Behavior anomaly analysis method and system based on target tracking

Abstract

The invention provides a behavior anomaly analysis method and system based on target tracking, and relates to the technical field of artificial intelligence, and the method comprises the steps: monitoring the motion state of a key target in real time, and obtaining a first monitoring video; extracting a target movement track generated by the target position information from a video frame in the first monitoring video, and obtaining a first behavior feature of the key target; and performing anomaly analysis and alarm on the first behavior characteristic based on the business logic of the fuel operation scene. Obtaining a first monitoring video by monitoring the motion state of the key target in different fuel operation scenes in real time; based on a target motion track generated by extracting target position information from a video frame in the first monitoring video by adopting a set target tracking algorithm, obtaining a first behavior feature of the key target; and performing anomaly analysis and alarm on the first behavior feature based on the corresponding business logic of the current fuel operation scene, thereby realizing high-precision tracking of the target object in the monitoring scene and improving flexibility of behavior anomaly alarm.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly relates to a method and system for analyzing abnormal behaviors based on target tracking. Background Art

[0002] Most traditional monitoring systems rely on manual monitoring, which is not only inefficient but also prone to missed alarms and false alarms. In recent years, the target tracking technology based on computer vision has developed rapidly, providing the possibility for automated and intelligent behavior analysis. However, the existing target tracking algorithms still face challenges in dealing with complex scenarios, and the judgment of abnormal behaviors often relies on simple rule settings, lacking flexibility and accuracy. Therefore, how to achieve target tracking and accurate abnormal behavior analysis in complex scenarios has become one of the current research focuses.

[0003] Therefore, the present invention provides a method and system for analyzing abnormal behaviors based on target tracking. Summary of the Invention

[0004] The present invention provides a method and system for analyzing abnormal behaviors based on target tracking, which is used to obtain a first monitoring video by real-time monitoring the motion states of key targets in different fuel operation scenarios; obtain the first behavior characteristics of the key targets based on the target motion trajectories generated by extracting the target position information from the video frames in the first monitoring video by using a set target tracking algorithm; perform abnormal analysis and alarm on the first behavior characteristics based on the corresponding business logic of the current fuel operation scenario, and can achieve high-precision tracking of target objects in the monitoring scenario and improve the flexibility of abnormal behavior alarms.

[0005] The present invention provides a method for analyzing abnormal behaviors based on target tracking, including: Step 1: Use a set shooting tool to real-time monitor the motion states of key targets in different fuel operation scenarios to obtain a first monitoring video; Step 2: Use a set target tracking algorithm to extract the position information of the key targets from the video frames in the first monitoring video to generate a target motion trajectory; Step 3: Based on the target motion trajectory, obtain the first behavior characteristics of the current key targets; Step 4: Based on the corresponding business logic of the current fuel operation scenario, perform abnormal analysis on the first behavior characteristics and give an alarm when there is an abnormal behavior.

[0006] Preferably, using a set shooting tool to real-time monitor the motion states of key targets in different fuel operation scenarios to obtain a first monitoring video includes: Set target monitoring points within the corresponding fuel operation areas of different fuel operation scenarios; Install a set shooting tool at each target monitoring point, and the set shooting tool monitors the running state of the key target in real time to obtain the first monitoring video; The set shooting tool uploads the obtained first monitoring video to the fuel supervision platform in real time.

[0007] Preferably, use a set target tracking algorithm to extract the position information of the key target from the video frames in the first monitoring video and generate a target motion trajectory, including: Preprocess the video frames in the first monitoring video to obtain a key frame sequence; Input the first key frame in the key frame sequence into a target detection model established in advance based on deep learning to obtain a target detection result; According to the target detection result and the key frame sequence, use a set target tracking algorithm to continuously track the current key target and obtain the first position information of the current key target in each key frame; Based on all the obtained first position information, use a set interpolation algorithm to generate the target motion trajectory of the current key target.

[0008] Preferably, based on the target motion trajectory, obtain the first behavior characteristics of the current key target, including: Perform data standardization processing on the trajectory data of the target motion trajectory to obtain the first trajectory data; According to the timestamps and position coordinates in the first trajectory data, determine the average speed of the current key target in each preset analysis time period; According to the average speed of the key target in each preset analysis time period, determine the acceleration of the current key target; Calculate the first difference in longitude and latitude of each pair of adjacent points on the target motion trajectory, and convert the first difference to obtain the first direction angle; Construct a direction angle change curve for all the obtained first direction angles in the order of acquisition; Extract trend characteristics from the direction angle change curve and calculate the weighted average to obtain the direction angle change coefficient; Statistically count the first position coordinates where the key target stays for more than the set time threshold, and the corresponding stay time is regarded as the abnormal stay time; Output the average speed, acceleration, first direction angle, direction angle change coefficient, abnormal stay time, and first position coordinates as the first behavior characteristics of the current key target.

