Climbing behavior analysis method and system and related equipment

Abstract

The invention relates to the field of machine vision, in particular to a climbing behavior analysis method and system and related equipment. The method comprises the steps of selecting a target frame image based on a video image, obtaining detection frame information containing the position, size and confidence of a target detection frame from the target frame image, screening out a first target detection frame with the confidence higher than a preset confidence threshold, matching the first target detection frame with a second target detection frame with the highest similarity in an original track, and calculating an IoU distance score, if the score is greater than the preset score, updating the target tracking trajectory; judging whether the target person is in a climbing area based on the target tracking trajectory, if so, identifying body key points of the target person, and selecting a plurality of target key points for attitude calculation; whether the calculated value meets a climbing threshold value or not is judged, if yes, the climbing behavior is recognized, and a third-level alarm is given out; the invention further comprises a corresponding analysis system, electronic equipment and a storage medium. According to the invention, accurate identification, rapid response and early warning can be carried out on the climbing behavior, and the safety guarantee and monitoring efficiency are enhanced.

Description

Technical Field

[0001] This application relates to the field of machine vision, and in particular, to a climbing behavior analysis method, system and related devices. Background Art

[0002] To ensure the safety of ship navigation, there are many restricted areas on ships, such as engine rooms, bridges, cargo holds, and areas where fuel or other dangerous goods are stored. In addition, to prevent people from falling into the sea, the edges of the deck are usually restricted areas. To prevent people from approaching these areas, guardrails or other obstacles are usually set around these areas. However, there are still people who climb over these obstacles for some reasons, resulting in safety accidents. Currently, ships mainly monitor key areas through video surveillance. Since climbing behavior relies on manual judgment and recognition, and there are usually many monitored areas, it is easy for managers to miss or overlook situations in a timely manner.

[0003] Therefore, how to design a behavior analysis method that can automatically analyze human climbing behavior, discover it in a timely manner and quickly inform the staff to take corresponding measures is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0004] The purpose of this application is to overcome the above technical problems and provide a climbing behavior analysis method, system and related devices.

[0005] In a first aspect, an embodiment of this application discloses a climbing behavior analysis method, which adopts the following scheme: A climbing behavior analysis method includes: selecting a target frame image based on a video image, and obtaining detection box information from the target frame image, where the detection box information includes the positions, sizes and confidences of at least one target detection box, and a target person is calibrated in the target detection box; comparing the confidences of the at least one target detection box with a preset confidence threshold, and screening out the target detection boxes with confidences higher than the preset confidence threshold as the first target detection boxes; selecting the second target detection box with the highest similarity to the first target detection box from the detection box queue of the original trajectory for matching, and calculating the IoU distance score based on the positions and sizes of the first target detection box and the second target detection box. If the IoU distance score is greater than a preset score, the first target detection box is incorporated into the detection box queue of the original trajectory to obtain an updated target tracking trajectory, where at least one detection box is included in the original trajectory; judging whether the area where the target person is currently located is a climbing area based on the target tracking trajectory. If so, identifying the key points of the target person's body, selecting multiple target key points from the key points for pose calculation to obtain calculation values; judging whether the calculation values meet the climbing threshold. If so, identifying a climbing behavior and issuing a level-three alarm.

[0006] By adopting the above technical solution, selecting a target frame image based on a video image and obtaining detection box information can clarify the position, size, and confidence of at least one target detection box, accurately calibrate the target person; comparing the confidence of the target detection box with a preset confidence threshold and screening out the first target detection box can exclude detection boxes with low confidence and improve the accuracy of subsequent analysis; selecting the second target detection box with the highest similarity from the original trajectory detection box queue for matching and calculating the IoU distance score, and then updating the target tracking trajectory can continuously and accurately track the action trajectory of the target person; judging whether the area where the target person is located is a climbing area based on the target tracking trajectory, then identifying key points and performing pose calculation to obtain a calculated value can effectively judge the action pose of the target person in a specific area; judging whether the calculated value meets the climbing threshold, thereby identifying the climbing behavior and issuing a level-three alarm can timely discover and warn of dangerous climbing behaviors and ensure safety.

