiment, by acquiring the overhead image of the double-door channel and the barrel light curtain data to be annotated at the same time node, using the existing human image detection algorithm to detect all human targets in the overhead image, and implementing the annotation of batches of barrel light curtain data to be annotated by position mapping, both the annotation accuracy and the annotation efficiency are improved.

[0101] Furthermore, in step S3, the method for real-time tracking and detecting the state of the associated pedestrian target includes the steps of:

[0102] Pedestrian detection is performed on the middle barrel light curtain data through the human posture detection model, and the characteristic values ​​of all suspicious pedestrian targets are output;

[0103] Match all suspicious pedestrian target feature values ​​with the pedestrian target feature values ​​in the pedestrian feature database, and filter out the unmatched suspicious pedestrian target feature values;

[0104] Update the tracking position of the pedestrian target according to the matching results;

[0105] The state detection result of the corresponding pedestrian target is output according to the updated tracking position change.

[0106] It is understandable that this embodiment updates the position of pedestrian targets based on the feature value layer, so that multiple pedestrian targets can be tracked synchronously and interference targets can be filtered out. The state detection result of the corresponding pedestrian target is also output according to the tracking position change to detect abnormal pedestrian behavior in time.

[0107] A risk interception system includes a monitoring platform and a double-door channel as described above, through which multiple pedestrians are checked simultaneously. When identity inconsistency is detected by the document recognition unit 70 and the biometric recognition unit 90, the identity detection result is sent to the monitoring platform.

[0108] Compared with the prior art, the risk interception system provided by the present application includes a monitoring platform and a double-door channel as described above. After the monitoring platform receives an identity detection result that is inconsistent with the identity, the abnormal pedestrian is intercepted between the front barrel gate 20 and the rear barrel gate 40, making it convenient for the staff to implement interception at the first time.

[0109] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A double-door channel light curtain pedestrian dynamic detection method, applied to a double-door channel, wherein the double-door channel is sequentially provided with a front barrel transverse light curtain, a front barrel gate, a middle barrel transverse light curtain and a rear barrel gate, characterized in that: The front barrel gate is provided with a first vertical light curtain and a certificate recognition unit, and the rear barrel gate is provided with a biometric recognition unit; the double-door channel light curtain pedestrian dynamic detection method comprises the steps of: The document information is recognized by the document recognition unit, and the front barrel gate is opened after successful recognition; The front bucket light curtain data is obtained through the front bucket horizontal light curtain and the first vertical light curtain, a curve graph of the trigger point number y of the front bucket light curtain data with respect to the time node x is established, and the front bucket gate is closed after matching the human body sensing curve characteristics from the curve graph; Detect pedestrian targets from the front-end data of the middle barrel light curtain obtained from the middle barrel transverse light curtain, associate the pedestrian targets with the certificate information, and track and detect the status of the associated pedestrian targets in real time; The biometric identification unit identifies the identity information of all pedestrians in the horizontal light curtain of the center barrel according to the passage order, and opens the rear barrel gate after the identity information matches the certificate information successfully; The method for matching the human body sensing curve characteristics from the curve graph includes the steps of: Collect the pre-collected data of the light curtain of the human body passing through the front barrel gate at different speeds; Establish a curve graph of the trigger point number y of the light curtain pre-collected data with respect to the time node x, and save it to the human body sensing curve feature database; Match the curve graph established based on the front bucket light curtain data with the human body sensing curve feature database; If there is a curve graph whose average cosine distance is less than the distance threshold, the human body sensing curve feature is matched from the curve graph.

2. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 1, characterized in that: The method for matching a curve graph established based on front bucket light curtain data with a human body sensing curve feature database comprises the steps of: Detect the current speed of pedestrians based on the front barrel transverse light curtain; Filter out a curve graph matching the current passing speed from a human body sensing curve feature database; Then, the curve graph established based on the front bucket light curtain data is matched with the screened human body sensing curve feature database.

3. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 2, characterized in that: The method for detecting the current speed of pedestrians based on the front barrel transverse light curtain comprises the following steps: Based on the front bucket transverse light curtain, obtain the front bucket transverse light curtain data within the time range from when the front bucket gate is opened to when all light spots in the front bucket transverse light curtain are blocked at least once; Define the total number of light spots on the front barrel transverse light curtain as n, where n is a positive even number, T i is the time point when the i-th light spot is blocked for the first time, L is the maximum measured length of the front barrel transverse light curtain, and the calculation formula for the current passing speed V is: 。 4. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 3, characterized in that: The method for matching a curve graph established based on front bucket light curtain data with a screened human body sensing curve feature database comprises the steps of: In addition, y>c, c is a positive integer, and a peak feature F1 in a curve graph established based on the front bucket light curtain data is obtained, and a peak feature set F2 based on a screened human body sensing curve feature database is obtained; The peak feature F1 is matched with the peak feature set F2.

5. The double-door channel light curtain pedestrian dynamic detection method according to any one of claims 1 to 4, characterized in that: The middle barrel horizontal light curtain includes an upper row of light curtains and a lower row of light curtains. The upper row of light curtains is used to sense the upper body of a human body, and the lower row of light curtains is used to sense the lower body of a human body.

6. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 5, characterized in that: The method for detecting a pedestrian target from the front segment data of the middle barrel light curtain data obtained by the middle barrel transverse light curtain comprises the steps of: Train a human posture detection model based on light curtain detection data; Intercept the front-end data of the middle barrel light curtain data; Inputting the preceding data of the continuous multiple frames into a human posture detection model to detect pedestrian targets; The pedestrian targets detected in each frame are converted into pedestrian target feature values ​​and saved in the pedestrian feature library.

7. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 6, characterized in that: The training method of the human posture detection model comprises the steps of: Acquire a top view image of the double-door passage based on a top-down perspective in a real environment, and acquire the middle barrel light curtain data to be annotated through the middle barrel transverse light curtain at the same time node; Detect all human targets in the overhead image, and map the positions of all human targets to the middle barrel light curtain data to be labeled, so as to obtain the labeled middle barrel light curtain data; The human posture detection model is trained through several labeled mid-barrel light curtain data.

8. The double-door channel light curtain pedestrian dynamic detection method as claimed in claim 7, characterized in that: The method for real-time tracking and detecting the state of the associated pedestrian target comprises the steps of: Pedestrian detection is performed on the middle barrel light curtain data through the human posture detection model, and the characteristic values ​​of all suspicious pedestrian targets are output; Match all suspicious pedestrian target feature values ​​with the pedestrian target feature values ​​in the pedestrian feature database, and filter out the unmatched suspicious pedestrian target feature values; Update the tracking position of the pedestrian target according to the matching results; The state detection result of the corresponding pedestrian target is output according to the updated tracking position change.

9. A risk interception system, based on the double-door channel light curtain pedestrian dynamic detection method according to any one of claims 1 to 8, characterized in that: It includes a monitoring platform and a double-door channel, through which multiple pedestrians are checked at the same time. When the identity is inconsistent detected by the document recognition unit and the biometric recognition unit, the identity detection result is sent to the monitoring platform.