An overweight vehicle tracking and identification management system based on image recognition

Abstract

The invention relates to the technical field of tracking and identifying overload control vehicles, and specifically to an overload control vehicle tracking and identifying management system based on image recognition, comprising: a data acquisition module, used for identifying a target vehicle and acquiring video data of the target vehicle; a size analysis module, used for analyzing the video data of the target vehicle, acquiring size information corresponding to the vehicle appearance, analyzing the size information, and obtaining a size analysis result; a behavior mapping module, used for analyzing the video data, extracting at least one related behavior vector of the target vehicle; decomposing the behavior vector, and acquiring a mapping behavior interval corresponding to the behavior vector; the behavior vector includes the relative change amount of the driving time, position, speed and vehicle posture of the target vehicle; the accuracy and reliability of vehicle monitoring are achieved, and the efficiency of overload control management is improved.

Description

Technical Field

[0001] The invention relates to the technical field of tracking and identifying overweight vehicles, and in particular to an overweight vehicle tracking and identifying management system based on image recognition. Background Art

[0002] In traditional overload vehicle management, manual measurement, weighing or simple sensor data are often relied on to determine whether a vehicle is overloaded. However, these methods have many shortcomings, such as inaccurate measurement, incomplete data, and non-real-time monitoring. With the continuous development of image recognition technology, its application in the field of vehicle monitoring and management is becoming more and more extensive. Based on this background, the current scheme proposes an overload vehicle tracking and identification management system based on image recognition, aiming to solve the problems existing in traditional methods and improve the efficiency and accuracy of overload management.

[0003] For example, Chinese patent publication number CN118643994A discloses a digital city overload and overload control supervision system, which relates to the technical field of overload and overload control supervision. The present invention performs overload and overload monitoring on each passing vehicle on the monitored road, screens out each overloading vehicle and each overloaded vehicle, and then calculates the corresponding driving safety value of each overloading vehicle and the corresponding driving safety value of each overloaded vehicle according to the driving information of each overloading vehicle and the passing information of the overloaded vehicle, and obtains the education type and education level of each overloading vehicle and the overloaded vehicle according to the historical education information, and then analyzes and sends the online education courses corresponding to each online-educated overloading driver and each online-educated overloaded driver, which greatly reduces the labor cost of road supervision, accurately screens out the education type that is effective for drivers, improves the accuracy and effect of driver education, and then reduces the number of subsequent violations by drivers, reduces the occurrence of traffic accidents, and maintains the safety of the transportation system.

[0004] Existing technologies tend to process driving information and user education. However, in the case of overload management, it is necessary to identify the corresponding size of the vehicle driven by the current user and related movement information in real time to determine whether the vehicle driven by the current user has corresponding violations, and track the relevant vehicles in time to improve the effectiveness of traffic control and management. Summary of the invention

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: an overweight vehicle tracking and identification management system based on image recognition, including: a data acquisition module for identifying a target vehicle and acquiring video data of the target vehicle.

[0006] The dimension analysis module is used to analyze the video data of the target vehicle, obtain the dimension information corresponding to the vehicle appearance, analyze the dimension information, and obtain the dimension analysis result.

[0007] The behavior mapping module is used to analyze the video data and extract at least one relevant behavior vector of the target vehicle; decompose the behavior vector and obtain the mapping behavior interval corresponding to the behavior vector; the behavior vector includes the relative change of the target vehicle's driving time, position, speed and vehicle posture.

[0008] The behavior analysis module is used to analyze the behavior vector and determine the changing trend of the mapping behavior in the mapping behavior interval within a unit time. The changing trend includes the behavior characteristic variation, speed relative quantity, posture relative quantity, behavior conversion rate and behavior warning rate. According to the changing trend of the mapping behavior, the trend analysis coefficient is calculated.

[0009] The behavior tendency module is used to analyze the behavior vector, determine the behavior tendency of the target vehicle when it moves, decompose the behavior tendency of the target vehicle, and obtain the tendency analysis coefficient.

[0010] The vehicle evaluation module is used to analyze and identify the changing trend of the mapping behavior, determine the probability distribution of the mapping behavior, and obtain the vehicle evaluation result based on the probability distribution of the mapping behavior, the tendency analysis coefficient and the size analysis result.

[0011] The beneficial effects of the present invention are as follows: the present invention realizes real-time acquisition and accurate judgment of vehicle appearance dimensions through image recognition technology, thereby improving the accuracy of over-limit identification; through the collaborative work of multiple modules, a comprehensive analysis of vehicle behavior is performed, thereby improving the accuracy and reliability of vehicle monitoring; the output results of multiple modules are integrated to conduct a comprehensive assessment of the vehicle, thereby improving the accuracy and credibility of the assessment results; through an automated and intelligent management system, the labor cost is reduced and the efficiency of over-limit management is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0013] Figure 1 It is a system framework diagram of an overweight vehicle tracking and identification management system based on image recognition.

[0014] Figure 2 The present invention is a system diagram of an overweight vehicle tracking and identification management system based on image recognition.

[0015] Figure 3 The present invention is a flow chart of trend analysis coefficients in a behavior analysis module of an overload vehicle tracking and identification management system based on image recognition.