[0009] Preferably, based on the corresponding business logic of the current fuel operation scenario, perform abnormal analysis on the first behavior characteristics and give an alarm when there is an abnormal behavior, including: Calculate the associated behavior features based on the characteristics of the first row, and perform normalization processing on the associated behavior features to obtain the key-associated behavior feature values; Obtain the corresponding scenario business logic according to the current fuel operation scenario, and then extract the list of associated behavior thresholds from the scenario business logic; Use the associated behavior features as matching conditions to extract the corresponding associated behavior threshold ranges from the list of associated behavior thresholds; Optimize the associated behavior threshold ranges based on the analysis of historical data to obtain the target behavior threshold ranges; Compare the key-associated behavior feature values with the corresponding target behavior threshold ranges, and mark the associated behavior features whose key-associated behavior feature values belong to the corresponding target behavior threshold ranges as normal associated behavior features; Mark the associated behavior features whose key-associated behavior feature values do not belong to the corresponding target behavior threshold ranges as abnormal associated behavior features; If there are abnormal associated behavior features, determine that there are abnormal behaviors in the current key target; Determine the abnormal behavior types of the current key target according to the abnormal associated behavior features; Generate an alarm signal based on the abnormal behavior type and the target basic information of the key target, and transmit it to the fuel supervision system for corresponding response.

[0010] Preferably, the associated behavior features include speed standard deviation, acceleration mean, acceleration standard deviation, direction angle change rate, and the proportion of abnormal stay time.

[0011] Preferably, based on the analysis of historical data, optimize the associated behavior threshold ranges to obtain the target behavior threshold ranges, including: Obtain the historical optimization data of the associated behavior threshold ranges for each associated behavior feature within the historical time period; According to the historical optimization data, construct a threshold upper limit change curve for each associated behavior feature with the historical optimization value of the associated behavior threshold upper limit as the dependent variable in chronological order; Extract the trend features from the obtained threshold upper limit change curve and perform weighted average to obtain the threshold upper limit change coefficient; Use the threshold upper limit change coefficient to optimize the associated behavior threshold upper limit of the current associated behavior feature to obtain the optimized threshold upper limit; Among them, the calculation formula for the optimized threshold upper limit is as follows: ; In the formula, represents the optimized threshold upper limit of the current associated behavior feature; represents the corresponding optimized threshold upper limit at the previous optimization moment of the current associated behavior feature; The upper limit change coefficient of the threshold represented as the current associated behavior feature; Represents the historical time period; Represents the time interval from the current to the previous optimization moment; e represents a constant with a value of 2.7; Taking the historical optimization value of the lower limit of the associated behavior threshold as the dependent variable, construct the lower limit change curve of the threshold for each associated behavior feature in chronological order; Extract the trend feature from the obtained lower limit change curve of the threshold and perform weighted average to obtain the lower limit change coefficient of the threshold; Using the lower limit change coefficient of the threshold, optimize the lower limit of the associated behavior threshold of the current associated behavior feature to obtain the optimized lower limit of the threshold; Among them, the calculation formula of the optimized lower limit of the threshold is as follows: ; In the formula, Represents the optimized lower limit of the threshold of the current associated behavior feature; Represents the corresponding optimized lower limit of the threshold at the previous optimization moment of the current associated behavior feature; Represents the lower limit change coefficient of the threshold of the current associated behavior feature; Represents the historical time period; Represents the time interval from the current to the previous optimization moment; e represents a constant with a value of 2.7; Using the optimized upper limit and optimized lower limit of the threshold, establish the initial optimized threshold range of the current key behavior feature; Calculate the deviation in the size of the associated behavior threshold between adjacent historical optimization moments of the current associated behavior feature; Combine the deviation in the size of the associated behavior threshold between adjacent historical optimization moments with the corresponding time interval between adjacent historical optimization moments to calculate the threshold-duration ratio; Average all the obtained threshold-duration ratios to obtain the target threshold-duration ratio; Combine the time interval between the previous optimization moment and the current moment with the target threshold-duration ratio to calculate the estimated size of the associated behavior threshold; If the size of the threshold of the initially obtained optimized threshold range is not greater than the estimated size of the associated behavior threshold, then output the initially obtained optimized threshold range as the target behavior threshold range; If the size of the threshold of the initially obtained optimized threshold range is greater than the estimated size of the associated behavior threshold, then gradually and overall reduce the initially obtained optimized threshold range according to a set ratio until the size of the threshold is not greater than the estimated size of the associated behavior threshold, and then output it as the target behavior threshold range.

