Method for identifying behavior of workers based on GIS and visual analysis

Abstract

This application relates to the field of behavior analysis technology, specifically to a method for identifying worker behavior based on GIS and visual analysis. The method includes: acquiring GIS location data and visual motion data of all workers within a target work area; determining the degree of motion variability of each worker at each time step based on the visual motion data of at least two workers; determining the combination conformity of the target worker at each time step based on the differences in motion variability and position between the target worker and reference workers; and determining the anomaly level of the target worker based on changes in combination conformity, position, and the number of reference workers over multiple time steps. This method can improve the accuracy of abnormal behavior identification.

Description

Technical Field

[0001] This application relates to the field of behavior analysis technology, specifically to a method for recognizing worker behavior based on GIS and visual analysis. Background Technology

[0002] Currently, the integrated use of the positioning capabilities of geographic information systems (GIS) and the visual analysis capabilities of video surveillance to achieve automated identification and risk warning of worker behavior has become an important direction for improving the level of intelligent safety management.

[0003] In related technologies, electronic fences are usually preset through GIS systems to delineate static safe working areas and dangerous areas. At the same time, monitoring videos are analyzed based on computer vision models, and abnormal behavior is identified when personnel enter dangerous areas.

[0004] However, the judgment rules of the above methods are too rigid and may lead to false alarms. For example, when the team is conducting collaborative operations or security patrols at the boundary of the safe zone, their position may briefly exceed the fixed electronic fence range. In this case, the above methods will directly judge it as abnormal due to the position crossing the boundary, resulting in inaccurate behavior analysis results and failing to meet the intelligent safety monitoring needs in complex industrial scenarios. Summary of the Invention

[0005] To address the technical problem that existing rules for judging abnormal behavior are too rigid and may lead to false alarms, this application aims to provide a method for recognizing worker behavior based on GIS and visual analysis. The specific technical solution adopted is as follows: The GIS and visual analysis-based method for identifying worker behavior includes: acquiring GIS location data and visual motion data of all workers within a target work area, wherein the GIS location data includes the location at multiple times, and the visual motion data includes the type and score of each action; determining the degree of change of each worker's actions at each time based on the visual motion data of at least two workers, wherein the at least two workers include the target worker and a reference worker for the target worker; determining the combination conformity degree of the target worker at each time based on the difference in the degree of change of actions and the difference in location between the target worker and the reference worker, wherein the combination conformity degree is used to characterize the degree of similarity between the worker's behavior and that of the reference worker; and determining the anomaly level of the target worker based on the changes in the combination conformity degree, the changes in location, and the changes in the number of reference workers at multiple times.

[0006] Optionally, the above-mentioned determination of the degree of change of each worker's actions at each time moment based on the visual motion data of each of at least two workers among all workers includes: segmenting the visual motion data based on multiple time moments included in the GIS positioning data to obtain motion segments corresponding to each time moment, wherein the motion segment corresponding to a time moment is the motion segment between that time moment and the previous time moment; determining the degree of change of the target worker's actions at the first time moment based on the number of motion types included in the work segment corresponding to the target worker at the first time moment, the motion confidence of each motion type, and the average number of motion types included in the motion segments corresponding to multiple time moments, wherein the first time moment is any time moment among multiple time moments.

[0007] Optionally, determining the variability of the target worker's actions at the first moment based on the number of action types included in the work segment corresponding to the target worker at the first moment, the action confidence of each action type, and the average number of action types included in the action segments corresponding to multiple moments includes: determining the ratio of the number of action types included in the work segment corresponding to the target worker at the first moment to the average number of action types as the action complexity of the target worker at the first moment; determining the minimum action confidence and standard deviation of the action confidence for each action type included in the work segment corresponding to the first moment; determining the variability of the target worker's actions at the first moment based on the difference between the action confidence of each action type included in the work segment corresponding to the first moment and the minimum action confidence for each action type, and the standard deviation of the action confidence for each action type; and determining the variability of the target worker's actions at the first moment based on the action complexity and the action variability.

[0008] Optionally, the method further includes: determining the distance between the target worker and each other worker based on the position of all workers at the first moment, wherein the other workers are all workers except the target worker; and identifying other workers whose distance from the target worker is less than or equal to a preset distance threshold as reference workers for the target worker at the first moment.

[0009] Optionally, the method further includes: sorting the distances between the target worker and other workers in ascending order to obtain a distance sequence; determining the difference between every two adjacent distances in the distance sequence and determining the maximum value of the difference; and using the minimum value of the two distances corresponding to the maximum value of the difference as the preset distance threshold.