[0016] Figure 4 The present invention is a flow chart of tendency analysis coefficient in a behavior tendency module of an overload vehicle tracking and identification management system based on image recognition. DETAILED DESCRIPTION

[0017] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0018] See also Figure 1 , Figure 2 , an overweight vehicle tracking and identification management system based on image recognition, including: a data acquisition module, a size analysis module, a behavior mapping module, a behavior analysis module, a behavior tendency module and a vehicle evaluation module.

[0019] The data acquisition module outputs the identified video data to the size analysis module; the size analysis module is used to identify whether the size of the vehicle is normal at this time, and transmits the corresponding data to the behavior mapping module, the behavior mapping module performs behavior analysis on the current vehicle to divide the data interval corresponding to the behavior, and outputs it to the behavior analysis module; the behavior analysis module is used to perform trend analysis on the vehicle's behavior, and outputs the corresponding data to the behavior tendency module; after completing the processing of the behavior tendency, the behavior tendency module outputs it to the vehicle evaluation module, and the vehicle evaluation module completes the tracking and identification analysis of the current vehicle.

[0020] The data acquisition module is used to identify the target vehicle and acquire the video data of the target vehicle.

[0021] The dimension analysis module is used to analyze the video data of the target vehicle, obtain the dimension information corresponding to the vehicle appearance, analyze the dimension information, and obtain the dimension analysis result.

[0022] The behavior mapping module is used to analyze the video data, extract at least one relevant behavior vector of the target vehicle, decompose the behavior vector, and obtain the mapping behavior interval corresponding to the behavior vector.

[0023] The behavior analysis module is used to analyze the behavior vector and determine the changing trend of the mapping behavior in the mapping behavior interval within a unit time. The changing trend includes the behavior characteristic variation, speed relative quantity, posture relative quantity, behavior conversion rate and behavior warning rate. According to the changing trend of the mapping behavior, the trend analysis coefficient is calculated.

[0024] The behavior tendency module is used to analyze the behavior vector, determine the behavior tendency of the target vehicle when it moves, decompose the behavior tendency of the target vehicle, and obtain the tendency analysis coefficient.

[0025] The vehicle evaluation module is used to analyze and identify the changing trend of the mapping behavior, determine the probability distribution of the mapping behavior, and obtain the vehicle evaluation result based on the combination of the probability distribution of the mapping behavior, the tendency analysis coefficient and the size analysis result.

[0026] In one embodiment of the present invention, the data acquisition module is used to identify the target vehicle and acquire video data of the target vehicle.

[0027] At this time, the camera will be used to obtain the vehicle's video data to identify the vehicle's corresponding appearance and whether the vehicle is overloaded or over-limited under the current circumstances; the acquired video data includes static pictures and continuous video frames, which are mainly used to identify the vehicle border and the trajectory of the vehicle's movement to determine whether the vehicle is overloaded, over-limited, or has abnormal overtaking when moving in the video data, and to track the vehicle in a timely manner to achieve the effect of vehicle management and control.

[0028] In one embodiment of the present invention, a dimensional analysis module is used to analyze video data of a target vehicle, obtain dimensional information corresponding to the vehicle's appearance, analyze the dimensional information, and obtain dimensional analysis results to determine whether the length, width or height of the vehicle exceeds a specified standard, such as identifying whether the cargo currently placed on the vehicle exceeds the boundary of the vehicle compartment and extends to the sides or rear, thereby obtaining dimensional analysis results.

[0029] In this module, the size information is mainly used to determine whether the current vehicle is obviously overloaded. Overload will directly identify whether the cargo of the current truck exceeds the edge of the car body, and whether the size of the current vehicle exceeds the regulations. At the same time, the identified vehicle size can assist in vehicle overload management. For example, based on the current vehicle size, it can assist in determining the relative positions between multiple vehicles to identify the relative deviation angles and relative distances of the vehicles. Based on these values, the vehicle's behavior vector can be classified to obtain a more complete description of the current vehicle.

[0030] Therefore, the implementation method of the size analysis module also includes identifying the target vehicle according to the timestamp of the video data, determining the size coordinates of the target vehicle at each timestamp, obtaining the size information of the target vehicle, comparing the size information of the target vehicle with the standard size, determining whether the target vehicle exceeds the limit, and outputting the over-limit size information as the size analysis result.

[0031] At this time, the contour recognition algorithm will be used to identify the dimensional coordinates corresponding to the length, width, and height of the vehicle, calculate the size information of the current vehicle, and obtain the actual length, actual width, and actual height of the vehicle. These values ​​will be compared with the vehicle's carrying standards to determine whether the vehicle has corresponding problems at this time. At the same time, when the value in the dimensional information is too large, it will also affect the judgment and processing of the vehicle's behavior, resulting in changes in the distance between the identified vehicles and the relative deviation angle.

[0032] When the dimensional analysis results are output, the specific excess position and the excess value will be output to determine the actual length, actual width, and actual height of the current vehicle; when the size of the target vehicle meets the requirements, the output value of the dimensional analysis result is 0. When the size of the target vehicle exceeds the standard, the values ​​exceeding the standard will be summed, and the summed values ​​will be normalized and output as the dimensional analysis results.