[0012] The present invention provides a behavior anomaly analysis system based on target tracking, including: Monitoring module: It is used to monitor the motion state of key targets in different fuel operation scenarios in real time by using a set shooting tool, and obtain a first monitoring video; Trajectory analysis module: It is used to extract target position information from video frames in the first monitoring video by using a set target tracking algorithm, and generate a target motion trajectory; Feature analysis module: It is used to obtain the first behavior feature of the current key target based on the target motion trajectory; Abnormality analysis module: It is used to perform abnormality analysis on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, and give an alarm when there is an abnormal behavior.

[0013] Compared with the prior art, the beneficial effects of the present application are as follows: By monitoring the motion state of key targets in different fuel operation scenarios in real time to obtain a first monitoring video; based on the target motion trajectory generated by extracting target position information from video frames in the first monitoring video by using a set target tracking algorithm, obtaining the first behavior feature of the key target; performing abnormality analysis and alarm on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, it is possible to achieve high-precision tracking of target objects in the monitoring scenario and improve the flexibility of behavior abnormality alarm.

[0014] Other features and advantages of the present invention will be described in the following description, and part of them will be obvious from the description, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written description and the drawings.

[0015] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0016] The drawings are used to provide a further understanding of the present invention, and constitute a part of the description. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 It is a flowchart of a method for analyzing behavior abnormality based on target tracking in an embodiment of the present invention; Figure 2 It is a structural diagram of a system for analyzing behavior abnormality based on target tracking in an embodiment of the present invention. Detailed Embodiments

[0017] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0018] The embodiment of the present invention provides a method for analyzing behavior abnormality based on target tracking, asFigure 1 As shown in Step 1: Use a set shooting tool to monitor the motion state of key targets in different fuel operation scenarios in real time, and obtain a first monitoring video; Step 2: Adopt a set target tracking algorithm to extract the position information of key targets from the video frames in the first monitoring video, and generate a target motion trajectory; Step 3: Based on the target motion trajectory, obtain the first behavior feature of the current key target; Step 4: Based on the corresponding business logic of the current fuel operation scenario, perform anomaly analysis on the first behavior feature, and give an alarm when there is an abnormal behavior.

[0019] In this embodiment, the set shooting tool refers to a high-definition camera; the fuel operation scenario refers to a specific environment involving fuel processing, storage, transportation and other related activities; the key targets refer to staff, vehicles and movable mechanical equipment; the first monitoring video refers to a monitoring video captured and generated by a high-definition camera in real time within the fuel operation area; the set target tracking algorithm refers to a pre-set algorithm for continuously tracking targets in a sequence of key frames, such as optical flow method, Kalman filter, etc.; the first behavior features include speed, acceleration, direction change, residence time, etc.; the business logic refers to a series of rules and standards formed according to the operation process, safety specifications, equipment operation requirements, etc. during the fuel operation process, such as an associated behavior threshold list.

[0020] The beneficial effects of the above technical solution are: obtaining a first monitoring video by monitoring the motion state of key targets in different fuel operation scenarios in real time; obtaining the first behavior feature of the key target based on the target motion trajectory generated by extracting the target position information from the video frames in the first monitoring video by adopting a set target tracking algorithm; performing anomaly analysis and alarm on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, which can realize high-precision tracking of target objects in the monitoring scenario and improve the flexibility of behavior anomaly alarm.

[0021] An embodiment of the present invention provides a method for analyzing behavior anomalies based on target tracking, which uses a set shooting tool to monitor the motion state of key targets in different fuel operation scenarios in real time, and obtains a first monitoring video, including: Set target monitoring points in the corresponding fuel operation areas of different fuel operation scenarios; Install a set shooting tool at each target monitoring point, and the set shooting tool monitors the running state of key targets in real time to obtain a first monitoring video; The set shooting tool uploads the obtained first monitoring video to the fuel supervision platform in real time.

[0022] In this embodiment, the fuel operation scenario refers to the specific environment involving fuel handling, storage, transportation, and other related activities; the fuel operation area refers to the specific geographical scope specifically demarcated in the fuel operation scenario for conducting fuel-related operations; the target monitoring point refers to the position where a preset shooting tool is installed and set to monitor, which can cover key operation points (such as fuel storage areas, transportation pipelines, and operation consoles); the set shooting tool refers to a high-definition camera; the first monitoring video refers to the monitoring video captured and generated in real time by the high-definition camera within the fuel operation area; the fuel supervision platform refers to a system platform that centrally manages the monitoring data of the fuel operation scenario, with functions such as video monitoring, data analysis, and abnormal alarm, and can receive and process the monitoring video data from the high-definition camera in real time.

[0023] The beneficial effects of the above technical solution are: By using the set shooting tool to monitor the motion states of key targets in different fuel operation scenarios in real time, the first monitoring video is obtained, which can provide a data basis for subsequent target tracking and abnormal behavior analysis.