[0007] Optionally, comparing the confidence of the at least one target detection box with a preset confidence threshold, screening out the target detection box higher than the preset confidence threshold as the first target detection box further includes: regarding the target detection box lower than the preset confidence threshold as the third target detection box; wherein, if the IoU distance score calculated based on the positions and sizes of the first target detection box and the second target detection box is less than the preset score, then matching the third target detection box with the fourth target detection box with the highest similarity, calculating the IoU distance score based on the positions and sizes of the third target detection box and the fourth target detection box, and if the IoU distance score is greater than the preset score, incorporating the third target detection box into the original trajectory to obtain an updated target tracking trajectory.

[0008] By adopting the above technical solution, setting the target detection box lower than the preset confidence threshold as the third target detection box, when the IoU distance score between the first target detection box and the second target detection box is less than the preset score, performing matching and IoU distance score calculation on the third target detection box and the fourth target detection box, and if it is greater than the preset score, incorporating the third target detection box into the original trajectory can more comprehensively update the target tracking trajectory, avoid losing the tracking information of the target person due to the low confidence of some target detection boxes, and improve the accuracy and integrity of target tracking.

[0009] Optionally, the calculation formula of the IoU distance score is: ; wherein, the position of the first target detection box / third target detection box is , is the area; the positions of the second target detection box / fourth target detection box are , is the area.

[0010] By adopting the above technical solution, the specific calculation formula of the IoU distance score is clarified, providing a basis for accurately calculating the distance score between the first target detection box / third target detection box and the second target detection box / fourth target detection box, thereby helping to more accurately incorporate the eligible detection boxes into the detection box queue of the original trajectory to update the target tracking trajectory.

[0011] Optionally, there are 18 key points, and 3 of the 18 key points are the target key points; among them, the 18 key points are nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, left ear; the 3 target key points are right hip, right knee, right ankle or left hip, left knee, left ankle.

[0012] By adopting the above technical solution, 18 specific human key points (such as nose, neck, etc.) are clearly selected, and 3 target key points (right hip, right knee, right ankle or left hip, left knee, left ankle) are determined from them for pose calculation, making the pose calculation more targeted and accurate, so as to be able to more accurately judge whether the target person has a climbing behavior.

[0013] Optionally, the pose is obtained by calculating the angles between the knee and the ankle, hip, and the calculation formula of the pose is: ; where θ is the angle between the knee and the ankle, hip, AB is the length between the hip and the knee, BC is the length between the knee and the ankle, and AC is the length between the ankle and the hip.

[0014] By adopting the above technical solution, the angle between the knee and the ankle, hip can be accurately calculated based on the specific pose calculation formula using the lengths between the hip and the knee, knee and the ankle, ankle and the hip, thereby assisting in judging whether the target person has a climbing behavior and improving the accuracy of climbing behavior recognition.

[0015] Optionally, based on the target tracking trajectory, a first-level alarm is issued when the target person is in the first area, and a second-level alarm is issued when the target person is in the second area.

[0016] By adopting the above technical solution, the area where the target person is located can be grasped in time based on the target tracking trajectory. When the target person is in the first area, a first-level alarm is issued, and when the target person is in the second area, a second-level alarm is issued. Gradual warnings can be issued according to the different areas where the target person is located to prevent climbing behavior from occurring in advance.

[0017] Optionally, when the unmatched detection frame in the original trajectory is still not matched within a preset time, it is deleted from the original trajectory.

[0018] By adopting the above technical solution, invalid detection frames that have not been matched for a long time in the original trajectory can be removed to avoid them occupying data storage space and interfering with subsequent target tracking calculations, making the original trajectory data more concise and accurate, which helps to improve the efficiency and accuracy of climbing behavior analysis.