[0033] In one embodiment of the present invention, a behavior mapping module is used to analyze video data and extract at least one relevant behavior vector of the target vehicle; the behavior vector is used to describe the current driving behavior of the target vehicle, and the behavior vector is decomposed to obtain a mapping behavior interval corresponding to the behavior vector.

[0034] The purpose of vehicle behavior vector analysis in this module is to determine whether the vehicle is overloaded while driving. For example, analyzing the vehicle's driving status, such as acceleration, deceleration, and turning behavior, can indirectly determine whether the vehicle may be overloaded. Overloaded vehicles may appear to be more laborious when starting, the braking distance will be longer, and the stability will be poor when turning. At this time, it is necessary to analyze the vehicle's driving behaviors, determine the time, position, speed of these behaviors, and whether the vehicle's posture has changed, and can also identify whether the vehicle is sinking and the tires are deformed.

[0035] The behavior vector includes the target vehicle's travel time, position, speed, and relative change in vehicle posture.

[0036] The relative change in driving time is expressed as the time taken by a vehicle to travel from one monitoring point to another. It is used to determine the average moving speed of the target vehicle to identify whether the vehicle is speeding or has abnormal behavior.

[0037] The relative change in position is used to describe the geographic location of the vehicle, record the position of the vehicle with other vehicles, and the corresponding distance relationship, such as the distance between the center of mass of the vehicles, to identify whether the vehicle has abnormal behavior while driving.

[0038] The relative change in speed is used to indicate the real-time speed of the vehicle under monitoring, focusing on the instantaneous speed of the vehicle, and is used to identify the speed of the vehicle while driving.

[0039] The relative change of vehicle attitude is used to indicate the change of vehicle attitude during driving, including changes in pitch angle, roll angle and yaw angle; pitch angle: the angle of vehicle tilting forward and backward; roll angle: the angle of vehicle tilting left and right; yaw angle: the angle of vehicle rotation around the vertical axis; By analyzing the change of vehicle attitude, the stability of the vehicle can be judged. For example, an overloaded vehicle may have a large roll angle change when turning, indicating that the vehicle is unstable; and when braking, the pitch angle may change greatly, indicating that the vehicle's braking performance is affected.

[0040] The mapping behavior interval is represented as a driving time interval, a position interval, a speed interval and a posture change interval. The mapping behavior interval defines the range of each feature in the current behavior feature by the identified value; the analysis of the current vehicle behavior is completed based on the analysis of the feature value; these mapping intervals correspond to the driving time, position, speed and vehicle posture respectively.

[0041] In the process of decomposing the behavior vector, the segmentation points of the behavior interval can also be determined. These segmentation points divide the behavior vector into different segments or intervals, each of which corresponds to a specific behavior or behavior stage. The segmentation point data helps to more accurately locate and analyze the change points and turning points of the behavior.

[0042] For example, the behavior vector is decomposed and the implementation method of obtaining the mapping behavior interval corresponding to the behavior vector can be as shown below: the first behavior interval and the second behavior interval of the behavior vector are obtained, the behavior vector is divided according to the first behavior interval and the second behavior interval, the behavior vector is mapped and analyzed according to the first behavior interval and the second behavior interval, and the behavior vector after the mapping analysis is output as a mapping behavior interval.

[0043] The first behavior interval is used to represent the main behavior interval of the target vehicle during driving, and represents the normal interval value corresponding to the driving time, position, speed, and vehicle posture of the vehicle under normal circumstances.

[0044] The second behavior interval indicates a behavior interval for selecting a behavior interval where a large change occurs at an adjacent position of the first behavior interval, and is used to describe a portion where the behavior of the target vehicle changes greatly during driving.

[0045] The mapping analysis can be performed in a manner that a mapping relationship is constructed between the first behavior interval, the second behavior interval, and the behavior vector according to the time and range of the values, and the behavior labels corresponding to the first behavior interval and the second behavior interval are determined. The behavior labels can be expressed as high-speed driving, low-speed driving, low driving distance, driving on the left, driving on the right, large driving deflection angle, small driving deflection angle, straight driving, turning, changing lanes, overtaking, and the like. Based on the obtained behavior labels, the behavior stage of the current behavior vector is determined, and the first behavior interval and the second behavior interval corresponding to each behavior stage are output as mapping behavior intervals.

[0046] At this time, the behavior stage mainly indicates the specific movement stage of the behavior vector combined by the current behavior label, to describe the specific movement of the current vehicle. According to this specific movement, the current vehicle status can be obtained.

[0047] In one embodiment of the present invention, a behavior analysis module is used to analyze the behavior vector and determine the changing trend of the mapping behavior in the mapping behavior interval within a unit time. The changing trend includes the behavior characteristic variation, speed relative quantity, posture relative quantity, behavior conversion rate and behavior warning rate. According to the changing trend of the mapping behavior, the trend analysis coefficient is calculated.