[0024] An embodiment of the present invention provides a method for analyzing abnormal behavior based on target tracking, which uses a set target tracking algorithm to extract the position information of key targets from the video frames in the first monitoring video and generate a target motion trajectory, including: Preprocess the video frames in the first monitoring video to obtain a sequence of key frames; Input the first key frame in the sequence of key frames into a target detection model established in advance based on deep learning to obtain a target detection result; According to the target detection result and the sequence of key frames, use the set target tracking algorithm to continuously track the current key target, and obtain the first position information of the current key target in each key frame; Based on all the obtained first position information, use a set interpolation algorithm to generate the target motion trajectory of the current key target.

[0025] In this embodiment, preprocessing refers to denoising and grayscale processing; the sequence of key frames refers to the frame sequence generated after preprocessing the video frames in the first monitoring video; the target detection model is used to identify targets (such as people, vehicles, etc.) and their positions in the video frames. It is obtained by collecting a large number of image frame data, annotating the target categories and position information in the images, then performing data preprocessing, and finally using the preprocessed data as training data to train a neural network; the target detection result refers to the position and category (such as people, vehicles, etc.) of the key target output after inputting the key frame into the target detection model; the set target tracking algorithm is used to continuously track the target in the sequence of key frames, such as the optical flow method, Kalman filter, etc.; the first position information includes a timestamp and position coordinates; the set interpolation algorithm is predetermined, generally referring to the linear interpolation algorithm.

[0026] The beneficial effects of the above technical solution are as follows: By adopting a set target tracking algorithm to extract target position information from video frames in the first monitoring video and generate a target motion trajectory, it can provide data support for subsequent abnormal behavior analysis.

[0027] An embodiment of the present invention provides a method for analyzing abnormal behavior based on target tracking. Based on the target motion trajectory, the first behavior characteristics of the current key target are obtained, including: Perform data standardization processing on the trajectory data of the target motion trajectory to obtain first trajectory data; According to the timestamps and position coordinates in the first trajectory data, determine the average speed of the current key target in each preset analysis time period; According to the average speed of the key target in each preset analysis time period, determine the acceleration of the current key target; Calculate the first difference in longitude and latitude between each pair of adjacent points on the target motion trajectory, and transform the first difference to obtain a first direction angle; Construct a direction angle change curve for all the obtained first direction angles in the order of acquisition; Extract trend features from the direction angle change curve and calculate the weighted average to obtain a direction angle change coefficient; Statistically count the first position coordinates where the key target stays for more than the set time threshold, and the corresponding stay time is regarded as an abnormal stay time; Output the average speed, acceleration, first direction angle, direction angle change coefficient, abnormal stay time, and first position coordinates as the first behavior characteristics of the current key target.

[0028] In this embodiment, the trajectory data includes information such as timestamps, longitudes, and latitudes; data standardization processing refers to converting the position coordinates of the target motion trajectory into a unified reference system; the first trajectory data refers to the data obtained after performing data standardization processing on the trajectory data; the preset analysis time period is pre-determined, such as 30s; acceleration is used to reflect the rapid change degree of the target motion state and is determined by using the average speed of the key target in the preset analysis time period. For example, there is a key target 1 with an average speed of in the preset analysis time period , and an average speed of in the preset analysis time period , then at this time, the acceleration .

[0029] In this embodiment, the first difference refers to the difference in longitude between each pair of adjacent points on the target motion trajectory or the difference in latitude between each pair of adjacent points; the first direction angle is used to represent the direction change of the target motion, and is calculated by using the arctangent function after converting the difference in longitude and the difference in latitude between each pair of adjacent points on the target motion trajectory into radians respectively; the direction angle change curve is constructed with the obtained first direction angle as the dependent variable according to the acquisition order; the direction angle change coefficient is used to reflect the stability and regularity of the target motion direction change, and is obtained by extracting the trend features from the direction angle change curve and calculating the weighted average, and the formula is expressed as , where is expressed as the direction angle change coefficient; is expressed as the j-th trend feature extracted from the direction angle change curve, j = 1, 2, 3, 4; is expressed as the influence weight of the j-th trend feature extracted from the direction angle change curve on the calculation of the direction angle change coefficient. Among them, the trend features include curve slope, the sum of the deviations of extreme points from the average value, curvature, and acceleration; the weights assigned to the trend features are obtained by solving the matrix constructed by pairwise comparison and relative importance scoring of the trend features using the analytic hierarchy process; the set time threshold is determined in advance; the abnormal stay time refers to the stay time that exceeds the set time threshold; the first position coordinate refers to the corresponding position coordinate when the key target stays for a time exceeding the set time threshold.

[0030] The beneficial effects of the above technical solution are: By obtaining the first behavior feature of the current key target based on the target motion trajectory, it helps to understand the motion pattern of the target and provides strong support for subsequent decision-making and analysis.