[0010] Optionally, the determination of the combination compatibility of the target operator at each time moment based on the difference in the degree of change of movement and the difference in position between the target operator and the reference operator at each time moment includes: determining the combination compatibility between the target operator and each reference operator at the first time moment based on the difference in the degree of change of movement between the target operator and the reference operator at the first time moment, the difference in position between the target operator and the reference operator at the first time moment, and the difference in position change between the target operator and the reference operator at the second time moment and the first time moment, wherein the second time moment is the time moment preceding the first time moment; and determining the average combination compatibility between the target operator and each reference operator at the first time moment as the combination compatibility of the target operator at the first time moment.

[0011] Optionally, determining the combination suitability between the target worker and each reference worker based on the difference in the degree of motion change between the target worker and the reference worker at the first moment, the difference in position between the target worker and the reference worker at the first moment, and the difference in position change between the target worker and the reference worker from the second moment to the first moment, includes: determining a first relative degree of motion change based on the difference in the degree of motion change between the target worker and the reference worker at the first moment, wherein the first relative degree of motion change is the relative difference in the degree of motion change between the target worker and the first reference worker; and determining the combination suitability between the target worker and the first reference worker based on the difference in position. A first distance is determined, which is the distance between the target worker and the first reference worker at a first moment, where the first reference worker is any reference worker for the target worker; based on the difference in positional change between the target worker and the reference worker between a second moment and a first moment, a first trajectory similarity is determined, which is the trajectory similarity between the target worker and the first reference worker between the second moment and the first moment; based on the first relative motion change, the first distance, and the first trajectory similarity, a first combination conformity is determined, which is the combination conformity between the target worker and the first reference worker at a first moment.

[0012] Optionally, the above-mentioned determination of the anomaly level of the target worker based on changes in the combination compliance, position, and number of reference workers at multiple times includes: determining the displacement distance of the target worker based on the position changes of the target worker at the current time and the previous time; determining the behavioral safety level of the target worker based on the difference between the combination compliance of the target worker at the current time and the combination compliance of the previous time, the displacement distance, and the ratio of the number of reference workers at the current time to the number of reference workers at the previous time; and determining the anomaly level of the target worker based on the behavioral safety level and a preset mapping relationship, wherein the preset mapping relationship is a mapping relationship between multiple behavioral safety value ranges and multiple anomaly levels.

[0013] Optionally, the method further includes: if the number of locations included in the GIS positioning data of the first operator is greater than a preset number threshold, then the first operator is identified as the target operator, and the first operator is any operator.

[0014] Optionally, the method further includes: if the number of locations included in the GIS positioning data of the first operator is less than or equal to a preset number threshold, determining the anomaly level of the first operator as the lowest anomaly level.

[0015] This application has the following beneficial effects: By analyzing the GIS positioning data and visual motion data of workers within the target work area, the temporal changes in the workers' own behavior are analyzed to obtain the degree of motion change for each worker. The similarity analysis of the behavior of the target workers and the surrounding groups of people yields the degree of combination conformity. By combining the dynamic change trends at multiple times, the anomaly level is comprehensively assessed. This method can accurately distinguish between positioning drift, normal team work, and genuine individual abnormal behavior. By classifying anomalies, the accuracy of the behavior analysis results is improved, thus providing a basis for intelligent monitoring. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for recognizing worker behavior based on GIS and visual analysis, provided as an embodiment of this application; Figure 2This is a flowchart illustrating another method for worker behavior recognition based on GIS and visual analysis, provided as an embodiment of this application. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a worker behavior recognition method based on GIS and visual analysis proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0020] With the development of visual analytics technology, human behavior within a scene can be analyzed and identified in the background, improving scene perception capabilities in hazardous environments. System deployment can be completed directly using footage captured by existing monitoring equipment, reducing deployment costs and maximizing the use of existing equipment for safety monitoring. Furthermore, by integrating mobile wearable GIS devices, location monitoring can be conducted during personnel work, and methods such as electronic fences can reduce personnel entering hazardous areas, further ensuring operational safety. GIS can also supplement blind spots in visual monitoring, achieving dynamic linkage of safety monitoring.

