An AI vision-based static electricity release behavior monitoring method

Abstract

This invention relates to an AI vision-based method for monitoring static electricity release behavior. It solves the technical problem of lacking a method for monitoring whether static electricity is being released and whether the release behavior is standardized in industrial safety production. The method includes the following steps: 1) Initializing parameters and starting the algorithm service; 2) Acquiring scene images; 3) Sending the scene images to the algorithm detector for detection; 4) Deleting personnel data information not in the static electricity release area; 5) Obtaining tracking numbers; 6) Binding hand detection boxes to corresponding human body detection boxes; 7) Performing minimum pixel filtering; 8) Determining whether the remaining personnel are entering or leaving the workshop; 9) Obtaining the determination result of whether the remaining personnel are releasing static electricity; 10) Obtaining the determination result of whether the remaining personnel have completed releasing static electricity; 11) Obtaining the determination result of whether the remaining personnel have left the static electricity release area; 12) Determining whether an alarm is triggered; 13) Repeating steps 2) to 12) until the monitoring period ends.

Description

Technical Field

[0001] This invention relates to an intelligent monitoring method for industrial safety production, specifically to a method for monitoring static electricity release behavior based on AI vision. Background Technology

[0002] Safety in production is the foundation and prerequisite of all work. In many industrial production processes, releasing static electricity from the human body before entering the production workshop is an important safety requirement, especially in some electronics workshops. Personnel entering the workshop need to pass through a designated static electricity release area at the entrance and touch the static eliminator next to that area to release static electricity. Currently, there are various types of static eliminators used, among which the cylindrical static sphere is the most common. (See [link to relevant documentation]). Figure 2 Personnel entering the workshop can release static electricity by touching the static ball 11. Compared with other static elimination devices, this type of device has the advantages of low cost and convenient and simple use.

[0003] However, the following problems exist in the process of releasing static electricity:

[0004] 1. Due to the lack of safety awareness among some workers, the workers rely heavily on supervision to release static electricity by touching the static ball before entering the workshop. Without supervision, they will not actively release static electricity and will directly enter the workshop.

[0005] 2. Improper use of columnar electrostatic balls. Touching the electrostatic ball 11 requires at least 3 seconds, but in actual use, many operators do not strictly follow this standard.

[0006] 3. Currently, apart from the human supervision of regulatory personnel, there is a lack of intelligent video methods to alert and capture violations such as failure to release static electricity or improper release of static electricity. Summary of the Invention

[0007] The purpose of this invention is to solve the technical problem in existing industrial safety production where, apart from the human supervision of supervisors, there is a lack of intelligent and automated monitoring methods for whether static electricity is released and whether the release behavior is standardized. The invention provides a static electricity release monitoring method based on AI vision.

[0008] The concept of this invention is to use AI vision technology to analyze the behavior of personnel releasing static electricity, thereby providing real-time alarms and automatic capture of violations by personnel who enter the work area without releasing static electricity or without properly releasing static electricity, so as to improve the safety awareness of workers and provide technical support for safety production management.

[0009] The technical solution of this invention:

[0010] A method for monitoring static electricity release behavior based on AI vision, characterized by the following steps:

[0011] 1) Initialize parameters and start the algorithm service;

[0012] The initialization parameters refer to setting the scene monitoring time period; setting the calibration position of the electrostatic ball on the electrostatic device; setting the minimum human body detection frame size threshold; setting the specified electrostatic discharge area in the scene; and setting the threshold for the duration of electrostatic discharge by personnel.

[0013] 2) Obtain the scene image at time t within the monitoring period;

[0014] The scene image includes a designated electrostatic discharge area at the workshop entrance and an electrostatic device next to the designated electrostatic discharge area inside the workshop.

[0015] 3) Send the scene image at time t to the algorithm detector for detection. If a person is detected, obtain the person's data information and execute step 4); if no person is detected, return to step 2).

[0016] The data information includes human body detection boxes, human body detection box confidence scores, human hand detection boxes, and human hand detection box confidence scores.

[0017] 4) Delete the data of personnel who are not in the electrostatic discharge area through area filtering; if there are still personnel remaining after deletion, proceed to step 5); if there are no personnel remaining, return to step 2).