[0048] In this module, mapping behavior is used to describe the changes in driving time, position, speed, and vehicle posture in the behavior vector within a unit time. For example, to determine the change in driving time in the behavior vector at this time, the corresponding speed difference between two time points can be calculated to quantify the relative speed; the change in vehicle posture uses the relative posture quantity, which mainly identifies the posture difference between the current vehicle and the adjacent vehicles; the behavior conversion rate is used to determine the frequency of changes in the current target vehicle behavior stage and the changes in vehicle movement behavior; the behavior warning rate is the probability of reaching the boundary warning value within the mapping behavior interval corresponding to the behavior vector; the relative speed quantity and the relative posture quantity are calculated relative to a certain standard value; the behavior feature variation is the difference between the four features of driving time, position, speed, and vehicle posture at each time point relative to the average value, and these differences are combined to determine the behavior feature variation.

[0049] At the same time, the above mapping behavior is achieved by selecting a behavior represented by each behavior label under the mapping behavior interval as the mapping behavior to be identified at this time. The mapping behavior is to decompose the changes in the target vehicle during driving into multiple small behaviors, and analyze these small behaviors to identify whether overloading, overtaking and other behaviors occur.

[0050] like Figure 3 As shown, the trend analysis coefficient may be realized by obtaining the periodic trend coefficient corresponding to the behavior conversion rate and the behavior warning rate.

[0051] Obtain the trend correlation coefficients corresponding to the behavior characteristic variation, speed relative quantity and posture relative quantity.

[0052] The periodic trend coefficient is used to analyze the behavior conversion rate and behavior warning rate, and determine the corresponding changes in the mapping behavior within a unit time; the trend correlation coefficient is used to analyze the behavior characteristic variation, speed relative quantity and posture relative quantity, and obtain the trend of the mapping behavior in terms of value.

[0053] The periodic trend coefficient and the trend correlation coefficient are calculated comprehensively to obtain the trend analysis coefficient.

[0054] The behavior feature variation is expressed as: obtaining the index values ​​of driving time, position, speed and vehicle posture in the behavior vector, normalizing the index values ​​of driving time, position, speed and vehicle posture, and performing weighted summation to obtain the behavior feature variation.

[0055] The behavioral characteristic variation can be expressed as: Among them, δ 1 Indicates the variation of behavioral characteristics, Pt in Indicates the normalized driving time index value, Tt in Represents the index value of the normalized position, V in The index value representing the normalized speed, θ in represents the index value of the normalized vehicle posture, The standard value of the index value indicating the speed, The standard value of the index value representing the vehicle posture, α 1 represents the weight factor of travel time and location, α 2 Represents the weight factor of speed, α 3 The weight factors representing the vehicle posture can be set to 0.3, 0.4, and 0.3 in the order of driving time and position, speed, and vehicle posture.

[0056] The index value for driving time represents the time difference between adjacent mapping behaviors in unit time, the index value for position represents the difference in moving distances between adjacent mapping behaviors, the index value for speed represents the difference in speeds between adjacent mapping behaviors, and the index value for vehicle posture represents the sum of the differences in pitch angle, roll angle, and yaw angle between adjacent mapping behaviors.

[0057] The purpose of dividing the driving time index value by the position index value at this time is to determine the change in the acceleration of the target vehicle at this time. When an overloaded vehicle accelerates, the acceleration will be smaller, so as to verify whether it is normal; the speed index value is mainly used to verify the speed stabilization time and the value at which the speed stabilizes; the vehicle posture is to verify whether the current vehicle will have a larger body tilt when turning due to its high center of gravity and large mass, so as to identify whether the current vehicle behavior is normal.

[0058] The relative amount of speed is expressed as follows: the travel time, position, and speed in the behavior vector are obtained, the average moving speed and instantaneous speed of the target vehicle are calculated, and the relative amount of speed is calculated.

[0059] For example, the relative speed is expressed as: Among them, δ 2Relative speed, represents the average moving speed, V′ represents the instantaneous speed, τ 1 Represents the speed relative coefficient. At this time, the speed relative coefficient can be set to 1.5 to amplify the changes between the corresponding speeds.

[0060] The relative speed quantity mainly analyzes the speed change of the vehicle, which helps to identify whether the vehicle has frequent speed fluctuations and then determine whether the vehicle's behavior is stable.

[0061] The relative quantity of the posture is expressed as obtaining the heading angle, pitch angle and roll angle of the vehicle posture in the behavior vector, and calculating the relative quantity of the posture of the target vehicle; at this time, Euler angles can be used to express it, which uses three angles, such as heading angle, pitch angle and roll angle to describe the rotation of one coordinate system relative to another coordinate system, to obtain the relative quantity of the posture at this time.

[0062] For example, the relative quantity of the posture is expressed as obtaining the weighted average of the heading angle, pitch angle and roll angle in the vehicle posture as the relative quantity of the posture. At this time, the weights used can be 0.4, 0.4, and 0.2 to complete the description of the vehicle posture in sequence.

[0063] The behavior conversion rate is expressed as: obtaining the number of conversions of the mapped behavior per unit time, and setting the number of conversions of the mapped behavior as the behavior conversion rate; at this time, it is determined whether the mapped behavior interval changes under the corresponding mapped behavior. If it changes, it is recorded as one conversion number; the behavior conversion rate refers to the frequency of the vehicle converting from one behavior interval to another behavior interval per unit time; the identified number of conversions is divided by the length of the unit time period to obtain the behavior conversion rate at this time.