[0021] The process of combining GIS location data with visual analysis of surveillance images presents challenges. Aligning different data structures can be difficult, hindering the integration of location and visual analysis anomalies to improve the sensitivity of abnormal behavior identification. Traditional fixed-range cross-validation methods, particularly in scenarios involving movement at the edge of dangerous areas, or group movements during teamwork, may lead to individuals moving outside the defined area and being directly flagged as abnormal. This fails to distinguish between risky operations and safety patrols moving within critical zones. For example, guiding vehicle parking from outside the effective area allows for more accurate confirmation of parking positions. Therefore, directly marking users who have crossed the boundary as abnormal effectively disconnects their current behavior from the surrounding clustered population.

[0022] The following description, in conjunction with the accompanying drawings, details a specific scheme for a worker behavior recognition method based on GIS and visual analysis provided in this application.

[0023] Please see Figure 1The diagram illustrates a flowchart of a method for recognizing worker behavior based on GIS and visual analysis, provided in one embodiment of this application.

[0024] like Figure 1 As shown, the operator behavior recognition method based on GIS and visual analysis includes S101-S104.

[0025] S101. Obtain GIS positioning data and visual motion data of all workers within the target work area.

[0026] The GIS positioning data includes locations at multiple times, and the visual motion data includes the motion type and motion confidence level for each motion.

[0027] It should be understood that GIS positioning data refers to location data collected in real time through wearable GIS positioning devices worn by workers, which is used to directly determine the geographical location of personnel and to determine whether personnel have entered dangerous areas or crossed boundaries.

[0028] Optionally, the GIS positioning data may also include personnel identifiers, and the location at a given time may be a coordinate point.

[0029] Understandably, this visual motion data consists of the actions of each worker extracted from the surveillance video, along with the analysis results of those actions (i.e., action type and action confidence).

[0030] In one alternative implementation, surveillance video of the target work area can be collected using a monitoring and shooting device. The video frames from this surveillance video are then input into a pre-trained preset action recognition model (e.g., a YOLO model) to obtain the action type and action confidence of each worker.

[0031] The preset action recognition model is a model trained based on historical actions, capable of action feature extraction and action classification.

[0032] It should be understood that the action confidence score of an action is used to characterize the reliability of the model's classification results for the action type. The value range can be set to [0, 1] or [0, 100]. The higher the score, the higher the reliability of the action classification results.

[0033] It should be noted that the time in the GIS positioning data and the timestamp in the visual motion data should be synchronized to ensure the temporal consistency of the two types of data.

[0034] Optionally, the duration of the acquired GIS positioning data and visual motion data can be determined based on the type of work scenario. For example, for simple work, 1-2 minutes of GIS positioning data and visual motion data can be acquired; for complex work, 3-5 minutes of GIS positioning data and visual motion data can be acquired; and for high-risk and dangerous work, 2-3 minutes of GIS positioning data and visual motion data can be acquired.

[0035] Optionally, the worker behavior recognition method based on GIS and visual analysis provided in this application embodiment is performed in real time, and for example, it can be executed once every 10 seconds.

[0036] S102. Based on the visual motion data of each of at least two of the workers, determine the degree of motion variation of each worker at each time step.

[0037] Among them, the at least two operators include the target operator and the reference operator of the target operator.

[0038] It should be understood that the target worker is the worker currently located in the target work area, and the reference person for the target worker is the person next to the target worker.

[0039] It is understandable that movement at the edge of a danger zone could represent both risky operations and safety patrols, and GIS positioning also has errors. Therefore, it is possible to determine whether it is a real anomaly by combining the overall changes of the team, that is, by combining the visual and motion data of the reference personnel to determine whether it is an isolated case.

[0040] In one alternative implementation, the distance between the target worker and each other worker can be determined based on the position of all workers at the first moment; other workers whose distance from the target worker is less than or equal to a preset distance threshold are identified as reference workers for the target worker at the first moment.

[0041] Among them, the other workers are all workers except the target worker, and the first moment is any moment among multiple moments, such as the current moment.

[0042] Optionally, the preset distance threshold can be set according to the distance between the team and other teams during team operations.

[0043] In one alternative implementation, the distances between the target worker and other workers can be sorted in ascending order to obtain a distance sequence; the difference between every two adjacent distances in the distance sequence can be determined, and the maximum difference can be determined; the minimum of the two distances corresponding to the maximum difference can be used as the preset distance threshold.

[0044] Alternatively, the difference between every two adjacent distances can be obtained based on the first-order difference (i.e., back minus front).