[0031] An embodiment of the present invention provides a method for analyzing abnormal behavior based on target tracking. Based on the corresponding business logic of the current fuel operation scenario, abnormal analysis is performed on the first behavior feature, and an alarm is given when there is abnormal behavior, including: According to the first behavior feature, calculate the associated behavior feature, and perform normalization processing on the associated behavior feature to obtain the key-associated behavior feature value; According to the current fuel operation scenario, obtain the corresponding scenario business logic, and then extract the list of associated behavior thresholds from the scenario business logic; Using the associated behavior feature as a matching condition, extract the corresponding associated behavior threshold range from the list of associated behavior thresholds; Based on the analysis of historical data, optimize the associated behavior threshold range to obtain the target behavior threshold range; Compare the key-associated behavior feature values with the corresponding target behavior threshold ranges, and mark the associated behavior features whose key-associated behavior feature values belong to the corresponding target behavior threshold ranges as normal associated behavior features; Mark the associated behavior features whose key-associated behavior feature values do not belong to the corresponding target behavior threshold ranges as abnormal associated behavior features; If there are abnormal associated behavior features, it is determined that there are abnormal behaviors in the current key target; Determine the type of abnormal behavior of the current key target according to the abnormal associated behavior features; Generate an alarm signal based on the abnormal behavior type and the target basic information of the key target, and transmit it to the fuel supervision system for corresponding response.

[0032] In this embodiment, the associated behavior features include speed standard deviation, acceleration mean, acceleration standard deviation, direction angle change rate, and abnormal stay time ratio; the key-associated behavior feature values are obtained by normalizing the associated behavior features, aiming to eliminate the dimensional differences between different feature values; the scenario business logic refers to a series of rules and standards formed according to the operation process, safety specifications, equipment operation requirements, etc. during the fuel operation process, such as the associated behavior threshold list; the associated behavior threshold list is a table composed of the initial normal value ranges set for various associated behavior features (such as speed standard deviation, acceleration mean, etc.) and the corresponding associated behavior features.

[0033] In this embodiment, the associated behavior threshold range is the initial normal value range extracted from the associated behavior threshold list with the associated behavior features as the matching condition; the target behavior threshold range refers to the optimized value range obtained by combining historical data analysis on the basis of the associated behavior threshold list; the normal associated behavior feature refers to the associated behavior feature whose key-associated behavior feature value belongs to the corresponding target behavior threshold range; the abnormal associated behavior feature refers to the associated behavior feature whose key-associated behavior feature value does not belong to the corresponding target behavior threshold range; the abnormal behavior type is determined according to the specific performance of the abnormal associated behavior features, combined with the scenario business logic and operation specifications. For example, if there is an abnormal associated behavior feature 1 that the movement direction of the key target is opposite to the preset normal movement direction, the corresponding abnormal behavior type is that the key target is driving in reverse; the target basic information includes the number of the key target, the fuel business type to which it belongs, and the fuel operation area.

[0034] The beneficial effects of the above technical solution are: By performing abnormal analysis on the first behavior feature based on the corresponding business logic of the current fuel operation scenario and giving an alarm when there are abnormal behaviors, potential safety hazards can be detected in a timely manner, accidents can be avoided, and operation safety can be improved.

