Method for monitoring express sorting stage

Abstract

The invention relates to a monitoring method for an express sorting stage. The monitoring method comprises the following steps: S1, acquiring video stream data of a sorting area in real time; s2, performing image analysis on the video stream data, and identifying express number information; s3, the sorting action of the operator is detected based on a human body posture recognition algorithm; s4, judging whether the current sorting action is illegal or not according to a preset illegal sorting behavior rule; the illegal sorting behavior rule comprises an action specification type illegal behavior, an operation process type illegal behavior and a health and safety type illegal behavior; if it is judged that the current operation of the courier conforms to the violation behavior, recording the violation operation time, operator information, a corresponding express number and violation image evidence; and S5, generating a sorting violation statistical report and outputting alarm information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of express sorting, in particular to a monitoring method for express sorting stage. BACKGROUND

[0002] In the modern express logistics industry, the sorting link as the key hub connecting the collection and delivery and the dispatch directly affects the overall operation efficiency and customer experience. With the rapid development of e-commerce economy, the quantity of express packages is growing exponentially, and the traditional manual sorting mode is facing severe challenges. On the one hand, high-intensity repetitive labor is easy to cause operator fatigue, leading to increased sorting error rate and increased risk of package damage; on the other hand, the lack of effective supervision means makes it difficult to find and correct irregular operations, such as violent throwing, process disorder, and health and safety hazards, which not only threatens the health of employees, but also causes loss of reputation and economic loss to the enterprise.

[0003] In the prior art, video monitoring systems have been widely used in sorting sites, but mostly stay at the level of simple video storage and rely on manual playback review, which has defects such as poor timeliness, high missed detection rate, and difficult evidence tracing. Although some automated sorting systems can improve efficiency, their ability to identify complex irregular behaviors is limited, especially in subtle aspects such as human motion specifications and operation process compliance, and there is still a lack of precise and efficient intelligent analysis means. In addition, potential risk factors such as package weight abnormalities have not been effectively linked with visual monitoring, resulting in late problem discovery and difficulty in achieving early warning and intervention

[0004] Therefore, we propose a monitoring method for express sorting stage. SUMMARY

[0005] The main purpose of the present application is to provide a monitoring method for express sorting stage, aiming to solve the problems of insufficient intelligent monitoring of sorting link, late identification of irregular behaviors, difficult evidence tracing, and lack of multi-source data linkage in the prior art.

[0006] To achieve the above purpose, the present application provides a monitoring method for express sorting stage, comprising the following steps:

[0007] S1, real-time acquisition of video stream data of the sorting area;

[0008] S2, image analysis of the video stream data to identify express order number information;

[0009] S3, detection of the sorting action of the operator based on a human pose recognition algorithm;

[0010] S4, judging whether the current sorting action is in violation of a preset rule of sorting behavior, the rule of sorting behavior including action specification type violation behavior, operation flow type violation behavior, and health and safety type violation behavior; if it is determined that the current operation of the courier complies with the above violation behaviors, recording the time of the violation operation, operator information, corresponding express number, and violation image evidence;

[0011] S5, generating a sorting violation statistical report and outputting alarm information.

[0012] Preferably, the step S3 of detecting the sorting action of the operator based on the human body posture recognition algorithm specifically comprises the following steps: using an OpenPose algorithm framework to extract key skeleton point coordinates of the operator in real time, the key skeleton point coordinates including movement trajectories of hands, elbows, shoulders, and waists; pre-processing the extracted key skeleton point coordinate data, and constructing a space-time movement trajectory model of the hands, elbows, shoulders, and waists based on the pre-processed key skeleton point coordinates; representing action changes and spatial position relationships of each part of the operator in the sorting process through the constructed space-time movement trajectory model, and calculating instantaneous speed, acceleration, and angle changes of key skeleton points of the operator.

[0013] Preferably, the step S4 of judging the action specification type violation behavior specifically comprises: when the instantaneous speed of the hand skeleton point coordinates of the operator in the sorting process exceeds a threshold value and the movement trajectory is a parabola upward or downward, it is determined that the operator makes a throwing action in the sorting process; when the acceleration mutation value of the hand skeleton point coordinates of the operator in the sorting process exceeds a preset threshold value, it is determined that there is an abnormal action of rapidly shaking the package; when the relative angle change frequency of the elbow and shoulder joint of the operator exceeds a preset threshold value, it is determined that there is a violation situation of excessively large sorting action amplitude or frequent limb shaking; when the included angle between the waist skeleton point of the operator and the ground projection is less than a first preset angle threshold value, it is determined that there is a violation posture of excessive bending; when the movement activity level of a unilateral hand skeleton point and the activity level of the other hand differ by more than a second preset threshold value, it is determined that there is a non-standard behavior of single-hand operation; when the movement trajectory of the hand to the target sorting area deviates from the optimal path and the deviation amount exceeds a third preset threshold value, it is determined that there is an inefficient action of detouring or circuitous operation.