[0045] For example, assuming the distance sequence (in meters) is 5, 6, 8, 12, 18, 19, 20, the maximum difference corresponding to the distance sequence is 6, and the two distances corresponding to the maximum difference are 12 and 18. The preset distance threshold can be 12.

[0046] In one optional implementation, if the number of locations included in the GIS location data of the first worker is greater than a preset threshold, the first worker is identified as the target worker, and the first worker can be any worker.

[0047] It should be understood that a worker may not always be in the target work area, and determining the degree of abnormality of movement requires sufficient data support. Therefore, the number of GIS positioning data that each worker has successfully acquired can be determined. When the number of GIS positioning data exceeds a preset threshold, it can be determined that the degree of change of movement of the worker can be analyzed. At this time, the worker is identified as the target worker.

[0048] For example, the preset quantity threshold can be 5.

[0049] It is understandable that, since the monitoring equipment is also used to monitor the target work area, and the amount of data in GIS location data is easier to analyze, the target workers can be identified based on this GIS location data.

[0050] In another optional implementation, if the number of locations included in the GIS location data of the first worker is less than or equal to a preset threshold, the anomaly level of the first worker is determined to be the lowest anomaly level.

[0051] It is understandable that if the number of locations included in the GIS positioning data of the first worker is less than or equal to the preset threshold, it means that the behavior of the first worker cannot be analyzed. This may be due to reasons such as the first worker just entering the target work area or the GIS positioning device just being started. If the degree of change of action is forcibly analyzed at this time, the analysis results are likely to be distorted due to insufficient data. Therefore, the abnormality level of the first worker can be determined as the lowest abnormality level.

[0052] It should be understood that the lowest level of abnormality represents the mildest degree of abnormality.

[0053] Optionally, a handling strategy can be set for each anomaly level. The handling strategy corresponding to the lowest anomaly level can be: continue data collection without issuing a warning.

[0054] It should be understood that this degree of change in action reflects the degree of abnormality of the target operator compared to their own historical actions.

[0055] S103. Based on the differences in the degree of change of action and the differences in position between the target operator and the reference operator at each time point, determine the combination conformity of the target operator at each time point.

[0056] The combined conformity score is used to characterize the degree of similarity between the behavior of the operator and that of the reference operator.

[0057] It should be understood that this combination conformity characterizes the degree of consistency between the target worker and the group.

[0058] It is understandable that the smaller the difference in movement and position between a target worker and a reference worker, the higher the degree of compatibility of the target worker combination.

[0059] In one implementation of this application, the combination compatibility degree between the target operator and each reference operator is determined based on the difference in the degree of change of the action between the target operator and the reference operator at the first moment, the difference in the position between the target operator and the reference operator at the first moment, and the difference in the change of the position between the target operator and the reference operator from the second moment to the first moment; the average combination compatibility degree between the target operator and each reference operator at the first moment is determined as the combination compatibility degree of the target operator at the first moment.

[0060] The second moment is the moment preceding the first moment.

[0061] In one alternative implementation, a first relative degree of motion change can be determined based on the difference in motion change between the target worker and the reference worker at the first moment; a first distance can be determined based on the positional difference between the target worker and the first reference worker at the first moment; a first trajectory similarity can be determined based on the positional change difference between the target worker and the reference worker from the second moment to the first moment; and then a first combination conformity can be determined based on the first relative degree of motion change, the first distance, and the first trajectory similarity.

[0062] Wherein, the first relative motion change degree is the relative difference between the motion change degree of the target worker and the first reference worker; the first distance is the distance between the target worker and the first reference worker at the first moment; the first reference worker is any reference worker of the target worker; the first trajectory similarity is the trajectory similarity between the target worker and the first reference worker from the second moment to the first moment; and the first combination conformity is the combination conformity between the target worker and the first reference worker at the first moment.

[0063] Optionally, the trajectory (i.e., displacement vector) of a worker can be determined based on the worker's position at two times, and then the trajectory similarity between the two trajectories can be determined based on the cosine similarity between the two displacement vectors.

[0064] Optionally, the first distance can be a Euclidean distance.

[0065] Optionally, the combination fit between an operator and a reference operator satisfies the following formula: in, Indicates the workers Its reference workers The degree of compatibility between the combinations Indicates the workers From time At the time The displacement vector generated between them Indicates reference personnel From time At the time The displacement vector generated between them express and The similarity between the trajectories, i.e., the first trajectory similarity. Indicates time Operators Its reference workers The distance between them, i.e., the first distance, is 0.001 to prevent the denominator from being zero. Indicates the workers At any moment Degree of motion variation Indicates reference personnel At any moment Degree of motion variation This indicates taking the absolute value.