[0018] 5) Obtain the tracking numbers of the remaining personnel through an algorithm tracker;

[0019] 6) Bind the hand detection frames of the remaining personnel to the corresponding human body detection frames;

[0020] 7) Filter the remaining people by the minimum pixel of the human detection box. If there are still people remaining after filtering, proceed to step 8). If there are no people remaining, return to step 2).

[0021] 8) Determine whether the remaining personnel enter or leave the workshop using the frame difference method, and record and save the determination results;

[0022] 9) Determine whether the remaining person is releasing static electricity by checking whether the hand detection frame overlaps with the calibrated position of the static ball;

[0023] 10) Obtain the determination result of whether the remaining personnel have completed the release of static electricity through static electricity release behavior analysis;

[0024] 11) By identifying whether the same tracking number human detection box appears in the previous and next frame images, the determination result of whether the remaining person has left the electrostatic discharge area can be obtained;

[0025] 12) Based on the judgment results of steps 8), 10), and 11), if the personnel with the same tracking number simultaneously meet the following conditions: entering the workshop, failing to complete the static electricity release, and leaving the static electricity release area, then a voice alarm is triggered and the scene image at that moment is saved; otherwise, no alarm is triggered and step 13) is executed.

[0026] 13) Repeat steps 2) to 12) until the monitoring period ends.

[0027] Furthermore, in step 3), the algorithm detectors are YOLOv5, YOLOv6, and YOLOv7; in step 5), the algorithm tracker is DeepSort.

[0028] Furthermore, the training method for the algorithm detector weights is as follows: collect scene images of 10 workshop entrances, and collect 500 sample images of each workshop entrance in three time periods: morning, noon and evening; then label the sample images, labeling the targets as human body and hand; input the labeled sample images into the algorithm detector for training to obtain training weights.

[0029] Further, in step 4), the area filtering is as follows: select the midpoint of the lower edge of the human body detection frame as the reference point, and then use the lead line method to determine whether the reference point is within the closed polygon formed by the electrostatic release area. If it is, the person is considered to be within the electrostatic release area; if not, the person is considered to be outside the specified electrostatic release area, and the person is filtered out.

[0030] Further, the binding method described in step 6) is as follows: calculate the overlap area between the hand detection box and each human body detection box, and bind the hand detection box to the human body detection box corresponding to the largest overlap area; if the hand detection box overlaps with multiple human body detection boxes at the same time, determine the distance between the center position of the hand detection box and the lower edge of each human body detection box surrounding it, and bind the hand detection box to the human body detection box with the largest distance.

[0031] Further, in step 7), the method for minimum pixel filtering is as follows: compare the size of the human body detection box with the minimum human body detection box size threshold set in step 1). If the size of the human body detection box is less than the minimum human body detection box size threshold, delete the person's data information; if it is greater than or equal to the threshold, do not delete it.

[0032] Furthermore, in step 8), the frame difference method determines the direction of human movement based on the displacement of the human detection box at time t and time t-1.

[0033] Further, step 9) specifically involves: calculating whether the hand detection frame bound to the human body detection frame overlaps with the calibrated position of the electrostatic ball set in step 1). If there is an overlap, it is determined that the corresponding person in the specified electrostatic release area is touching the electrostatic ball to release static electricity, and the determination result of releasing static electricity and the timestamp at time t are bound to the human body detection frame; otherwise, it is determined that the corresponding person has not released static electricity, and the determination result of not releasing static electricity is bound to the human body detection frame.

[0034] Further, in step 10), the electrostatic discharge behavior analysis method includes the following steps:

[0035] 10.1) Determine whether the duration of the person correctly releasing static electricity in the current scene image has reached the set value; if yes, update the person's human detection box and hand detection box, and proceed to step 11); if no, proceed to step 10.2);

[0036] 10.2) Determine if this is the first time the person has released static electricity. If yes, initialize the person's timestamp, set the person's status to "temporary," and output the result indicating that the static electricity release was not completed. Proceed to step 11); otherwise, proceed to step 10.3).