[0019] In a second aspect, an embodiment of the present application discloses a climbing behavior analysis system, which adopts the following scheme: A climbing behavior analysis system comprises: an acquisition module, used for selecting a target frame image based on a video image, and obtaining detection frame information from the target frame image, wherein the detection frame information comprises the position, size and confidence of at least one target detection frame, and a target person is marked in the target detection frame; a comparison module, used for comparing the confidence of the at least one target detection frame with a preset confidence threshold, and screening out a target detection frame having a confidence threshold higher than the preset confidence threshold as a first target detection frame; a selection module, used for selecting a second target detection frame having the highest similarity with the first target detection frame from a detection frame queue of an original trajectory for matching, so as to obtain a target person detected by the first target detection frame based on the confidence of the first target detection frame and the first target detection frame. The position and size of the second target detection frame are calculated to calculate the IoU distance score. If the IoU distance score is greater than the preset score, the first target detection frame is merged into the detection frame queue of the original trajectory to obtain an updated target tracking trajectory, wherein the original trajectory includes at least one detection frame; a first judgment module is used to judge whether the current area where the target person is located is a climbing area based on the target tracking trajectory. If so, the key points of the target person's body are identified, and multiple target key points are selected from the key points for posture calculation to obtain a calculated value; a second judgment module is used to judge whether the calculated value meets the climbing threshold. If so, the climbing behavior is identified and a third-level alarm is issued.

[0020] By adopting the above technical solution, the acquisition module selects the target frame image based on the video image and obtains the detection frame information, which can provide basic data for subsequent analysis, avoid blind detection, and improve analysis efficiency; the comparison module compares the confidence of the target detection frame with the preset confidence threshold and screens out the first target detection frame, which can exclude low-confidence interference information and improve detection accuracy; the selection module updates the target tracking trajectory by matching and calculating the IoU distance score, which can accurately track the movement trajectory of the target person and facilitate real-time grasp of the dynamics of the target person; the first judgment module judges the area where the target person is located according to the target tracking trajectory and performs posture calculation, which can determine whether the target person enters the climbing area and his climbing posture; the second judgment module identifies the climbing behavior by judging whether the calculated value meets the climbing threshold and issues a three-level alarm, which can timely detect dangerous climbing behavior and ensure the safety of related areas.

[0021] In a third aspect, an embodiment of the present application discloses an electronic device, which adopts the following solution: An electronic device comprises: a memory and a processor, wherein the memory is used to store a computer program; and the processor is used to implement the steps of any of the above methods when executing the computer program.

[0022] In a fourth aspect, an embodiment of the present application discloses a computer-readable storage medium, which adopts the following scheme: A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: 1. By obtaining detection frame information based on video images and selecting high-confidence detection frames, the target personnel around the prohibited entry area of ​​the ship can be locked more accurately, avoiding the problem of untimely discovery or omission caused by multiple monitoring areas during manual monitoring of the ship; 2. Using the IoU distance score to match the detection frame to update the target tracking trajectory, the actions of target personnel around the prohibited entry area of ​​the ship can be monitored in real time and comprehensively, making up for the defect that manual monitoring of the ship cannot achieve real-time and comprehensive monitoring; 3. By identifying the key points of the target person's body and calculating the posture, the climbing behavior can be accurately judged and an alarm can be issued. It can promptly detect people who illegally climb over obstacles on the ship, and quickly inform the staff to take corresponding measures to avoid safety accidents caused by climbing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of a process of analyzing a climbing behavior disclosed in an embodiment of the present application; Figure 2 Schematic diagram of detecting frame selection for a target person recognized on a target frame image; Figure 3 Schematic diagram of identifying key points of the target person's body; Figure 4 Schematic diagram of identifying target key points in the intercepted detection frame; Figure 5 Schematic diagram of the area position divided in the video image; Figure 6 Schematic diagram of the structure of an electronic device disclosed in another embodiment of the present application; Figure 7 Schematic diagram of a computer-readable storage medium disclosed in another embodiment of the present application. Detailed implementation manners

[0025] The present application will be further described in detail below with reference to the accompanying drawings.

[0026] The embodiments of the present application will be described in more detail below with reference to the drawings. Although the embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0027] The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The singular forms "a" and "the" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0028] It should be understood that although the terms "first", "second", etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0029] The technical solutions of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0030]

First Embodiment

[0031] The method specifically includes the following steps: S10. Select a target frame image based on the video image and obtain detection box information from the target frame image; Among them, the detection box information includes the position, size, and confidence of at least one target detection box. A target person is calibrated in the first target detection box (see Figure 2 ), and thus the picture without people is directly discarded. The selection of the target frame image can adopt a timed sampling method, such as selecting one frame of image every certain time. For the acquisition of the target detection box, existing target detection algorithms can be used, such as the yolov5 algorithm (a single-stage target detection algorithm). The number of target detection boxes is based on the number of people identified from the target frame image. The position of the target detection box can be represented by the upper left and lower right coordinates of the rectangular box, and the size can be measured by the length and width of the rectangular box. The confidence indicates the possibility that there is a target person in the detection box, and it is calculated by the target detection algorithm according to the image features.