[0014] Preferably, the judgment of operation flow type violation behavior in step S4 is specifically: when the hand skeleton point does not enter the scanning device coordinate range within 5 seconds after the package arrives at the sorting area, and the corresponding express number does not trigger the system scanning record, it is determined that the operation of not performing code scanning verification is violated; when the package motion trajectory does not enter the target sorting area according to the preset conveyor belt path, and the contact time of the hand skeleton point and the package exceeds the sorting operation standard time length, it is determined that the sorting sequence is disordered; when the same express number corresponding package stays in the sorting area for more than a set threshold, and the hand skeleton point does not generate an effective operation trajectory for the package, it is determined that the timeout stay is not processed; when the same express number is repeatedly grabbed in the sorting process more than the threshold of the sorting number recorded by the system, it is determined that the repeated sorting operation is abnormal.

[0015] Preferably, the judgment of operation flow type violation behavior in step S4 is specifically: when the package stacking height detection value of the sorting area ground exceeds 2 / 3 of the safe passage height, and the distance between the stacking edge and the operator foot skeleton point is less than 50 cm, it is determined that the safe passage is blocked. A package stacking height detection model is established through video stream data of the sorting area, and when the vertical projection height of the package three-dimensional space coordinates of the sorting target area exceeds the safety threshold, it is determined that the stacking area is out of limit violation.

[0016] Preferably, the violation image evidence recorded in step S5 includes: N frames of time sequence images before the violation action occurs, local close-up images of the violation action, and same frame images containing express numbers and operator ID cards.

[0017] Preferably, the method further comprises: generating a multi-dimensional statistical report according to the violation record of step S5, including a sorting violence index heat map, an operator violation ranking, and a package damage risk warning list.

[0018] Preferably, the method further comprises: deploying a weight sensor in the sorting pipeline, and triggering a video review process when it is detected that the package weight deviates from the waybill record by more than 10%.

[0019] The beneficial effects of the technical scheme of the present application are:

[0020] When the operator performs the action of carrying a large package, the system quickly extracts the coordinates of the key skeleton points through the OpenPose algorithm framework, and combines the space-time motion trajectory model to accurately determine whether the operator adopts a standard squatting and standing posture, thereby avoiding waist injury caused by bending over to carry and preventing damage to goods caused by the package falling. If the operator is detected to have abnormal shaking in the motion trajectory of the hands and elbows when grabbing a small express delivery, and the instantaneous acceleration calculated by the system exceeds the threshold value, the system will immediately issue a warning to prompt the operator to adjust the posture, effectively reducing the situation of express delivery falling caused by unstable hands, and avoiding the problems of missed judgment and misjudgment existing in traditional manual monitoring or simple sensor detection.

[0021] At the same time, by calculating the instantaneous speed, acceleration and angle change of the key skeleton points, the motion standardization of the operator is further quantified. In the e-commerce warehouse sorting link, when a new employee is learning the action of placing goods on a designated shelf, the system records the skeleton point data of the standard placement action of an old employee to form a standard action model. When the new employee operates, the system compares the action data of the new employee with the standard model in real time, such as the arm stretching angle and waist twisting amplitude, to provide data support for the standardized training of sorting actions, shorten the learning period of new employees, and reduce the training cost of enterprises.

[0022] At the same time, the system can timely issue a warning by monitoring the action abnormity in real time, for example, when the operator is sorting fragile goods, if the arm swinging speed is too fast, the system will immediately prompt to slow down, effectively preventing damage to equipment or goods caused by unreasonable operation, and reducing the operating loss of enterprises; the long-term accumulated action data can also be used to optimize the sorting process, help enterprises realize intelligent and refined management, and significantly improve the overall efficiency and safety of the logistics sorting link. DETAILED DESCRIPTION

[0023] The embodiments described below are exemplary and are intended to be illustrative of the present application, and are not to be construed as limiting the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0024] In addition, if the description of "first", "second" and the like is involved in the present application, it is only for the purpose of description, such as for distinguishing the same or similar elements, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the fact that a person of ordinary skill in the art can realize it, and when the combination of technical solutions appears to be contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, and is not within the scope of protection required by the present application.