[0037] 10.3) The following three situations shall be handled:

[0038] If the person has not released static electricity and this is not the first time, update the person's human body detection frame and hand detection frame, initialize its timestamp, and output the judgment result that static electricity has not been released, and execute step 11);

[0039] If the person has not yet released static electricity and this is the first time this has happened, then output the judgment result that the static electricity release has not been completed, and proceed to step 11);

[0040] If the person is currently releasing static electricity and this is not the first time, proceed to step 10.4);

[0041] 10.4) Update the human body detection box and hand detection box for this person and obtain the current timestamp;

[0042] 10.5) Calculate the difference between the current timestamp and the timestamp of the person's last appearance. Compare the timestamp difference with a set value. If the difference is less than the set value, set the status to "temporary" and proceed to step 11). If the difference is greater than or equal to the set value, set the person's status to "qualified" and output the result of completing the static electricity release. Proceed to step 11).

[0043] Further, the judgment method in step 11) is as follows: if the tracking number bound to the human body detection box appears at time t-1 but does not appear at time t, it is determined that the corresponding person has walked out of the electrostatic discharge area; otherwise, the corresponding person has not walked out of the electrostatic discharge area. If the current time t is the first frame of the scene image, that is, there is no detection result at time t-1, then the judgment result of the human body detection box at the current time t is initialized to not having walked out.

[0044] The beneficial effects of this invention are:

[0045] 1. This invention provides an AI vision-based method for monitoring static electricity release behavior. It utilizes AI vision technology to monitor the static electricity release behavior of personnel entering a workshop. By capturing scene images through cameras and then analyzing these images using algorithms, it promptly alerts and reminds violators and allows for the tracing of violations. It also supports the analysis of complex situations such as multiple people releasing static electricity simultaneously, achieving intelligent and automated monitoring, reducing reliance on manual labor, improving workers' safety awareness, and providing technical support for safe production management.

[0046] 2. This invention provides an AI-based vision-based method for monitoring static electricity release behavior. A violation must be determined simultaneously if the following conditions are met: a) the person is entering the workshop; b) the person is not releasing static electricity or is not releasing it correctly; c) the person has left the static electricity release area. This invention fully considers the complexity of the scenario and the dynamic nature of personnel, ensuring accurate and reliable judgment results.

[0047] 3. The present invention provides an AI vision-based method for monitoring static electricity release behavior, which provides correct static electricity release specifications and supports the judgment of static electricity release duration, thereby effectively improving the reliability of static electricity release.

[0048] 4. The present invention provides an AI vision-based method for monitoring static electricity release behavior. It utilizes the target detection algorithms YOLOv5, YOLOv6, and YOLOv7 and the multi-target tracking algorithm DeepSort in the field of artificial intelligence. It has high detection accuracy, fast speed, and good tracking effect, and can effectively reduce interference from personnel occlusion.

[0049] 5. The present invention provides an AI vision-based method for monitoring static electricity release behavior. By filtering personnel within a specified static electricity release area, non-target personnel are effectively excluded, ensuring that only personnel entering and leaving the specific area are monitored.

[0050] 6. The present invention provides a method for monitoring static electricity release behavior based on AI vision, which effectively eliminates abnormal samples by filtering the minimum pixel of the human body detection box.

[0051] 7. The present invention provides an AI vision-based method for monitoring static electricity release behavior. It determines whether a person has touched the static electricity device by judging whether the center of the human hand detection frame overlaps with the calibrated position of the static electricity device. This method is simple and efficient.

[0052] 8. The present invention provides a method for monitoring static electricity release behavior based on AI vision, which can judge the movement status of personnel through frame difference method, thereby helping to improve the accuracy and reliability of the judgment results. Attached Figure Description

[0053] Figure 1 This is a layout diagram of the camera, the designated electrostatic discharge area, and the columnar electrostatic sphere in this invention;

[0054] Figure 2 This is a schematic diagram of the columnar electrostatic sphere structure in this invention;

[0055] Figure 3 This is a flowchart illustrating the vision-based intelligent monitoring technology for static electricity release in this invention.