[0032] S20. Compare the confidence of at least one target detection box with the preset confidence threshold, and screen out the target detection boxes with a confidence higher than the preset confidence threshold as the first target detection boxes; Among them, the preset confidence threshold is preset and can be adjusted according to the actual situation. If you want to more strictly screen out certain target people, the threshold can be set higher; conversely, if you want to cover as many possible target people as possible, the threshold can be set lower. By comparing the confidence of the target detection box with the preset confidence threshold in step S20, the more reliable target detection boxes can be preferentially used for the next trajectory tracking and matching, which can not only improve the accuracy of subsequent analysis but also reduce the amount of data processing.

[0033] S30. Select the second target detection box with the highest similarity to the first target detection box from the detection box queue of the original trajectory for matching. Based on the positions and sizes of the first target detection box and the second target detection box, calculate the IoU distance score. If the IoU distance score is greater than the preset score, incorporate the first target detection box into the detection box queue of the original trajectory to obtain an updated target tracking trajectory; Among them, the original trajectory includes at least one detection box. When an image containing a person is recognized for the first time, the person is boxed to automatically serve as the first detection box in the original trajectory. In this step S30, the similarity of the detection boxes is calculated by the IoU (Intersection Over Union) algorithm, and the Hungarian algorithm is used to select the second target detection box with the highest similarity to the first target detection box from the detection box queue of the original trajectory for matching.

[0034] The calculation formula for the IoU distance score is as follows: ; Among them, the position of the first target detection box / the third target detection box is , is the area of; the position of the second target detection box / the fourth target detection box is , is the area of.

[0035] Further, in this embodiment, step S20 includes: regarding the target detection box with a confidence threshold lower than the preset value as the third target detection box; Correspondingly, if the IoU distance score calculated based on the positions and sizes of the first target detection box and the second target detection box in step S30 is less than the preset score, the third target detection box is matched with the fourth target detection box with the highest similarity, and the IoU distance score is calculated based on the positions and sizes of the third target detection box and the fourth target detection box. If the IoU distance score is greater than the preset score, the third target detection box is incorporated into the original trajectory to obtain an updated target tracking trajectory.

[0036] Among them, the fourth target detection box can be the same target detection box as the second target detection box. In this embodiment, in the case where the high-confidence first target detection box fails to match the second target detection box, the low-confidence third target detection box is used to match the fourth target detection box again, so as to avoid losing the tracking information of the target person due to the low confidence of some target detection boxes and improve the accuracy and integrity of target tracking. Correspondingly, if the high-confidence detection box matches successfully, the low-confidence detection box is no longer matched, which can improve the matching efficiency.

[0037] In addition, if the detection box that has not been matched in the original trajectory has not been matched within a preset time, such as 30 frames, it is deleted from the original trajectory, which can avoid too many invalid detection boxes in the original trajectory and reduce the calculation amount and data redundancy.

[0038] S40. Determine whether the current area where the target person is located is a climbing area based on the target tracking trajectory. If so, identify the key points of the target person's body, select multiple target key points from the key points for pose calculation, and obtain the calculated value. Among them, in this embodiment, by calculating the IoU between the detection box of the target person and this area, if it is greater than the set threshold, it is determined that the person is in this climbing area. See Figure 5 , the setting of the climbing area can be pre-divided in the video image, and the position information of the target detection box is used to determine whether the target person enters a certain area.

[0039] See Figure 3 and Figure 4 , identify the key points of the target person's body through the OpenPose - a deep learning-based human pose estimation library. In this embodiment, 18 key points of the body are detected through pre-trained weights, which are set to 18, namely the nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, left ear. Based on the identified key points, select multiple target key points, for example, 3 for pose calculation. The 3 key points are such as the right hip, right knee, right ankle or the left hip, left knee, left ankle.

[0040] Specifically, the pose is obtained by calculating the angles between the knee and the ankle, and the hip. The corresponding pose calculation formula is: ; See Figure 4 , θ is the angle between the knee and the ankle, and the hip. AB is the length between the hip and the knee, BC is the length between the knee and the ankle, and AC is the length between the ankle and the hip.