[0035] An embodiment of the present invention provides a method for analyzing abnormal behaviors based on target tracking. Based on the analysis of historical data, the associated behavior threshold range is optimized to obtain the target behavior threshold range, including: Obtain the historical optimization data of the associated behavior threshold range for each associated behavior feature within a historical time period; According to the historical optimization data, taking the historical optimization value of the upper limit of the associated behavior threshold as the dependent variable, construct the threshold upper limit change curve for each associated behavior feature in chronological order; Extract the trend features from the obtained threshold upper limit change curve and perform weighted averaging to obtain the threshold upper limit change coefficient; Use the threshold upper limit change coefficient to optimize the upper limit of the associated behavior threshold of the current associated behavior feature to obtain the optimized threshold upper limit; Among them, the calculation formula for the optimized threshold upper limit is as follows: ; In the formula, represents the optimized threshold upper limit of the current associated behavior feature; represents the corresponding optimized threshold upper limit at the previous optimization moment of the current associated behavior feature; represents the threshold upper limit change coefficient of the current associated behavior feature; represents the historical time period; represents the time interval from the current to the previous optimization moment; e represents a constant with a value of 2.7; Taking the historical optimization value of the lower limit of the associated behavior threshold as the dependent variable, construct the threshold lower limit change curve for each associated behavior feature in chronological order; Extract the trend features from the obtained threshold lower limit change curve and perform weighted averaging to obtain the threshold lower limit change coefficient; Use the threshold lower limit change coefficient to optimize the lower limit of the associated behavior threshold of the current associated behavior feature to obtain the optimized threshold lower limit; Among them, the calculation formula for the optimized threshold lower limit is as follows: ; In the formula, represents the optimized threshold lower limit of the current associated behavior feature; represents the corresponding optimized threshold lower limit at the previous optimization moment of the current associated behavior feature; represents the threshold lower limit change coefficient of the current associated behavior feature; represents the historical time period; represents the time interval from the current to the previous optimization moment; e represents a constant with a value of 2.7; Use the optimized threshold upper limit and the optimized threshold lower limit to establish the initial optimized threshold range of the current key behavior feature; Calculate the deviation of the associated behavior threshold size at the adjacent historical optimization moment of the current associated behavior feature; Combine the deviation of the associated behavior threshold size at the adjacent historical optimization moment with the corresponding time interval at the adjacent historical optimization moment to calculate the threshold-duration ratio; Average all the obtained threshold-duration ratios to obtain the target threshold-duration ratio; Combine the time interval between the previous optimization moment and the current moment with the target threshold-duration ratio to calculate the estimated size of the associated behavior threshold; If the size of the threshold in the initially obtained initial optimization threshold range is not greater than the estimated size of the associated behavior threshold, then output the initially obtained initial optimization threshold range as the target behavior threshold range; If the size of the threshold in the initially obtained initial optimization threshold range is greater than the estimated size of the associated behavior threshold, then gradually and overall reduce the initially obtained initial optimization threshold range according to a set ratio until the threshold size is not greater than the estimated size of the associated behavior threshold, and then output it as the target behavior threshold range.

[0036] In this embodiment, the historical time period is pre-determined, such as 3 months; the historical optimization data refers to the historical optimization records of the associated behavior threshold range of the associated behavior feature; the threshold upper limit change curve is constructed in chronological order with the historical optimization values of the associated behavior threshold upper limit as the dependent variable; the threshold upper limit change coefficient is used to characterize the historical optimization change degree of the associated behavior threshold upper limit of the associated behavior feature, and is obtained by extracting trend features from the obtained threshold upper limit change curve and performing weighted average calculation, where the trend features include curve slope, the sum of the deviations of extreme points from the average value, curvature, and acceleration; the weights assigned to the trend features are obtained by solving the matrix constructed by pairwise comparison and relative importance scoring of the trend features using the analytic hierarchy process; the optimized threshold upper limit is obtained by optimizing the associated behavior threshold upper limit using the threshold upper limit change coefficient.

[0037] In this embodiment, the threshold lower limit change curve is constructed in chronological order with the historical optimization values of the associated behavior threshold lower limit as the dependent variable; the threshold lower limit change coefficient is used to characterize the historical optimization change degree of the associated behavior threshold lower limit of the associated behavior feature, and is obtained by extracting trend features from the obtained threshold lower limit change curve and performing weighted average calculation, where the trend features include curve slope, the sum of the deviations of extreme points from the average value, curvature, and acceleration; the weights assigned to the trend features are obtained by solving the matrix constructed by pairwise comparison and relative importance scoring of the trend features using the analytic hierarchy process; the optimized threshold lower limit is obtained by optimizing the associated behavior threshold lower limit using the threshold lower limit change coefficient.

[0038] In this embodiment, the initial optimization threshold range is composed of an optimization threshold upper limit and an optimization threshold lower limit; the threshold-duration ratio is obtained by dividing the corresponding time interval between adjacent historical optimization times by the deviation of the associated behavior threshold magnitude between adjacent historical optimization times, where the deviation of the associated behavior threshold magnitude refers to the difference between the upper threshold difference and the lower threshold difference of the associated behavior threshold between adjacent historical optimization times; the target threshold-duration ratio is obtained by averaging all the obtained threshold-duration ratios; the threshold magnitude of the initial optimization threshold range is obtained by subtracting the optimization threshold lower limit from the optimization threshold upper limit in the initial optimization threshold range; the set ratio is determined in advance; the estimated associated behavior threshold magnitude is calculated by combining the time interval between the previous optimization time and the current time and the target threshold-duration ratio, and the estimated associated behavior threshold magnitude is expressed as , where is expressed as the estimated associated behavior threshold magnitude; is expressed as the time interval between the previous optimization time and the current time; is expressed as the target threshold-duration ratio.

[0039] In this embodiment, for example, the associated behavior threshold range of the historical optimization time q1 with the associated behavior feature 1 is , and the associated behavior threshold range of the adjacent previous historical optimization time q0 is , then the deviation of the associated behavior threshold magnitude is , and the threshold-duration ratio is expressed as , where is expressed as the time interval between the historical optimization time q1 and the historical optimization time q0.