[0064] The behavior warning rate is expressed as: obtain the standard threshold of the behavior vector and calculate the behavior warning rate. The behavior warning rate indicates the ratio of mapping behaviors that require warning to the total number of behaviors. The mapping behaviors that require warning can be selected from those that differ from the standard threshold by less than 5%. If the indicator value corresponding to the mapping behavior is greater than the standard threshold, it is also regarded as a mapping behavior that requires warning, and the ratio of these data that require warning to the total data is taken as the value of the behavior warning rate at this time.

[0065] The periodic trend coefficient is expressed as: After normalizing the behavior conversion rate and the behavior warning rate, the periodic trend coefficient is calculated.

[0066] Among them, γ 1 represents the periodic trend coefficient, δ 4 represents the behavior conversion rate, δ 5 represents the behavioral warning rate, Indicates the weight coefficient of behavior conversion rate, Represents the weight coefficient of the behavior warning rate; the weight coefficients of the behavior conversion rate and the behavior warning rate used here are expressed as the average value of the weight coefficients of the behavior conversion rate and the behavior warning rate in the historical data.

[0067] This periodic trend coefficient tends to identify changes in mapping behavior within a unit time, to identify changes in the current vehicle's driving conditions within a relatively short period of time. At this time, the selected mapping behavior will select behaviors that change within 5 seconds, mainly identifying changes in a shorter period of time, such as changes in the vehicle within 5 minutes, or the comprehensive changes in the mapping behavior in a shorter period of time.

[0068] The trend correlation coefficient is expressed as follows: the behavior characteristic variation, the speed relative quantity and the posture relative quantity are obtained, and after normalizing the behavior characteristic variation, the speed relative quantity and the posture relative quantity, the trend correlation coefficient is calculated.

[0069] Among them, γ 2 represents the trend correlation coefficient, δ 1 represents the variation of behavioral characteristics, δ 2 Represents the relative speed, δ 3 Indicates the relative amount of posture, The weight coefficient representing the variation of behavioral characteristics, The weight coefficient representing the relative amount of speed, represents the weight coefficient of the relative amount of posture, and e represents the exponential constant; the weight coefficients of the behavioral characteristic variation, the relative amount of speed, and the relative amount of posture are all the average values ​​of the corresponding weight coefficients in the historical data.

[0070] The trend correlation coefficient mainly determines the relationship between specific values ​​in the behavior vector to reflect the similarity of the behavior vector in terms of driving time, position, speed, vehicle posture, etc. When the trend correlation coefficient is large, it means that the correlation between these values ​​is large, which helps to identify the similarities and potential connections between the mapped behaviors of the target vehicle, and facilitates the judgment of whether the trend presented by the target vehicle is normal.

[0071] The trend analysis coefficient is expressed as: the period trend coefficient and the trend correlation coefficient are calculated comprehensively to obtain the trend analysis coefficient.

[0072] Among them, γ 3 Expressed as the trend analysis coefficient, represents the weight coefficient of the periodic trend coefficient, Represents the weight coefficient of the trend correlation coefficient; at this time, the weight coefficients of the period trend coefficient and the trend correlation coefficient can be set to 0.5 and 0.5 respectively.

[0073] By outputting the trend correlation coefficient, the behavior analysis of the current target vehicle can be completed, the subsequent prediction performance and explanation of vehicle behavior can be improved, and abnormal situations in the vehicle can be discovered in time.

[0074] In one embodiment of the present invention, the behavior tendency module is used to analyze the behavior vector, determine the behavior tendency of the target vehicle when moving, decompose the behavior tendency of the target vehicle, and obtain the tendency analysis coefficient.

[0075] In this module, the behavior tendency can be decomposed into obtaining the speed fluctuation value, the distance from the center of mass, and the relative deviation angle of the vehicle to obtain the behavior tendency of the target vehicle; the speed fluctuation value can be expressed as the fluctuation of the difference between the vehicle speeds of the target vehicle and the adjacent vehicles on the average value. For example, the average value of the difference between the speeds of the target vehicle and the adjacent vehicles is first calculated to obtain the average speed difference, and then the average value of the average speed difference and the sum of the squares of the speed differences between the target vehicle and the adjacent vehicles is calculated as the speed fluctuation value at this time; the distance from the center of mass is expressed as the average value of the distance between the center points of the target vehicle and the adjacent vehicles, and the relative deviation angle represents the average value of the angle difference between the target vehicle and the adjacent vehicles; to obtain the safety assessment value existing at this time.