[0025] The application provides a monitoring method for an express sorting stage, comprising the following steps:

[0026] S1, real-time acquisition of video stream data of a sorting area;

[0027] S2, image analysis on the video stream data to identify express order number information;

[0028] S3, detection of sorting actions of an operator based on a human body posture recognition algorithm;

[0029] S4, judgment of whether the current sorting action is illegal according to a preset illegal sorting behavior rule; the illegal sorting behavior rule comprises action specification type illegal behavior, operation process type illegal behavior and health and safety type illegal behavior; if it is judged that the current operation of the express operator conforms to the illegal behavior, the time of illegal operation, operator information, corresponding express order number and illegal image evidence are recorded;

[0030] S5, generation of a sorting illegal behavior statistical report and output of alarm information.

[0031] In the embodiment, the video stream data of the sorting area is acquired in real time, and image analysis and action detection are synchronously performed, so that a manager can understand the specific situation of the sorting site in real time. The real-time nature ensures whole-process monitoring of the sorting process, potential problems and risks are found in time, and illegal behavior is prevented from further expansion. For example, when illegal operation occurs, the manager can know and handle it in the first time, so that greater damage to express packages or influence on sorting efficiency is prevented.

[0032] Further, once it is judged that the operator has illegal behavior, the system immediately records the time of illegal operation, operator information, corresponding express order number and illegal image evidence, which provides a strong basis for subsequent illegal behavior analysis and processing, facilitates the manager to trace the source of problems and understand the specific details of illegal behavior, so that targeted improvement measures are taken. Meanwhile, the video stream data of the sorting area is acquired in real time, and image analysis and action detection are synchronously performed, so that the manager can understand the specific situation of the sorting site in real time, whole-process monitoring of the sorting process is ensured, potential problems and risks are found in time, and illegal behavior is prevented from further expansion.

[0033] In one embodiment, step S3, which detects the operator's sorting actions based on a human posture recognition algorithm, specifically includes the following steps: using the OpenPose algorithm framework to extract the coordinates of the operator's key skeletal points in real time, the coordinates of the key skeletal points including the motion trajectories of the hand, elbow, shoulder, and waist; preprocessing the extracted key skeletal point coordinate data, and constructing a spatiotemporal motion trajectory model of the hand, elbow, shoulder, and waist based on the preprocessed key skeletal point coordinates; using the constructed spatiotemporal motion trajectory model to represent the changes in the operator's movements and spatial positional relationships during the sorting process, and calculating the instantaneous velocity, acceleration, and angle changes of the operator's key skeletal points.

[0034] In this embodiment, the above steps enable high-precision real-time analysis of operator sorting actions. For example, in a parcel sorting scenario, when an operator is handling large packages, the system uses the OpenPose algorithm framework to quickly extract the coordinates of key skeletal points. Combined with a spatiotemporal motion trajectory model, this allows for accurate judgment of whether the operator is using a proper squatting and standing posture, preventing back injuries from bending over to handle packages and preventing packages from slipping and damaging goods. If abnormal shaking is detected in the operator's hand and elbow movements when grasping small parcels, and the instantaneous acceleration calculated by the system exceeds a threshold, an immediate warning will be issued, prompting the operator to adjust their posture. This effectively reduces the chance of parcels falling due to hand instability and avoids the problems of missed or false judgments inherent in traditional manual monitoring or simple sensor detection.

[0035] Simultaneously, by calculating the instantaneous velocity, acceleration, and angular changes of key skeletal points, the system further quantifies the standardization of operator movements. In the e-commerce warehouse sorting process, when new employees learn to place goods on designated shelves, the system records the skeletal point data of experienced employees' standard placement movements to form a standard movement model. When new employees operate, the system compares their movement data with the standard model in real time, such as arm extension angle and waist twisting amplitude, providing data support for standardized training of sorting movements, shortening the learning cycle for new employees, and reducing corporate training costs. At the same time, the system can issue timely warnings by monitoring abnormal movements in real time. For example, when an operator is sorting fragile items, if the arm swings too fast, the system immediately prompts for deceleration, effectively preventing equipment damage or goods damage caused by improper operation and reducing operational losses. The accumulated movement data can also be used to optimize the sorting process, helping enterprises achieve intelligent and refined management, and significantly improving the overall efficiency and safety of the logistics sorting process.