[0040] The beneficial effects of the above technical solution are: by optimizing the associated behavior threshold range based on the analysis of historical data to obtain the target behavior threshold range, it can help improve the accuracy of abnormal recognition of associated behavior features, and further improve the accuracy of behavior abnormal recognition.

[0041] An embodiment of the present invention provides a behavior anomaly analysis system based on target tracking, as Figure 2 shown, including: Monitoring module: used to use a set shooting tool to monitor the motion state of key targets in different fuel operation scenarios in real time to obtain a first monitoring video; Trajectory analysis module: used to extract target position information from video frames in the first monitoring video by using a set target tracking algorithm to generate a target motion trajectory; Feature analysis module: used to obtain the first behavior feature of the current key target based on the target motion trajectory; Abnormality analysis module: used to perform abnormality analysis on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, and issue an alarm when there is an abnormal behavior.

[0042] The beneficial effects of the above technical solution are as follows: the first monitoring video is obtained by real-time monitoring of the motion states of key targets in different fuel operation scenarios; the first behavior feature of the key target is obtained based on the target motion trajectory generated by extracting the target position information from the video frames in the first monitoring video by using a set target tracking algorithm; by performing abnormality analysis and alarm on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, high-precision tracking of the target object in the monitoring scenario can be realized and the flexibility of behavior abnormality alarm can be improved.

[0043] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A behavior abnormality analysis method based on target tracking, characterized in that: include: Step 1: Use a set shooting tool to monitor the movement status of key targets in different fuel operation scenarios in real time to obtain a first monitoring video; Step 2: extracting the location information of the key target from the video frames in the first surveillance video using a set target tracking algorithm to generate a target motion trajectory; Step 3: Based on the target motion trajectory, obtain the first behavior feature of the current key target; Step 4: Based on the corresponding business logic of the current fuel operation scenario, perform an abnormal analysis on the first behavior feature, and issue an alarm when abnormal behavior exists.

2. A method and system for analyzing abnormal behavior based on target tracking according to claim 1, characterized in that: The motion status of key targets in different fuel operation scenarios is monitored in real time using the set shooting tool to obtain the first monitoring video, including: Set target monitoring points in the corresponding fuel operation areas of different fuel operation scenarios; Installing a set shooting tool at each target monitoring point, the set shooting tool monitors the operating status of the key target in real time to obtain a first monitoring video; The setting shooting tool uploads the acquired first monitoring video to the fuel supervision platform in real time.

3. The method and system for analyzing abnormal behavior based on target tracking according to claim 1, characterized in that: The target tracking algorithm is used to extract the position information of the key target from the video frames in the first monitoring video to generate the target motion trajectory, including: Preprocessing the video frames in the first surveillance video to obtain a key frame sequence; Inputting the first key frame in the key frame sequence into a target detection model pre-established based on deep learning to obtain a target detection result; According to the target detection result and the key frame sequence, the current key target is continuously tracked by using a set target tracking algorithm to obtain the first position information of the current key target in each key frame; Based on all the first position information obtained, a set interpolation algorithm is used to generate the target motion trajectory of the current key target.

4. The method and system for analyzing abnormal behavior based on target tracking according to claim 1, characterized in that: Based on the target motion trajectory, a first behavior feature of the current key target is obtained, including: Performing data standardization processing on the trajectory data of the target running trajectory to obtain first trajectory data; Determine the average speed of the current key target in each preset analysis time period according to the timestamp and position coordinates in the first trajectory data; Determine the acceleration of the current key target based on the average speed of the key target in each preset analysis time period; Calculate a first difference in longitude and latitude between each pair of adjacent points on the target motion trajectory, and convert the first difference to obtain a first direction angle; For all the first direction angles obtained, construct a direction angle change curve according to the order of obtaining; Extract trend features from the direction angle change curve and calculate the weighted average to obtain the direction angle change coefficient; The first position coordinates where the key target stay time exceeds the set time threshold are counted, and the corresponding stay time is regarded as abnormal stay time; The average speed, acceleration, first direction angle, direction angle variation coefficient, abnormal residence time and first position coordinates are output as the first behavior features of the current key target.