[0041] S50. Determine whether the calculated value meets the climbing threshold. If so, identify the climbing behavior and issue a level-three alarm.

[0042] Among them, the climbing threshold can be set to 90°. When the angle θ between the knee and the ankle, and the hip is less than 90°, it means that the posture of the target person is in a knee-bending state, corresponding to the climbing behavior, and a level-three alarm will be triggered on this system to remind the staff.

[0043] In this embodiment, a hierarchical alarm mechanism is adopted. The corresponding divisions have different levels of early warnings for different fields. For different levels of alarms, different sound, light, or display methods can be used to distinguish, so that the staff can quickly understand the state of the target person.

[0044] See Figure 5When the target person is in the first area, a first-level alarm will be triggered. When the target person is in the second area, a second-level alarm will be triggered. Regarding the division of the area, it can be set according to the specific scenario.

[0045] In summary, a climbing behavior analysis method disclosed in the first embodiment of the present invention monitors the movement trajectory and posture of the target person by processing and analyzing video images and using object detection and tracking technologies. In the detection box screening process, some unreliable detection boxes are excluded through the confidence threshold, improving the accuracy of subsequent analysis. During the trajectory update process, the IoU distance score is used to judge the matching degree between detection boxes, enabling accurate tracking of the movement of the target person. When judging the area and posture, combined with the pre-set climbing area and posture calculation formula, climbing behavior can be accurately identified. The hierarchical alarm mechanism allows the staff to take corresponding measures according to different situations, greatly improving the timeliness and accuracy of ship key area monitoring compared with the traditional manual monitoring method, and reducing the probability of safety accidents.

[0046]

Second Embodiment

[0047] Among them, the acquisition module 210 is used to select a target frame image based on the video image and obtain detection box information from the target frame image. The detection box information includes the position, size, and confidence of at least one object detection box, and the object detection box is marked with the target person; The comparison module 220 is used to compare the confidence of the at least one object detection box with the pre-set confidence threshold, and screen out the object detection box with a confidence higher than the pre-set confidence threshold as the first object detection box; The selection module 230 is used to select the second object detection box with the highest similarity to the first object detection box from the detection box queue of the original trajectory for matching, and calculate the IoU distance score based on the positions and sizes of the first object detection box and the second object detection box. If the IoU distance score is greater than the preset score, the first object detection box is incorporated into the detection box queue of the original trajectory to obtain an updated target tracking trajectory, where the original trajectory includes at least one detection box; The first judgment module 240 is used to judge whether the area where the target person is currently located is a climbing area based on the target tracking trajectory. If so, identify the key points of the target person's body, select multiple target key points from the key points for posture calculation, and obtain a calculated value; The second determination module 250 is configured to determine whether the calculated value meets the climbing threshold. If so, it identifies a climbing behavior and issues a third-level alarm.

[0048] It should be noted that a climbing behavior analysis method implemented by a climbing behavior analysis system disclosed in the second embodiment of the present application is the same as that in the first embodiment, so it will not be described in detail here. Optionally, each module and the above other operations or functions in this embodiment are respectively for implementing the method in the foregoing embodiment.

[0049]

Third Embodiment

[0050] The technical effect of the electronic device provided in this embodiment in actual application is the same as that of the climbing behavior analysis method in the first embodiment.

[0051]

Fourth Embodiment

[0052] In addition, it can be understood that the foregoing embodiments are only exemplary descriptions of the present invention. On the premise that the technical features do not conflict, the structures do not contradict, and the invention purpose of the present invention is not violated, the technical solutions of the various embodiments can be arbitrarily combined and used.

[0053] In several embodiments provided by the present invention, it should be understood that the disclosed methods, systems, and devices can be implemented in other ways. For example, the modules included in the system described above are merely illustrative. The division of modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0054] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] In addition, in each embodiment of the present invention, each functional unit / module can be integrated in a processing unit / module, or each unit / module can exist physically alone, or two or more units / modules can be integrated in one unit / module. The above-mentioned integrated unit / module can be implemented in the form of hardware or in the form of hardware plus software functional unit / module.