[0036] Specifically, the rate of change of displacement of key skeleton points between adjacent frames is obtained through... Calculate; where, Represented as the coordinates of the key point at time t; This is the video frame interval; It is represented as the instantaneous velocity vector at time t;

[0037] The rate of change of velocity at key skeleton points was obtained through Calculate; where, It is represented as the instantaneous velocity vector at time t; It is represented as the instantaneous acceleration vector at time t;

[0038] The joint bending angle is calculated based on the vectors of adjacent key bone points; taking the elbow joint as an example (composed of three points: shoulder, elbow, and hand), the vector dot product formula is used. Calculate the included angle; where, ; ; For vector dot product, The vector magnitude;

[0039] The rate of change of joint flexion angle was obtained through calculate;

[0040] Furthermore, the judgment of violations related to action specifications in step S4 is specifically as follows: when the instantaneous velocity of the operator's hand skeletal coordinates exceeds a threshold and the trajectory is an upward or downward parabola, it is judged that the operator is making a throwing motion during the sorting process; when the acceleration change value of the operator's hand skeletal coordinates exceeds a preset threshold during the sorting process, it is judged that there is an abnormal action of rapidly shaking the package; when the relative angle change frequency of the operator's elbow and shoulder joints exceeds a preset threshold, it is judged that there is a violation of excessive sorting action amplitude or frequent limb shaking; when the angle between the operator's waist skeletal point and the ground projection is less than a first preset angle threshold, it is judged that there is a violation of excessive bending posture; when the difference between the motion activity of one hand skeletal point and the activity of the other hand exceeds a second preset threshold, it is judged that there is a non-standard behavior of single-handed operation; when the movement trajectory of the hand to the target sorting area deviates from the optimal path and the offset exceeds a third preset threshold, it is judged that there is an inefficient action of detouring or roundabout operation.

[0041] In this embodiment, the system uses both the instantaneous velocity threshold of the hand's skeletal points and the characteristics of the parabolic trajectory to accurately identify "violent throwing" behavior during sorting, avoiding misjudgment based on a single parameter (such as brief rapid movement without actual throwing). This reduces package damage and protects goods. Simultaneously, abnormal shaking is detected based on abrupt changes in the acceleration of the skeletal points (such as drastic acceleration changes within a short period). This identifies instances where operators forcefully shake packages for rapid sorting, preventing packaging breakage and item scattering due to excessive shaking and improving operational standardization.

[0042] The motion amplitude is quantified by the frequency of the relative angle change between the elbow and the shoulder joint. A high frequency may indicate that the motion amplitude is too large (e.g., a large arm swing) or that the limb is unconsciously shaking (e.g., tremor caused by fatigue). The operator's fatigue state or operation habit problems are monitored, early warning of high-load work is given, and the risk of muscle strain is reduced. The degree of bending is determined by the angle between the waist skeletal point and the ground projection. A small angle (e.g., close to 90 degrees) indicates that the bending is too much, and the long-term bending causes the lumbar muscle strain, lumbar spine injury, and other occupational health problems.

[0043] On the other hand, by comparing the difference in motion activity (such as displacement distance, speed, and trajectory complexity) between the unilateral hand and the other hand, a large difference indicates unilateral operation, which avoids the safety hazards caused by unilateral carrying of heavy objects, such as center of gravity imbalance and package sliding, and forces the standardization of bilateral cooperative work.

[0044] In one embodiment, the judgment of operation process class violation behavior in step S4 is as follows: when the hand skeletal point does not enter the scanning device coordinate range within 5 seconds after the package reaches the sorting area, and the corresponding express number does not trigger the system scanning record, it is determined that the operation of not performing code scanning verification is violated; when the package motion trajectory does not enter the target sorting area according to the preset conveyor belt path, and the contact time of the hand skeletal point with the package exceeds the standard length of the sorting operation, it is determined that the sorting order is disordered; when the residence time of the package corresponding to the same express number in the sorting area exceeds the set threshold, and the hand skeletal point does not generate an effective operation trajectory for the package, it is determined that the package is not processed due to timeout; when the same express number is repeatedly grabbed in the sorting process more than the threshold number of sorting times recorded by the system, it is determined that the repeated sorting operation is abnormal.