[0076] At the same time, it verifies whether the current vehicle has micro-behaviors. The obtained micro-behaviors include the safety assessment value, behavior characteristic value, and spacing change value of the current target vehicle, and the evaluation result of the current behavior is obtained, thereby completing the tendency analysis of the vehicle; micro-behaviors indicate whether the target vehicle has undergone multiple changes in a short period of time, and the behavior stage of the vehicle has changed according to the multiple changes; at this time, the safety assessment value, behavior characteristic value, spacing change value and micro-behavior threshold are compared to identify the current micro-behavior. When the set threshold is exceeded, the relevant data at this time is marked, and the behavior tendency of the current vehicle is analyzed according to these data values ​​to obtain the vehicle situation that needs to be identified, so as to identify and monitor the current vehicle to prevent the problem of incorrect identification of the corresponding vehicle during overload tracking.

[0077] The behavior feature value represents the difference between the target vehicle and the adjacent vehicles after the behavior vector is modulated. At this time, all values ​​in the behavior vector are first processed by standard processing so that all values ​​are dimensionless and then calculated. At this time, the main comparison is between the target vehicle's driving time, position, speed and vehicle posture index values ​​and those of the adjacent vehicles.

[0078] The distance change value indicates the distance between the target vehicle and the adjacent vehicle in actual size. This distance value is used to verify the distance between the vehicles.

[0079] like Figure 4As shown, the behavior tendency of the target vehicle is decomposed, and the implementation method of obtaining the tendency analysis coefficient can be expressed as obtaining the speed fluctuation value, the distance from the center of mass and the relative deviation angle corresponding to the behavior tendency.

[0080] After normalizing the speed fluctuation value, the distance from the center of mass and the relative deviation angle, the safety assessment value is calculated.

[0081] The safety assessment value, behavior characteristic value and spacing change value of the target vehicle are obtained, and a micro-behavior threshold is set. The micro-behavior threshold includes a safety assessment threshold, a behavior characteristic threshold and a spacing change threshold. At this time, the safety assessment threshold, the behavior characteristic threshold and the spacing change threshold are expressed as the threshold part of the safety assessment value, the behavior characteristic value and the spacing change value, which are used to determine whether the target vehicle exceeds the corresponding threshold when being monitored. The safety assessment value, the behavior characteristic value and the spacing change value of the target vehicle are compared with the micro-behavior threshold to obtain the number of micro-behavior occurrences and the duration under the corresponding micro-behavior threshold.

[0082] The number of occurrences and duration of micro-behaviors are monitored, the behavioral tendency of the target vehicle is identified based on the cumulative effect of micro-behaviors, and the tendency analysis coefficient of the target vehicle is calculated.

[0083] The safety assessment threshold, behavior feature threshold, and spacing change threshold in the micro-behavior threshold can select the average value of the top 60% of the historical data as the micro-behavior threshold used at this time; at the same time, the part that exceeds the micro-behavior threshold can be used to identify the part of the current micro-behavior that has changed, and the number and duration of changes in this part can be analyzed to obtain the tendency analysis coefficient at the corresponding frequency.

[0084] The safety assessment value is expressed as: SAV = α STV ×STV+α DTC ×(1-DTC)+α RAD ×RAD; where SAV represents the safety assessment value, STV represents the speed fluctuation value, DTC represents the distance from the center of mass, RAD represents the relative deviation angle, α STV Represents the weight coefficient of speed fluctuation value, α DTC Represents the weight coefficient of the distance from the centroid, α RAD Represents the weight coefficient of the relative deflection angle; the weight coefficient is set to 0.4, 0.3, and 0.3 respectively according to the speed fluctuation value, the distance from the center of mass, and the relative deflection angle.

[0085] The tendency analysis coefficient is expressed as: the weight corresponding to the duration of the micro-behavior is obtained, and the tendency analysis coefficient is calculated according to the safety assessment value, behavior characteristic value, interval change value and number of occurrences of the current micro-behavior.

[0086] Among them, PA represents the tendency analysis coefficient, SAV i represents the safety assessment value of the i-th micro-behavior, BV i represents the behavior characteristic value of the i-th micro-behavior, SV i represents the spacing change value of the i-th micro-behavior, n represents the number of micro-behaviors, i ranges from 1 to n, and w i represents the weight corresponding to the duration of the i-th micro-behavior, α SAV,i represents the weight coefficient of the safety assessment value of the i-th micro-behavior, α BV,i represents the weight coefficient of the behavior characteristic value of the i-th micro-behavior, α SV,i The weight coefficient representing the spacing change value of the i-th micro-behavior. The multiple weights set at this time can be selected from historical data to complete the current trend analysis coefficient; in the current formula, these identified data will be added according to the cumulative effect of micro-behaviors to determine the comprehensive situation of these values ​​to complete the trend analysis.

[0087] In the behavior tendency module, the relationship between the target vehicle and the adjacent vehicles is analyzed to verify whether the changes of the target vehicle under different mapping behaviors in the behavior analysis module are consistent, so as to analyze the driving conditions of the target vehicle on the road to determine the degree of influence of the target vehicle by the adjacent vehicles when the corresponding abnormal behavior occurs.

[0088] In one embodiment of the present invention, a vehicle evaluation module is used to analyze and identify the changing trend of the mapping behavior, determine the probability distribution of the mapping behavior, and obtain the vehicle evaluation result based on the probability distribution of the mapping behavior, the tendency analysis coefficient and the index value of the size analysis result.