[0056] Figure 4 This is a schematic diagram of the displacement deviation changes in four application scenarios of personnel entering the workshop in the four application scenarios of this invention;

[0057] Figure 5 This is a flowchart of the electrostatic discharge behavior analysis in this invention;

[0058] Reference numerals: 1-Columnar electrostatic ball device, 11-Electrostatic ball, 12-Base, 13-Warning sign, 2-Designated electrostatic discharge area, 3-Camera. Detailed Implementation

[0059] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0060] The scene and layout at the workshop entrance in an embodiment of the method of the present invention are as follows: Figure 1 As shown, a columnar electrostatic ball device 1 is fixed at the workshop entrance, and a designated electrostatic release area 2 is planned at the workshop entrance to ensure that all personnel entering the workshop pass through this area. A camera 3 is installed to ensure that it can capture clear and complete images of the workshop entrance. The structure of the columnar electrostatic ball device 1 is generally as follows: Figure 2 As shown, it includes an electrostatic ball 11, a base 12, and a warning sign 13.

[0061] This invention provides an AI-based vision-based method for monitoring static electricity release behavior. The method analyzes captured scene images using an algorithm to determine if personnel have violated regulations. If a violation is detected, a voice alarm is triggered, and the violation scene image is saved. Personnel entering the workshop must press and hold the static electricity ball 11 within the designated static electricity release area 2 for at least 3 seconds, then exit the designated static electricity release area 2 and enter the workshop. Otherwise, it is considered a violation. The algorithm determines a violation based on the following three conditions simultaneously: a) the personnel's movement direction is "entering the workshop"; b) the personnel's static electricity release status is "incomplete"; c) the personnel have "left" the static electricity release area. The specific method flow is as follows: Figure 3 As shown:

[0062] 1) Initialize parameters and start the algorithm service;

[0063] Initialization parameters include setting the scene monitoring time period, setting the calibration position of the electrostatic ball on the electrostatic device, setting the minimum human detection frame size threshold, setting the specified electrostatic discharge area in the scene, and setting the threshold for the duration of electrostatic discharge by personnel. After setting, the algorithm service is started. In this embodiment, the default scene monitoring time period is from 8:00 AM to 6:00 PM on weekdays.

[0064] 2) Acquire scene images at time t within the monitoring period using a camera;

[0065] The monitoring time period t can be any time within the working time period. It is agreed that only 3 frames of scene images will be acquired within 1 second. The scene images include the designated electrostatic discharge area at the entrance of the workshop and the electrostatic device next to the designated electrostatic discharge area in the workshop.

[0066] 3) The scene image at time t is fed into the algorithm detector for detection. If a person is detected, the person's data information is obtained, and step 4) is executed; if no person is detected, the process returns to step 2). The person's data information includes the human body detection box, the confidence score of the human body detection box, the hand detection box, and the confidence score of the hand detection box. The algorithm detector can be YOLOv5, YOLOv6, or YOLOv7; this embodiment uses YOLOv5, requiring custom weight training. The method for training the algorithm detector weights is as follows: Scene images of 10 workshop entrances are collected, with 500 sample images collected for each workshop entrance during three time periods (morning, noon, and evening); the sample images are then labeled with the targets being human bodies and hands; the labeled sample images are input into the algorithm detector for training to obtain the training weights.

[0067] 4) Traverse all human detection frames and delete the data of personnel not in the electrostatic discharge area through area filtering; if there are still personnel remaining after deletion, proceed to step 5); if there are no personnel remaining, return to step 2).

[0068] The area filtering is as follows: Select the midpoint of the lower edge of the human body detection box as the reference point, and then use the lead line method to determine whether the reference point is within the closed polygon formed by the electrostatic discharge area. If it is, the person is considered to be within the electrostatic discharge area; otherwise, the person is considered to be outside the specified electrostatic discharge area, and the person is filtered out.

[0069] 5) Iterate through the human detection boxes of all remaining personnel, and feed the human detection boxes and their confidence scores into the algorithm tracker to obtain the tracking numbers of the remaining personnel. The algorithm tracker is DeepSort, which can use open-source weights or train its own weights. In other embodiments, the algorithm tracker also supports the use of other mainstream multi-object tracking algorithms.

[0070] 6) Iterate through all remaining personnel's hand detection frames and bind them to the corresponding human detection frames. The binding method is as follows: calculate the overlap area between the hand detection frame and each human detection frame, and bind the hand detection frame to the human detection frame corresponding to the largest overlap area; if the hand detection frame overlaps with multiple human detection frames simultaneously, determine the distance between the center position of the hand detection frame and the bottom edge of each human detection frame surrounding it, and bind the hand detection frame to the human detection frame with the largest distance. This is because, relative to the camera, a human being at a larger distance is always in front of a human being at a smaller distance in spatial information, or it can be understood that a human being at a smaller distance is occluded by a human being at a larger distance, which is consistent with actual human body structure analysis.