5. The method and system for analyzing abnormal behavior based on target tracking according to claim 1, characterized in that: Based on the corresponding business logic of the current fuel operation scenario, the first behavior feature is analyzed for abnormalities, and an alarm is issued when abnormal behavior exists, including: Calculating an associated behavior feature according to the first behavior feature, and normalizing the associated behavior feature to obtain a key-associated behavior feature value; According to the current fuel operation scenario, obtain the corresponding scenario business logic, and then extract the associated behavior threshold list from the scenario business logic; Taking the associated behavior feature as a matching condition, extracting a corresponding associated behavior threshold range from an associated behavior threshold list; Based on the analysis of historical data, the associated behavior threshold range is optimized to obtain the target behavior threshold range; Comparing the key-associated behavior feature value with the corresponding target behavior threshold range, and marking the associated behavior feature whose key-associated behavior feature value belongs to the corresponding target behavior threshold range as a normal associated behavior feature; The associated behavior feature whose key-associated behavior feature value does not fall within the corresponding target behavior threshold range is marked as an abnormal associated behavior feature; If there are abnormal correlation behavior characteristics, it is determined that the current key target has abnormal behavior; Determine the abnormal behavior type of the current key target according to the abnormal associated behavior characteristics; An alarm signal is generated based on the abnormal behavior type and the basic target information of the key target and transmitted to the fuel supervision system for corresponding response.

6. A method and system for analyzing abnormal behavior based on target tracking according to claim 5, characterized in that: The associated behavior characteristics include speed standard deviation, acceleration mean, acceleration standard deviation, direction angle change rate, and abnormal residence time ratio.

7. The method and system for analyzing abnormal behavior based on target tracking according to claim 5, characterized in that: Based on the analysis of historical data, the associated behavior threshold range is optimized to obtain the target behavior threshold range, including: Obtain historical optimization data of the associated behavior threshold range for each associated behavior feature within a historical time period; According to the historical optimization data, the historical optimization value of the upper limit of the associated behavior threshold is used as the dependent variable, and the upper limit threshold change curve of each associated behavior feature is constructed in time series; Extract trend features from the obtained threshold upper limit change curve and perform weighted averaging to obtain the threshold upper limit change coefficient; Optimizing the upper threshold value of the current associated behavior feature by using the upper threshold value variation coefficient to obtain an optimized upper threshold value; The calculation formula for optimizing the upper threshold value is as follows: ; In the formula, It is expressed as the upper limit of the optimization threshold of the current associated behavior characteristics; It is represented as the upper limit of the corresponding optimization threshold at the last optimization moment of the current associated behavior feature; It is expressed as the coefficient of change of the upper threshold limit of the current associated behavior feature; Expressed as a historical time period; It is represented as the time interval between the current time and the last time of optimization; e is represented as a constant with a value of 2.7; Taking the historical optimization value of the lower limit of the associated behavior threshold as the dependent variable, a threshold lower limit change curve of each associated behavior feature is constructed in time series; Extract trend features from the obtained threshold lower limit change curve and perform weighted averaging to obtain the threshold lower limit change coefficient; Utilizing the threshold lower limit variation coefficient, optimizing the associated behavior threshold lower limit of the current associated behavior feature to obtain an optimized threshold lower limit; The calculation formula for optimizing the lower limit of the threshold is as follows: ; In the formula, It is expressed as the optimized lower threshold of the current associated behavior feature; It is represented as the lower limit of the corresponding optimization threshold at the last optimization moment of the current associated behavior feature; It is expressed as the coefficient of change of the lower threshold limit of the current associated behavior feature; Expressed as a historical time period; It is represented as the time interval between the current time and the last time of optimization; e is represented as a constant with a value of 2.7; Using the optimization threshold upper limit and the optimization threshold lower limit, an initial optimization threshold range of the current key behavior feature is established; Calculate the magnitude deviation of the associated behavior threshold of the adjacent historical optimization moments of the current associated behavior feature; The threshold-duration ratio is calculated by combining the magnitude deviation of the associated behavior thresholds of adjacent historical optimization moments with the corresponding time intervals of adjacent historical optimization moments; The target threshold-duration ratio is obtained by averaging all the obtained threshold-duration ratios; The time interval between the last optimization moment and the current moment and the target threshold-duration ratio are combined and calculated to obtain an estimated associated behavior threshold value; If the threshold value of the currently obtained initial optimization threshold range is not greater than the estimated associated behavior threshold value, the obtained initial optimization threshold range is output as the target behavior threshold range; If the threshold size of the currently obtained initial optimization threshold range is greater than the estimated associated behavior threshold size, the obtained initial optimization threshold range is gradually reduced as a whole according to the set ratio until the threshold size is no greater than the estimated associated behavior threshold size, and then output as the target behavior threshold range.

8. A behavior abnormality analysis system based on target tracking, characterized in that: include: Monitoring module: used to monitor the motion state of key targets in different fuel operation scenarios in real time by using a set shooting tool to obtain a first monitoring video; Trajectory analysis module: used to extract target position information from the video frames in the first monitoring video by using a set target tracking algorithm to generate a target motion trajectory; Feature analysis module: used for obtaining the first behavior feature of the current key target based on the target motion trajectory; Abnormal analysis module: used to perform abnormal analysis on the first behavior feature based on the corresponding business logic of the current fuel operation scenario, and issue an alarm when abnormal behavior exists.

Images