[0089] In this module, the probability distribution of the mapping behavior is actually the probability distribution value of the trend analysis coefficient. Based on the probability distribution value of the trend analysis coefficient, the tendency analysis coefficient and the size analysis result are combined to obtain the vehicle evaluation result. The vehicle evaluation result is used to judge the comprehensive situation of the current vehicle, and to assist in tracking the overtaking vehicle, so as to timely determine whether the current vehicle has changed, timely identify the vehicle's driving problems, and quickly locate the target vehicle.

[0090] Therefore, the vehicle evaluation result indicates that the probability distribution value of the trend analysis coefficient is obtained, and the probability distribution value of the trend analysis coefficient is used as the probability distribution of the mapping behavior. The vehicle evaluation coefficient is calculated according to the probability distribution of the mapping behavior, the tendency analysis coefficient and the index value of the size analysis result, and the vehicle evaluation coefficient is output as the vehicle evaluation result; the probability distribution value of the trend analysis coefficient is calculated through the probability density function in the normal distribution, that is, the current trend distribution coefficient value is input and calculated using the standard deviation and mean value of the historical data.

[0091] Among them, VER represents the index value of the vehicle evaluation result, P(γ) represents the probability distribution of the mapping behavior, PA represents the tendency analysis coefficient, DAR represents the index value of the dimensional analysis result, and the index value of the dimensional analysis result here is the corresponding value in the above-mentioned dimensional analysis result; β, χ, λ represent adjustment coefficients, and e represents the exponential constant; at this time, the value of the adjustment coefficient can be set to 2, 0.5, and 0.7 respectively.

[0092] The vehicle evaluation coefficient output at this time is a comprehensive situation of the numerical values ​​representing the probability distribution of the mapping behavior, the tendency analysis coefficient and the size analysis result, and these values ​​are combined into a vehicle evaluation coefficient. The vehicle evaluation coefficient is finally output to the external terminal as the vehicle evaluation result. According to this value, the current vehicle status can be evaluated to realize the tracking and monitoring of the vehicle.

[0093] If the final vehicle assessment result is different from the actual estimated value, the corresponding vehicle will be marked and the road conditions at that time will be obtained to compare the reason for the abnormal movement of the current target vehicle, so as to identify whether the current vehicle has speeding, overloading and other behaviors, and determine whether the vehicle needs to be tracked further.

[0094] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. An image recognition-based tracking and identification management system for overweight vehicles, characterized in that: include: A data acquisition module is used to identify the target vehicle and acquire video data of the target vehicle; The dimension analysis module is used to analyze the video data of the target vehicle, obtain the dimension information corresponding to the vehicle appearance, analyze the dimension information, and obtain the dimension analysis result; A behavior mapping module, used to analyze the video data and extract at least one relevant behavior vector of the target vehicle; Decomposing the behavior vector to obtain the mapping behavior interval corresponding to the behavior vector; The behavior vector includes the target vehicle’s travel time, position, speed, and relative changes in vehicle posture; The behavior analysis module is used to analyze the behavior vector and determine the change trend of the mapping behavior in the mapping behavior interval within a unit time. The change trend includes the behavior characteristic variation, speed relative quantity, posture relative quantity, behavior conversion rate and behavior warning rate. According to the change trend of the mapping behavior, the trend analysis coefficient is calculated; The behavior tendency module is used to analyze the behavior vector, determine the behavior tendency of the target vehicle when it moves, decompose the behavior tendency of the target vehicle, and obtain the tendency analysis coefficient; The vehicle evaluation module is used to analyze and identify the changing trend of the mapping behavior, determine the probability distribution of the mapping behavior, and obtain the vehicle evaluation result based on the probability distribution of the mapping behavior, the tendency analysis coefficient and the size analysis result.

2. According to the image recognition-based tracking and identification management system for overweight vehicles according to claim 1, it is characterized in that: The implementation method of the size analysis module also includes identifying the target vehicle according to the timestamp of the video data, determining the size coordinates of the target vehicle at each timestamp, obtaining the size information of the target vehicle, comparing the size information of the target vehicle with the standard size, determining whether the target vehicle exceeds the limit, and outputting the over-limit size information as the size analysis result.

3. According to the image recognition-based tracking and identification management system for overweight vehicles according to claim 1, it is characterized in that: The behavior vector is decomposed to obtain the mapping behavior interval corresponding to the behavior vector as follows: Obtaining a first behavior interval and a second behavior interval of the behavior vector, dividing the behavior vector according to the first behavior interval and the second behavior interval; performing mapping analysis on the behavior vector according to the first behavior interval and the second behavior interval, and outputting the behavior vector after the mapping analysis as a mapping behavior interval; The mapping analysis is performed by constructing a mapping relationship between the first behavior interval, the second behavior interval, and the behavior vector according to the time and range of the value, and determining the behavior labels corresponding to the first behavior interval and the second behavior interval; According to the obtained behavior label, the behavior stage of the current behavior vector is determined, and the first behavior interval and the second behavior interval corresponding to each behavior stage are output as mapping behavior intervals.