[0071] 7) Iterate through the human detection boxes of all remaining personnel, and filter the remaining personnel based on the minimum pixel size of the human detection box. If there are still remaining personnel after filtering, proceed to step 8); otherwise, return to step 2). The filtering method is as follows: compare the size of the human detection box with the minimum human detection box size threshold set in step 1). If the size of the human detection box is less than the minimum threshold, delete the human detection box, its confidence level, its tracking number, the hand detection boxes bound to it, and their confidence levels. If the size is greater than or equal to the minimum threshold, do not delete it.

[0072] 8) Traverse all human detection boxes, determine whether the remaining personnel are entering or leaving the workshop using the frame difference method, record and save the determination results, and bind the determination results such as "entering the workshop" or "leaving the workshop" to the human detection boxes. If the current time t is the first frame of the scene image acquired, i.e., there is no detection result at time t-1, then initialize the human movement direction at the current time t to "unknown". The frame difference method determines the human movement direction based on the displacement bias of the human detection boxes at time t and t-1. Figure 4In the scenario shown in (a), the difference in the Y-coordinate of the midpoint of the lower edge of the human detection box corresponding to the same tracking number at time t and time t-1 is compared. If the difference is greater than or equal to 0, the movement state is determined to be entering the workshop; otherwise, it is leaving the workshop. Figure 4 The determination method for (a) is as follows: Figure 4 In the scenario shown in (b), if the difference is less than 0, the movement state is determined to be entering the workshop; otherwise, it is leaving the workshop. For Figure 4 In the scenario shown in (c), the difference in the X-coordinate of the midpoint of the lower edge of the human detection box corresponding to the same tracking number at time t and time t-1 is compared. If the difference is greater than 0, the movement state is determined to be entering the workshop; otherwise, it is leaving the workshop. Figure 4 The determination method for (c) is as follows: Figure 4 In the scenario shown in (d), if the difference is less than 0, the movement state is determined to be entering the workshop; otherwise, it is leaving the workshop.

[0073] 9) Traverse all human body detection frames and calculate whether the hand detection frame bound to the human body detection frame overlaps with the calibration position of the electrostatic ball set in step 1). Obtain the determination result of whether the remaining person is releasing static electricity. If there is an overlap, it is determined that the corresponding person in the specified static electricity release area is touching the electrostatic ball to release static electricity, and the determination result of "releasing static electricity" and the timestamp at time t are bound to the human body detection frame. Otherwise, it is determined that the corresponding person has not released static electricity, and only the determination result of "not releasing static electricity" is bound to the human body detection frame.

[0074] 10) Traverse all human body detection frames. Based on step 9), determine whether the person has completed the release of static electricity using the static electricity release behavior analysis method. Bind the "completed" or "not completed" judgment result to the human body detection frame to obtain the judgment result of whether the release of static electricity has been completed.

[0075] The electrostatic discharge behavior analysis method includes the following steps:

[0076] 10.1) Determine whether the duration of the person correctly releasing static electricity in the current scene image has reached the set value; if yes, update the person's human detection box and hand detection box, and proceed to step 11); if no, proceed to step 10.2);

[0077] 10.2) Determine if this is the first time the person has released static electricity. If yes, initialize the person's timestamp, set the person's status to "temporary," and output the result indicating that the static electricity release was not completed. Proceed to step 11); otherwise, proceed to step 10.3).

[0078] 10.3) The following three situations shall be handled:

[0079] If the person has not released static electricity and this is not the first time, update the person's human body detection frame and hand detection frame, initialize its timestamp, and output the judgment result that static electricity has not been released, and execute step 11);

[0080] If the person has not yet released static electricity and this is the first time this has happened, then output the judgment result that the static electricity release has not been completed, and proceed to step 11);

[0081] If the person is currently releasing static electricity and this is not the first time, proceed to step 10.4);

[0082] 10.4) Update the human body detection box and hand detection box for this person and obtain the current timestamp;

[0083] 10.5) Calculate the difference between the current timestamp and the timestamp of the person's last appearance. Compare the timestamp difference with a set value. If the difference is less than the set value, set the status to "temporary" and proceed to step 11). If the difference is greater than or equal to the set value, set the person's status to "qualified" and output the result of completing the static electricity release. Proceed to step 11).