[0066] In this formula, This indicates the relative degree of change in motion, i.e., the operator's... Its reference workers The relative differences in the degree of change of motion, This indicates that the deviation of the calculated ratio from the ideal value "1" (representing perfect equality) is taken as the absolute value of the deviation, yielding a non-negative relative deviation. This indicates the degree of inconsistency in the behavioral change patterns of two workers. Subtracting the absolute deviation from 1 yields a value that decreases as the difference in the degree of change in the two workers' movements increases; the greater the deviation in the degree of change in the two workers' movements, the lower the value. The larger the value, the lower the compatibility of the combination. The collaboration relationship between the similarity of the two workers' movement directions and their spatial distance was comprehensively measured. The more consistent the movement directions and the closer the spatial distance, the greater the degree of combination.

[0067] The above-mentioned determination of the combination conformity degree comprehensively evaluates the similarity between two operators based on three dimensions: behavioral patterns, static spatial relationships, and dynamic movement trends. The combination conformity degree at a single moment is defined as the average of the combination conformity degrees with each reference operator, which makes the combination conformity degree more reflective of the consistency of group behavior and improves the accuracy of group relationship identification.

[0068] S104. Based on the changes in the combination compliance, location, and number of reference workers of the target workers at multiple times, determine the abnormality level of the target workers.

[0069] It should be understood that abnormal behavior of target workers depends not only on the consistency of a single moment's actions, but also directly on behavioral trends, movement range, and group correlation. A single-dimensional assessment may easily classify "long-distance movement during safety patrols" as abnormal, or overlook "risky operations that are far from the group and exhibit inconsistent behavior." Therefore, it is necessary to re-analyze changes in consistency of actions (behavioral similarity trends), changes in location (movement range), and changes in the number of reference personnel (group correlation) to assess the level of abnormality of target workers from multiple dimensions.

[0070] In one optional implementation, the displacement distance of the target worker can be determined based on the positional changes of the target worker at the current moment and the previous moment; the behavioral safety level of the target worker can be determined based on the difference between the combination compliance degree of the target worker at the current moment and the combination compliance degree of the previous moment, the displacement distance, and the ratio of the number of reference workers at the current moment to the number of reference workers at the previous moment; and the abnormality level of the target worker can be determined based on the behavioral safety level and the preset mapping relationship.

[0071] The preset mapping relationship is a mapping relationship between multiple behavioral safety value ranges and multiple anomaly levels.

[0072] It should be understood that the behavioral safety level of the target worker is used to characterize the degree of safety and credibility of the target worker's behavior at the current moment.

[0073] Optionally, before determining the safety of the behavior, the displacement distance can be normalized based on the historical maximum displacement distance, mapping the displacement distance to the interval [0, 1].

[0074] Optionally, the safety level of a worker's behavior satisfies the following formula: in, Indicates the workers behavioral safety, Indicates the workers At the present moment The corresponding number of reference workers, Indicates the workers In the previous moment The corresponding number of reference workers, Indicates the workers At the present moment The degree of compatibility of the combination Indicates the workers In the previous moment The degree of compatibility of the combination Indicates the normalized workers At the present moment Compared to the previous moment The displacement distance between them.

[0075] Based on this formula, it should be understood that This indicates the time from the previous moment to the current moment for the workers. The stability of the local group environment in which the workers are located; this value reflects the stability of the local group environment in which the workers are located. Should we get closer to the crowd or stay away from the crowd? This indicates the distance traveled per unit of time by the workers. The rate of change of the combination conformity, this value reflects the operator's During the movement, whether the coordination with the group is maintained, strengthened, or weakened, when the numerator is positive, regardless of the distance traveled, the workers... The safety of behavior increases when the numerator is negative, and the safety of behavior decreases when the numerator is negative.

[0076] Optionally, the safety level obtained above may be a positive value or a negative value. Therefore, in order to map to the abnormality level more accurately and quickly, the operation validity can be normalized and mapped to the [0,1] interval. Then, the normalized operation validity value is compared with the preset mapping relationship to determine the value interval to which it belongs, and thus the abnormality level corresponding to the target operator is obtained.

[0077] It should be understood that the anomaly level is arranged from low to high, representing a higher degree of anomaly.

[0078] Optionally, the preset mapping relationship can be based on the risk level classification requirements of the work scenario.