4. The image recognition-based tracking and identification management system for overweight vehicles according to claim 1 is characterized in that: The trend analysis coefficient is implemented as follows: Obtain the periodic trend coefficients corresponding to the behavior conversion rate and behavior warning rate; Obtain trend correlation coefficients corresponding to behavior characteristic variation, speed relative quantity and posture relative quantity; The periodic trend coefficient and the trend correlation coefficient are calculated comprehensively to obtain the trend analysis coefficient.

5. The image recognition-based tracking and identification management system for overweight vehicles according to claim 4 is characterized in that: The periodic trend coefficient is expressed as, after normalizing the behavior conversion rate and the behavior warning rate, the periodic trend coefficient is calculated; Among them, γ1 represents the periodic trend coefficient, δ4 represents the behavior conversion rate, and δ5 represents the behavior warning rate. Indicates the weight coefficient of behavior conversion rate, The weight coefficient representing the behavior warning rate; The trend correlation coefficient is expressed as follows: the behavior characteristic variation, the speed relative quantity and the posture relative quantity are obtained, and the behavior characteristic variation, the speed relative quantity and the posture relative quantity are normalized to calculate the trend correlation coefficient; Among them, γ2 represents the trend correlation coefficient, δ1 represents the behavioral characteristic variation, δ2 represents the relative speed, and δ3 represents the relative posture. The weight coefficient representing the variation of behavioral characteristics, The weight coefficient representing the relative amount of speed, It represents the weight coefficient of the relative amount of posture, and e represents the exponential constant.

6. The image recognition-based tracking and identification management system for overweight vehicles according to claim 5 is characterized in that: The trend analysis coefficient is expressed as, the period trend coefficient and the trend correlation coefficient are calculated comprehensively to obtain the trend analysis coefficient; Among them, γ3 represents the trend analysis coefficient, represents the weight coefficient of the periodic trend coefficient, Represents the weight coefficient of the trend correlation coefficient.

7. The image recognition-based tracking and identification management system for overweight vehicles according to claim 1 is characterized in that: The behavior tendency of the target vehicle is decomposed and the tendency analysis coefficient is obtained as follows: Obtain the speed fluctuation value, distance from the center of mass, and relative deviation angle corresponding to the behavior tendency; After normalizing the speed fluctuation value, the distance from the center of mass, and the relative deviation angle, the safety assessment value is calculated; Obtain the safety assessment value, behavior characteristic value and distance change value of the target vehicle, set the micro-behavior threshold, which includes the safety assessment threshold, behavior characteristic threshold and distance change threshold; compare the safety assessment value, behavior characteristic value and distance change value of the target vehicle with the micro-behavior threshold, and obtain the number of micro-behavior occurrences and duration under the corresponding micro-behavior threshold; The number of occurrences and duration of micro-behaviors are monitored, the behavioral tendency of the target vehicle is identified based on the cumulative effect of micro-behaviors, and the tendency analysis coefficient of the target vehicle is calculated.

8. The image recognition-based tracking and identification management system for overweight vehicles according to claim 7 is characterized in that: The safety assessment value is expressed as: SAV = a STV ×STV+a DTC ×(1-DTC)+α RAD ×RAD; Among them, SAV represents the safety assessment value, STV represents the speed fluctuation value, DTC represents the distance from the center of mass, RAD represents the relative deviation angle, α STV Represents the weight coefficient of speed fluctuation value, α DTC Represents the weight coefficient of the distance from the centroid, α RAD Represents the weight coefficient of the relative deflection angle.

9. The image recognition-based tracking and identification management system for overweight vehicles according to claim 7 is characterized in that: The tendency analysis coefficient is expressed as follows: the weight corresponding to the duration of the micro-behavior is obtained, and the tendency analysis coefficient is calculated according to the safety assessment value, behavior characteristic value, interval change value and number of occurrences of the current micro-behavior; Among them, PA represents the tendency analysis coefficient, SAV i represents the safety assessment value of the i-th micro-behavior, BV i represents the behavior characteristic value of the i-th micro-behavior, SV i represents the spacing change value of the i-th micro-behavior, n represents the number of micro-behaviors, i ranges from 1 to n, and w i represents the weight corresponding to the duration of the i-th micro-behavior, α SAV,i represents the weight coefficient of the safety assessment value of the i-th micro-behavior, α BV,i represents the weight coefficient of the behavior characteristic value of the i-th micro-behavior, α SV,i Represents the weight coefficient of the spacing change value of the i-th micro-behavior.

10. The image recognition-based tracking and identification management system for overweight vehicles according to claim 1 is characterized in that: The vehicle evaluation result is expressed by obtaining a probability distribution value of a trend analysis coefficient, using the probability distribution value of the trend analysis coefficient as the probability distribution of the mapping behavior, calculating a vehicle evaluation coefficient according to the probability distribution of the mapping behavior, the tendency analysis coefficient and the size analysis result, and outputting the vehicle evaluation coefficient as the vehicle evaluation result; Where VER represents the vehicle evaluation coefficient, P(γ) represents the probability distribution of the mapping behavior, PA represents the tendency analysis coefficient, DAR represents the index value of the size analysis result; β, χ, λ represent adjustment coefficients, and e represents the exponential constant.

Images