[0084] 11) Traverse all human detection boxes. By identifying whether the same tracking number human detection box appears in the preceding and following frames, obtain the determination result of whether the remaining person has left the electrostatic discharge area. Bind the determination result of "left" or "not left" to the human detection box. The determination method is as follows: if the tracking number bound to the human detection box appears at time t-1 but does not appear at time t, it is determined that the corresponding person has left the electrostatic discharge area; otherwise, the corresponding person has not left the electrostatic discharge area. If the current time t is the first frame of the scene image, that is, there is no detection result at time t-1, then the determination result of the human detection box at the current time t is initialized to "not left".

[0085] 12) Traverse all human detection frames. Based on the judgment results of steps 8), 10), and 11), if a person with the same tracking number simultaneously meets the following conditions: entering the workshop, not completing static electricity release, and leaving the static electricity release area, then trigger a voice alarm and save the scene image at that moment; otherwise, do not trigger an alarm and proceed to step 13). In practical applications, the judgment method is as follows: if the movement direction of the person bound to the human detection frame in step 8) is "entering the workshop," the static electricity release status of the person bound to the human detection frame in step 10) is "not completed," and the person bound to the human detection frame in step 11) has "left" the static electricity release area, then trigger an alarm at time t. Otherwise, do not trigger an alarm; if an alarm is determined, trigger a voice alarm, return the result, and save the scene image at time t. The returned result includes the visualized scene image, whether an alarm was triggered, and the person who violated the rules. The visualized scene image refers to the visualization of the human detection frame and the hand detection frame in the scene image.

[0086] 13) Repeat steps 2) to 12) until the scene monitoring time period set in step 1 is exceeded, then stop the service.

Claims

1. A method for monitoring static electricity release behavior based on AI vision, characterized in that, Includes the following steps: 1) Initialize parameters and start the algorithm service; The initialization parameters refer to setting the scene monitoring time period and setting the calibration position of the electrostatic ball on the electrostatic device. Set the minimum human detection frame size threshold; Set a specified area for static electricity release in the scene; set a threshold for the duration of static electricity release by personnel. 2) Obtain the scene image at time t within the monitoring period; The scene image includes a designated electrostatic discharge area at the workshop entrance and an electrostatic device next to the designated electrostatic discharge area inside the workshop. 3) Send the scene image at time t to the algorithm detector for detection. If a person is detected, obtain the person's data information and execute step 4); if no person is detected, return to step 2). The data information includes human body detection boxes, human body detection box confidence scores, human hand detection boxes, and human hand detection box confidence scores. 4) Delete personnel data information that are not in the electrostatic discharge area through area filtering; if there are still personnel remaining after deletion, proceed to step 5); if there are no remaining personnel, return to step 2). 5) Obtain the tracking numbers of the remaining personnel through an algorithm tracker; 6) Bind the hand detection frames of the remaining personnel to the corresponding human body detection frames; Calculate the overlap area between the hand detection box and each human body detection box, and bind the hand detection box to the human body detection box corresponding to the largest overlap area; If a hand detection box overlaps with multiple human detection boxes at the same time, the distance between the center position of the hand detection box and the bottom edge of each human detection box surrounding it is determined, and the hand detection box is bound to the human detection box with the largest distance. 7) Filter the remaining people using the minimum pixel size of the human detection bounding box. If there are still people remaining after filtering, proceed to step 8). If there are no people remaining, return to step 2). The minimum pixel filtering method is as follows: compare the size of the human body detection box with the minimum human body detection box size threshold set in step 1). If the size of the human body detection box is less than the minimum human body detection box size threshold, delete the person's data information; if it is greater than or equal to the minimum human body detection box size threshold, do not delete it. 8) Determine whether the remaining personnel enter or leave the workshop using the frame difference method, and record and save the determination results; 9) Determine whether the remaining person is releasing static electricity by checking whether the hand detection frame overlaps with the calibrated position of the static ball; 10) Obtain the determination result of whether the remaining personnel have completed the release of static electricity through static electricity release behavior analysis; 11) By identifying whether the same tracking number human detection box appears in the preceding and following frames, the determination result of whether the remaining person has left the electrostatic discharge area can be obtained; 12) Based on the judgment results of steps 8), 10), and 11), if the personnel with the same tracking number simultaneously meet the following conditions: entering the workshop, failing to complete the static electricity release, and leaving the static electricity release area, then a voice alarm is triggered and the scene image at that moment is saved; otherwise, no alarm is triggered, and step 13) is executed. 13) Repeat steps 2) to 12) until the monitoring period ends.