[0079] For example, there can be 4 abnormality levels, with each abnormality level corresponding to a value range: [0.75,1] corresponds to level 1, which represents the "normal level", [0.5,0.75) corresponds to level 2, which represents the "warning level", [0.25,0.5) corresponds to level 3, which represents the "proactive intervention level", and [0,0.25) corresponds to level 4, which represents the "emergency response level".

[0080] In one alternative implementation, different handling strategies can be adopted for different exception levels.

[0081] For example, for Level 1, the response strategy could be: continue data collection without issuing a warning; for Level 2, the response strategy could be: send a warning via the audible and visual signals integrated into the positioning device, requiring the target worker to confirm their status within a short period and track their subsequent movements. If there is no timely feedback or the anomaly persists, a higher level of intervention would be automatically triggered; for Level 3, the response strategy could be: immediately notify safety personnel to intervene, temporarily restrict some of the target worker's operational permissions, and link video surveillance for on-site verification, while requiring the target worker to immediately return to a safe area or provide a detailed explanation of the situation; for Level 4, the response strategy could be: completely suspend the target worker's operational permissions, automatically notify the emergency response team for emergency intervention, initiate location tracking and safety confirmation procedures, and prepare for external rescue to ensure the safety risk is controlled as quickly as possible.

[0082] Based on the behavioral safety level determined by the changes in displacement distance, combination compliance degree and the number of reference personnel, it can dynamically reflect the degree of risk of personnel behavior, and then realize graded response based on the preset mapping relationship.

[0083] The methods provided in S101-S104 above use GIS positioning data and visual motion data of workers in the target work area to analyze the temporal changes in the workers' own behavior to obtain the degree of change in each worker's actions. They also analyze the similarity between the behavior of the target workers and the surrounding group of people to obtain the degree of combination conformity. Furthermore, they combine the dynamic change trends at multiple times to comprehensively evaluate the level of abnormality. This can accurately distinguish between positioning drift, normal team work and genuine individual abnormal behavior. By classifying the level of abnormality, the accuracy of the behavior analysis results is improved, thus providing a basis for intelligent monitoring.

[0084] Combination Figure 1 ,like Figure 2As shown, in one implementation of this application embodiment, the above-mentioned S102 can be specifically implemented by S201-S202.

[0085] S201. Based on the multiple time points included in the GIS positioning data, the visual motion data is segmented to obtain the motion segment corresponding to each time point.

[0086] Among them, the action segment corresponding to a moment is the action segment between that moment and the previous moment.

[0087] Optionally, all consecutive moments in the GIS positioning data can be extracted, sorted in chronological order to form a time sequence, and the time interval can be used as the boundary between two adjacent moments to define the action segments of the corresponding moments.

[0088] Synchronously, retrieve the timestamp information of the visual motion data, filter out all motion frames whose timestamps fall within a certain time interval, and obtain the motion segment corresponding to that moment.

[0089] It should be understood that this step enables the alignment of GIS positioning data and visual motion data. It establishes a one-to-one correspondence between "time-moment" segments, providing a temporally matched motion data foundation for subsequent calculations of motion variability over time.

[0090] S202. Based on the number of action types included in the work segment corresponding to the target worker at the first moment, the action confidence of each action, and the average number of action types included in the action segments corresponding to multiple moments, determine the degree of action variability of the target worker at the first moment.

[0091] Here, the first moment can be any one of multiple moments.

[0092] Optionally, the number of action types for each action included in the work segment corresponding to each time moment of the target worker can be counted to obtain the number of action types corresponding to each time moment. Then, the number of action types corresponding to the first time moment can be extracted, and the average number of action types corresponding to the multiple time moments can be determined.

[0093] Optionally, since the first moment is not the current moment, the average number of actions of this type can also be determined based on the action segments corresponding to historical moments before the first moment.

[0094] It is understandable that the more types of actions there are, the higher the complexity of the actions, and the greater the variability of the actions should be; the greater the change in action confidence, the stronger the volatility, and the greater the variability of the actions.

[0095] In one optional implementation, the ratio of the number of action types included in the work segment corresponding to the target worker at the first moment to the average number of action types can be determined as the action complexity of the target worker at the first moment. Then, the minimum action confidence and standard deviation of action confidence for each action type included in the work segment corresponding to the first moment can be determined. Based on the difference between the action confidence of each action included in the work segment corresponding to the target worker at the first moment and the minimum action confidence of its corresponding action type, as well as the standard deviation of action confidence for each action type, the action variability of the target worker at the first moment can be determined. Finally, based on the action complexity and the action variability, the action change of the target worker at the first moment can be determined.