[0045] In this embodiment, the spatial relationship between the hand skeletal point and the scanning device coordinates is modeled in real time, and the system trigger state of the express number is verified twice, which can effectively identify the process violation behaviors such as missing scanning and wrong scanning. By converting the traditional code scanning link relying on manual checking into real-time system monitoring, the completeness of package information input is improved to more than 99.6%, and the risk of subsequent sorting errors and logistics disputes caused by missing information is significantly reduced. At the same time, based on the deviation analysis of the package motion trajectory and the preset path of the conveyor belt, combined with the dynamic threshold judgment of the hand contact time, the package mistakenly sorted into the non-target area can be captured in real time.

[0046] In addition, a coupling analysis model of package retention time and operation trajectory is constructed. When the system detects that the package retention of a sorting position exceeds a threshold and there is no operation trajectory, an audible and light warning is automatically triggered, and a work order is generated and pushed to the terminal of the nearest operator, so that the average processing time of the retained package is compressed from 12 minutes of manual patrol to 3.2 minutes, and the throughput of the sorting line is greatly improved. At the same time, the binding tracking technology of express number-skeleton trajectory is adopted, and a dynamic comparison mechanism of sorting times and system records is established. When it is monitored that the same package is repeatedly grabbed abnormally, the system immediately locks the package and suspends the surrounding sorting tasks, and at the same time, the operation video is called to analyze the root cause. This function makes the occurrence rate of repeated sorting decrease by 78%, avoiding package damage and waste of human resources caused by redundant operation.

[0047] In summary, the technical scheme changes the operation process violation judgment from result tracing to process blocking through the data linkage in the space-time dimension. Compared with the traditional single node monitoring based on RFID or barcode, the scheme uses computer vision and skeleton tracking technology to realize holographic perception of the three-dimensional interactive relationship of "person-object-field", and upgrades the sorting process compliance supervision from discrete sampling to continuous closed-loop control.

[0048] In one of the embodiments, the judgment of operation process class violation behavior in the step S4 is specifically: when the package stacking height detection value of the sorting area ground exceeds 2 / 3 of the safe passage height, and the distance between the stacking edge and the operator's foot skeleton point is less than 50 cm, it is judged that there is a risk of safe passage blockage; a package stacking height detection model is established through the video stream data of the sorting area, and when the vertical projection height of the three-dimensional space coordinates of the sorting target area exceeds the safety threshold, it is judged that there is a stacking area over-limit violation.

[0049] In this embodiment, the double judgment of package stacking height (more than 2 / 3 of the safe passage height) and operator foot distance (<50 cm) can accurately identify the risk of passage blockage. At the same time, based on the analysis of the three-dimensional space coordinates of the package through the video stream, the system can detect in real time whether the vertical projection height is over-limit, avoiding the lag of traditional manual patrol. And through the dynamic distance monitoring of the stacking edge of the package and the foot skeleton point, the audible and light warning can be triggered in advance when the operator approaches the high-risk stacking area, further shortening the response time of human collision risk.

[0050] In one of the embodiments, the violation image evidence recorded in the step S5 includes: a time sequence image of N frames before the violation action occurs, a close-up image of the violation action, and a same-frame image containing the express number and the operator's ID card.

[0051] In this embodiment, by recording the video pictures of the N frames (for example, N=5 or 10) before the violation occurs, the spatio-temporal context of the operation scene is completely restored, the brewing process of the violation action, the personnel movement trajectory and the on-site environment state are displayed, and an objective basis for determining the continuity of the violation behavior is provided. At the same time, the express single barcode (or electronic face sheet number) and the operator's card information are recorded in the same frame, ensuring the strong correlation between the violation behavior and the specific package and the person responsible.

[0052] In one of the embodiments, the method further comprises: generating a multi-dimensional statistical report according to the violation record of step S5, including a sorting violence index heat map, an operator violation ranking and a package damage risk warning list.

[0053] In one of the embodiments, the method further comprises: deploying a weight sensor at the sorting pipeline, and triggering a video review process when the deviation between the package weight and the shipping record is more than 10%.

[0054] It should be noted that in this document, the term "comprising" or "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, device, article or monitoring method of the sorting stage of express delivery including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, device, article or monitoring method of the sorting stage of express delivery. Without more limitations, the element defined by the statement "including a" does not exclude the existence of another same element in the process, device, article or monitoring method of the sorting stage of express delivery including the element.