[0056] The above-mentioned integrated unit / module implemented in the form of software functional unit / module can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including several instructions to enable one or more processors of a computer device (which can be a personal computer, a server, or a network device, etc.) to execute some steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0057] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for analyzing climbing behavior, characterized in that, Including: Selecting a target frame image based on a video image, and obtaining detection box information from the target frame image, where the detection box information includes the positions, sizes, and confidences of at least one target detection box, and a target person is calibrated in the target detection box; Comparing the confidences of the at least one target detection box with a preset confidence threshold, and screening out the target detection boxes higher than the preset confidence threshold as the first target detection boxes; Selecting the second target detection box with the highest similarity to the first target detection box from the detection box queue of the original trajectory for matching, and calculating the IoU distance score based on the positions and sizes of the first target detection box and the second target detection box. If the IoU distance score is greater than the preset score, incorporating the first target detection box into the detection box queue of the original trajectory to obtain an updated target tracking trajectory, where at least one detection box is included in the original trajectory; Judging whether the area where the target person is currently located is a climbing area based on the target tracking trajectory. If so, identifying the key points of the target person's body, selecting multiple target key points from the key points for pose calculation, and obtaining a calculated value; Judging whether the calculated value meets the climbing threshold. If so, identifying a climbing behavior and issuing a level-three alarm.

2. The method according to claim 1, characterized in that The step of comparing the confidences of the at least one target detection box with a preset confidence threshold and screening out the target detection boxes higher than the preset confidence threshold as the first target detection boxes further includes: Regarding the target detection boxes lower than the preset confidence threshold as the third target detection boxes; Wherein, if the IoU distance score calculated based on the positions and sizes of the first target detection box and the second target detection box is less than the preset score, matching the third target detection box with the fourth target detection box with the highest similarity, calculating the IoU distance score based on the positions and sizes of the third target detection box and the fourth target detection box, and if the IoU distance score is greater than the preset score, incorporating the third target detection box into the original trajectory to obtain an updated target tracking trajectory.

3. The method according to claim 2, wherein The calculation formula of the IoU distance score is: ； Among them, the position of the first target detection box / the third target detection box is , is the area of; the position of the second target detection box / the fourth target detection box is , is the area of.

4. The method according to claim 2, wherein There are 18 key points, and the target key points are 3 of the 18 key points; Among them, the 18 key points are the nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, left ear; the 3 target key points are the right hip, right knee, right ankle or the left hip, left knee, left ankle.

5. The method according to claim 4, wherein The pose is obtained by calculating the angles between the knee and the ankle, and the hip. The calculation formula of the pose is: ； Wherein, θ is the angle between the knee and the ankle, and the hip, AB is the length between the hip and the knee, BC is the length between the knee and the ankle, and AC is the length between the ankle and the hip.

6. The method according to claim 4, wherein Based on the target tracking trajectory, issuing a level-one alarm when the target person is in the first area, and issuing a level-two alarm when the target person is in the second area.

7. The method according to claim 2, characterized in that If the detection box that has not been matched in the original trajectory is still not matched within the preset time, it is deleted from the original trajectory.

8. A climbing behavior analysis system, characterized in that, It includes: An acquisition module, configured to select a target frame image based on a video image and obtain detection box information from the target frame image, where the detection box information includes the positions, sizes, and confidences of at least one target detection box, and a target person is calibrated in the target detection box; A comparison module, configured to compare the confidences of the at least one target detection box with a preset confidence threshold, and screen out the target detection boxes with confidences higher than the preset confidence threshold as the first target detection boxes; A selection module, configured to select a second target detection box with the highest similarity to the first target detection box from the detection box queue of the original trajectory for matching, and calculate the IoU distance score based on the positions and sizes of the first target detection box and the second target detection box. If the IoU distance score is greater than the preset score, the first target detection box is incorporated into the detection box queue of the original trajectory to obtain an updated target tracking trajectory, where at least one detection box is included in the original trajectory; A first judgment module, configured to judge whether the area where the target person is currently located is a climbing area based on the target tracking trajectory. If so, identify the key points of the target person's body, select multiple target key points from the key points for pose calculation, and obtain a calculated value; A second judgment module, configured to judge whether the calculated value meets the climbing threshold. If so, identify a climbing behavior and issue a level-three alarm.

9. An electronic device, characterized in that, It includes: A memory and a processor. The memory is used to store a computer program; the processor is used to implement the steps of the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images