[0096] Optionally, the degree of change in a worker's action at a given moment satisfies the following formula: in, Indicates the workers At any moment Degree of motion variation Indicates the workers At any moment The number of action types included in the corresponding action segment Indicates the workers At any moment One or more previous historical moments In the mean number of action types, Indicates the workers At any moment Identify the total number of actions. Indicates the workers The Action confidence of each action Indicates the workers At any moment In the corresponding action segment, the first The action type to which this action belongs The minimum confidence level of the action. Represents one or more historical moments In the middle, action type The standard deviation of the confidence level of the action.

[0097] Based on this formula, it should be understood that Indicates action complexity, used to measure time. The degree of behavioral change relative to the worker's own historical habits; This indicates the volatility of action, where, Indicates the first The absolute offset of the action confidence of each action. This is used to "standardize" the aforementioned absolute offset, eliminating the influence of inherent fluctuation differences between different types and making instances of different types comparable. In this formula, the cube operation is used to... "Perform a cube to achieve a nonlinear transformation, when..." When the value is very large (far exceeding 1), the cube value becomes extremely large, strongly indicating that this instance is an outlier. When the value of is very small (close to 0), the cube value is still very small, which makes the formula extremely sensitive to significant fluctuations, but insensitive to small noise.

[0098] The degree of behavioral variability derived from this formula reflects the overall extent to which a target worker's behavioral patterns deviate from their personal historical baseline over a given period. A greater degree of variability indicates highly abnormal behavior, potentially manifesting as: suddenly performing multiple uncommon actions (significantly increased complexity), or exhibiting highly unstable execution quality of routine actions (drastic fluctuations in confidence level), or both. In a safety monitoring context, this is often associated with states such as panic, risk-taking, violations, distraction, or equipment malfunction. Conversely, a greater degree of variability suggests stable and normal behavior, consistent with the individual's historical patterns. Even if the person is on the edge of a hazardous area, as long as their behavioral combinations and the quality of their actions are stable, they are more likely engaged in orderly patrols, standardized collaboration, or other safe work practices.

[0099] The methods provided in S201-S202 above accurately segment continuous visual motion data based on the time included in the GIS positioning data, ensuring that a motion segment strictly corresponds to a specific positioning time and time period. This time alignment method establishes a unified and reliable time benchmark for all subsequent analyses, enabling the calculation of motion variability to accurately reflect behavioral characteristics within a specific time period. Then, by separately quantifying motion complexity to reflect behavioral diversity and motion volatility to reflect motion execution stability, the motion variability is determined by combining the two. This allows the motion variability to simultaneously and keenly capture anomalies in both "what is being done" and "how well it is being done" in human behavior, thus improving the reliability of behavioral analysis.

[0100] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for recognizing worker behavior based on GIS and visual analysis, characterized in that, include: Acquire GIS positioning data and visual motion data of all workers within the target work area. The GIS positioning data includes the location at multiple times, and the visual motion data includes the type and score of each action. Based on the visual motion data of each of at least two workers among all workers, determine the degree of motion change of each worker at each time moment, wherein the at least two workers include the target worker and the reference worker of the target worker; Based on the differences in the degree of change of action and the difference in position between the target worker and the reference worker at each time point, the combination conformity of the target worker at each time point is determined. The combination conformity is used to characterize the degree of similarity between the behavior of the worker and the reference worker. The abnormality level of the target workers is determined based on the changes in the combination compliance, location, and number of reference workers at multiple times.

2. The method for recognizing worker behavior based on GIS and visual analysis according to claim 1, characterized in that, The determination of the degree of motion change of each worker at each time step, based on visual motion data of at least two workers among all workers, includes: Based on the multiple time points included in the GIS positioning data, the visual motion data is segmented to obtain motion segments corresponding to each time point. The motion segment corresponding to a time point is the motion segment between the time point and the previous time point. Based on the number of action types included in the work segment corresponding to the target worker at the first moment, the action confidence of each action type, and the average number of action types included in the action segments corresponding to multiple moments, the degree of action variability of the target worker at the first moment is determined. The first moment can be any moment among multiple moments.