[0055] The above description is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method of monitoring a stage of sorting of parcels, characterized in that, The method comprises the following steps: S1, acquiring video stream data of a sorting area in real time; S2, performing image analysis on the video stream data to identify express order information; S3, detecting the sorting action of an operator based on a human body posture recognition algorithm; S4, judging whether the current sorting action is in violation of a preset rule for sorting behavior according to the rule for sorting behavior; the rule for sorting behavior comprises action specification type violation behavior, operation process type violation behavior, and health and safety type violation behavior; if it is determined that the current operation of the express courier is in violation of the above-mentioned violation behavior, the time of the violation operation, operator information, corresponding express order number, and violation image evidence are recorded; S5, generating a sorting violation statistical report and outputting alarm information.

2. The method of claim 1, wherein, In the step S3, the sorting action of the operator is detected based on a human body posture recognition algorithm, and the detection specifically comprises the following steps: The key skeleton point coordinates of the operator are extracted in real time by using an OpenPose algorithm framework, and the key skeleton point coordinates comprise the motion trajectories of hands, elbows, shoulders, and waists; The extracted key skeleton point coordinate data is preprocessed, and a space-time motion trajectory model of hands, elbows, shoulders, and waists is constructed based on the preprocessed key skeleton point coordinates; The space-time motion trajectory model is used to represent the action change and spatial position relationship of each part of the operator in the sorting process, and the instantaneous speed, acceleration, and angle change of the key skeleton points of the operator are calculated.

3. A method of monitoring a stage of sorting of parcels according to claim 2, characterized in that, In the step S4, the judgment of the action specification type violation behavior is specifically as follows: When the instantaneous speed of the hand skeleton point coordinates of the operator in the sorting process exceeds a threshold value and the motion trajectory is a parabola upward or downward, it is determined that the operator makes a throwing action in the sorting process; When the acceleration mutation value of the hand skeleton point coordinates of the operator in the sorting process exceeds a preset threshold value, it is determined that there is an abnormal action of rapidly shaking the package; When the relative angle change frequency of the elbow and shoulder joint of the operator exceeds a preset threshold value, it is determined that there is a violation situation of excessively large sorting action amplitude or frequent limb shaking; When the included angle between the waist skeleton point of the operator and the ground projection is less than a first preset angle threshold value, it is determined that there is a violation posture of excessive bending; When the motion activity of a unilateral hand skeleton point and the activity of the other hand skeleton point differ by more than a second preset threshold value, it is determined that there is a non-standard behavior of single-hand operation; When the motion trajectory of the hand to the target sorting area deviates from the optimal path and the deviation exceeds a third preset threshold value, it is determined that there is an inefficient action of detouring or circuitous operation.

4. The method of claim 2, wherein, In the step S4, the judgment of the operation process type violation behavior is specifically as follows: When the hand skeleton point does not enter the scanning device coordinate range within 5 seconds after the package arrives at the sorting area, and the corresponding express order number does not trigger the system scanning record, it is determined that the violation operation of not performing code scanning verification is not executed; When the package motion trajectory does not enter the target sorting area according to the preset conveying belt path, and the contact time of the hand skeleton point with the package exceeds the standard length of the sorting operation, it is determined that the sorting sequence is disordered; When the stay time of the package corresponding to the same express order number in the sorting area exceeds a set threshold value, and the hand skeleton point does not generate an effective operation trajectory for the package, it is determined that the package is not processed due to timeout stay. When the same express delivery number is repeatedly picked up more than the system-recorded sorting frequency threshold in the sorting process, it is determined that the repeated sorting operation is abnormal.

5. The method of claim 2, wherein, The judgment of the operation flow class violation behavior in the step S4 is specifically: When the package stacking height detection value of the sorting area ground exceeds 2 / 3 of the height of the safety channel, and the distance between the stacking edge and the operator's foot skeleton point is less than 50 cm, it is determined that there is a risk of safety channel blockage. A package stacking height detection model is established through the video stream data of the sorting area. When the vertical projection height of the three-dimensional space coordinates of the sorting target area exceeds the safety threshold, it is determined that the stacking area is out of limit.

6. The method of claim 3, wherein, The violation image evidence recorded in the step S5 includes: the time sequence image of N frames before the violation action occurs, the close-up image of the violation action, and the same frame image containing the express delivery number and the operator's ID card.

7. The method of claim 3, wherein, The method further includes: generating a multi-dimensional statistical report according to the violation record of step S5, including a sorting violence index heat map, an operator violation ranking, and a package damage risk warning list.

8. The method of claim 3, wherein, The method further includes: deploying a weight sensor in the sorting pipeline, and triggering a video review process when the detected package weight deviates from the shipping record by more than 10%.