2. The method for monitoring static electricity release behavior based on AI vision according to claim 1, characterized in that, In step 3), the algorithm detectors are YOLOv5, YOLOv6, and YOLOv7; In step 5), the algorithm tracker is DeepSort.

3. The method for monitoring static electricity release behavior based on AI vision according to claim 2, characterized in that, The training method for the algorithm detector weights is as follows: Scene images were collected at the entrances of 10 workshops, with 500 sample images collected at each workshop entrance during three time periods: morning, noon, and evening. Then annotate the sample images, labeling the targets as human bodies and hands; The labeled sample images are input into the algorithm detector for training to obtain training weights.

4. A method for monitoring static electricity release behavior based on AI vision according to claim 1, 2, or 3, characterized in that, In step 4), the area filtering is as follows: select the midpoint of the lower edge of the human body detection frame as the reference point, and then use the lead line method to determine whether the reference point is within the closed polygon formed by the electrostatic discharge area. If it is, the person is considered to be within the electrostatic discharge area; if not, the person is considered to be outside the specified electrostatic discharge area, and the person is filtered out.

5. The method for monitoring static electricity release behavior based on AI vision according to claim 4, characterized in that, In step 8), the frame difference method determines the direction of human movement based on the displacement of the human detection box at time t and time t-1.

6. The method for monitoring static electricity release behavior based on AI vision according to claim 5, characterized in that, Step 9) Specifically: Calculate whether the hand detection frame bound to the human body detection frame overlaps with the calibration position of the electrostatic ball set in step 1). If there is an overlap, it is determined that the corresponding person in the specified electrostatic release area is touching the electrostatic ball to release static electricity. Then, the determination result of releasing static electricity and the timestamp at time t are bound to the human body detection frame. Otherwise, if the person is deemed not to have released static electricity, the result of not releasing static electricity will be bound to the human body detection frame.

7. The method for monitoring static electricity release behavior based on AI vision according to claim 6, characterized in that, In step 10), the electrostatic discharge behavior analysis method includes the following steps: 10.1) Determine whether the duration of static electricity release by the person in the current scene image has reached the set value; if yes, update the person's body detection box and hand detection box, and proceed to step 11); if no, proceed to step 10.2). 10.2) Determine if this is the first time the person has released static electricity. If yes, initialize the person's timestamp, set the person's status to "provisional," and output the result indicating that static electricity release was incomplete. Proceed to step 11). If no, proceed to step 10.3). 10.3) The following three situations shall be handled: If the person has not yet released static electricity and this is not the first time this has happened, update the person's body detection frame and hand detection frame, initialize their timestamp, and output the judgment result that static electricity has not been released, then proceed to step 11). If the person has not yet released static electricity and this is the first time this has happened, then output a result indicating that static electricity release has not been completed, and proceed to step 11). If the person is currently releasing static electricity and this is not the first time, proceed to step 10.4). 10.4) Update the human body detection box and hand detection box for this person and obtain the current timestamp; 10.5) Calculate the difference between the current timestamp and the timestamp of the person's last appearance. Compare the timestamp difference with a set value. If it is less than the set value, set the status to "temporary" and proceed to step 11. If it is greater than or equal to the set value, set the person's status to "qualified" and output the judgment result of "static discharge completed". Proceed to step 11.

8. The method for monitoring static electricity release behavior based on AI vision according to claim 7, characterized in that, Step 11) The judgment method is as follows: if the tracking number bound to the human body detection frame appears at time t-1 but does not appear at time t, it is determined that the corresponding person has walked out of the electrostatic discharge area; otherwise, the corresponding person has not walked out of the electrostatic discharge area. If the current time t is the first frame of the scene image acquired, i.e. there is no detection result at time t-1, then the judgment result of the human body detection box at the current time t is initialized to "not out of bounds".

Images