3. The method for recognizing worker behavior based on GIS and visual analysis according to claim 2, characterized in that, The determination of the target worker's variability in the first moment, based on the number of action types included in the work segment corresponding to the target worker at the first moment, the action confidence of each action type, and the average number of action types included in the action segments corresponding to multiple moments, includes: The ratio of the number of action types in the work segment corresponding to the target worker at the first moment to the average number of action types is determined as the action complexity of the target worker at the first moment. Determine the minimum confidence value and standard deviation of the action confidence for each action type included in the task segment corresponding to the first time step; Based on the difference between the action confidence of each action type and the minimum action confidence of each action type in the work segment corresponding to the target worker at the first moment, and the standard deviation of the action confidence of each action type, the action volatility of the target worker at the first moment is determined. Based on the complexity and volatility of the action, the degree of change in the target worker's action at the first moment is determined.

4. The method for recognizing worker behavior based on GIS and visual analysis according to claim 1, characterized in that, The method further includes: Based on the positions of all workers at the first moment, determine the distance between the target worker and each other worker, where the other workers are all workers except the target worker; Other workers whose distance from the target worker is less than or equal to a preset distance threshold are identified as reference workers for the target worker at the first moment.

5. The method for recognizing worker behavior based on GIS and visual analysis according to claim 4, characterized in that, The method further includes: Sort the distances between the target worker and other workers in ascending order to obtain a distance sequence; Determine the difference between every two adjacent distances in the distance sequence, and determine the maximum value of the difference; The minimum of the two distances corresponding to the maximum difference is taken as the preset distance threshold.

6. The method for worker behavior recognition based on GIS and visual analysis according to claim 1, characterized in that, The determination of the combination suitability of the target workers at each time moment, based on the differences in the degree of change of movement and the differences in position between the target workers and the reference workers, includes: Based on the difference in the degree of motion change between the target operator and the reference operator at the first moment, the difference in position between the target operator and the reference operator at the first moment, and the difference in position change between the target operator and the reference operator from the second moment to the first moment, the combination conformity between the target operator and each reference operator at the first moment is determined, and the second moment is the moment before the first moment. The average combination conformity between the target operator and each reference operator at the first moment is determined as the combination conformity of the target operator at the first moment.

7. The method for worker behavior recognition based on GIS and visual analysis according to claim 6, characterized in that, The determination of the combination compatibility between the target worker and each reference worker at the first moment, based on the differences in the degree of motion change between the target worker and the reference worker at the first moment, the differences in the position between the target worker and the reference worker at the first moment, and the differences in the position change between the target worker and the reference worker from the second moment to the first moment, includes: Based on the difference in the degree of change of the target operator and the reference operator at the first moment, the first relative degree of change of the action is determined, which is the relative difference in the degree of change of the action of the target operator and the first reference operator. Based on the positional difference between the target worker and the first reference worker at the first moment, a first distance is determined. The first distance is the distance between the target worker and the first reference worker at the first moment. The first reference worker is any reference worker of the target worker. Based on the difference in positional changes between the target worker and the reference worker between the second time and the first time, the first trajectory similarity is determined. The first trajectory similarity is the trajectory similarity between the target worker and the first reference worker between the second time and the first time. Based on the first relative motion change degree, the first distance, and the first trajectory similarity, a first combination conformity degree is determined. The first combination conformity degree is the combination conformity degree between the target operator and the first reference operator at the first moment.

8. The method for recognizing worker behavior based on GIS and visual analysis according to claim 1, characterized in that, The determination of the anomaly level of the target worker based on changes in the combination compliance, location, and number of reference workers at multiple times includes: Based on the positional changes of the target worker at the current moment and the previous moment, determine the displacement distance of the target worker; The safety level of the target worker's behavior is determined based on the difference between the combination compliance rate of the target worker at the current moment and the combination compliance rate at the previous moment, the displacement distance, and the ratio of the number of reference workers at the current moment to the number of reference workers at the previous moment. Based on the behavioral safety level and the preset mapping relationship, the abnormality level of the target operator is determined. The preset mapping relationship is a mapping relationship between multiple behavioral safety level value ranges and multiple abnormality levels.

9. The method for recognizing worker behavior based on GIS and visual analysis according to claim 1, characterized in that, The method further includes: If the number of locations included in the GIS location data of the first operator is greater than a preset threshold, the first operator is identified as the target operator, and the first operator can be any operator.

10. The method for worker behavior recognition based on GIS and visual analysis according to claim 1, characterized in that, The method further includes: If the number of locations included in the GIS location data of the first operator is less than or equal to a preset threshold, the anomaly level of the first operator will be determined as the lowest anomaly level.

Images