Accurate detection system for overloaded vehicles based on image analysis

Abstract

The present invention discloses an accurate overload vehicle detection system based on image analysis, which relates to the technical field of overload vehicle detection, and comprises a detection point initial setting unit, a layout scheme generating unit, a point cloud data extracting unit, a dimension analyzing unit, a vehicle volume calculating unit, a license plate recognizing unit and a result outputting unit. The present invention numbers all highways involved in a designated area through the detection point initial setting unit, increases the number of detection points corresponding to highways involved in areas with a large gross domestic product value, a large traffic volume and a high possibility of overload, and ensures that the initial establishment of detection points is more scientific and reasonable, and the layout scheme generating unit further optimizes the layout of detection points in the detection point initial setting unit, and considers multiple aspects such as detection coverage, the number of bridges included and detection point cost, to ensure that each detection point can play the best effect when it is put into use.

Description

Technical Field

[0001] The present invention relates to the technical field of overweight vehicle detection, and in particular to an accurate overweight vehicle detection system based on image analysis. Background Art

[0002] Overload transportation will cause the technical parameters of the vehicle to exceed the design standards, increase the risk of vehicle damage and mechanical accidents, and threaten the safety of drivers and pedestrians. By detecting overloaded vehicles, overload transportation behavior can be discovered and prevented in time, reducing the occurrence of road traffic safety accidents. The invention patent with application number 202210556786.X discloses "a vehicle overload detection method and terminal, and a vehicle overload detection system. The vehicle overload detection method is used to detect whether a vehicle entering a highway entrance is overlong, overhigh, overwide, and overweight; the method includes: obtaining a detection image of a target vehicle in an overload detection area at a highway intersection, and judging whether the target vehicle is overlong, overhigh, and overwide based on the detection image; if the target vehicle is not overlong, overhigh, and overwide, obtaining the weight information of the target vehicle; extracting the license plate information of the target vehicle from the detection image, determining the weight threshold of the target vehicle according to the license plate information, and judging whether the target vehicle is overweight based on the weight information and the weight threshold. The present invention can improve the efficiency of vehicle overload detection and reduce labor costs.";

[0003] The above-mentioned existing technology solves the problem of low efficiency in manual vehicle inspection. However, when the system is running, due to the lack of setting and optimization of the inspection point layout, the number of inspection points on each road may be unreasonable, and it cannot be ensured that each inspection point can achieve the best effect when put into use. At the same time, the system does not use point cloud data to analyze the size of the vehicle, resulting in a large error in the final result, and the system cannot calculate the actual volume of the current vehicle, making it impossible for the operator to take relevant measures to intervene in time. Summary of the invention

[0004] The purpose of the present invention is to provide an accurate detection system for overloaded vehicles based on image analysis to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an accurate overload vehicle detection system based on image analysis, comprising a vehicle volume calculation unit and a license plate recognition unit;

[0006] A detection point initialization unit, wherein the detection point initialization unit divides the designated area into a plurality of regions, counts the values ​​of the corresponding first-class indicators in each region, takes the numbered roads as nodes, determines the number of the first-class indicators and the total number of nodes, counts the values ​​of the second-class indicators corresponding to each numbered road, and uses a node importance analysis algorithm to calculate according to the weights and indicator values ​​corresponding to each first-class indicator and the second-class indicator to obtain the first importance value and the second importance value corresponding to the current node, and analyzes the total importance value corresponding to each node based on these values;

[0007] A layout scheme generating unit, wherein the layout scheme generating unit sets a mark bit for each node, determines a relevant target formula after setting the detection point according to the total importance value corresponding to each node, the number of bridges included, the cost of the detection point and the mark bit, adds constraints, continuously adjusts the number of detection points corresponding to each node through a layout optimization algorithm, calculates the target formula corresponding to each layout scheme, outputs an optimal layout scheme through a multi-objective optimization model, and sets detection points for the highway corresponding to each node according to the optimal layout scheme;

[0008] A point cloud data extraction unit, wherein the point cloud data extraction unit uses an IMU sensor to collect data related to the position and posture of the vehicle to be detected, uses a laser radar to scan the vehicle to be detected, calculates the difference between the horizontal angle of the earliest data point collected and the horizontal angle of the latest data point collected, uses the difference as the actual scanning angle of the current laser radar, analyzes the corresponding vehicle position, speed and posture according to the IMU data corresponding to each point cloud data, calculates these data through a vehicle position correction algorithm, obtains the actual data at each point cloud data collection moment, determines the actual three-dimensional coordinates of all point cloud data according to these data, and groups these data;

[0009] A dimension analysis unit, wherein the dimension analysis unit calculates the point cloud data using a dimension analysis algorithm to obtain the actual width and height corresponding to the vehicle to be tested. If it is determined that the vehicle to be tested enters the scanning area at the time of collection, the data collected at that time and thereafter are transmitted to the processor, and the actual length corresponding to the vehicle to be tested is calculated based on these data;

[0010] The result output unit uses a gravity sensor to collect weight data of the vehicle to be tested, transmits it to a storage device, queries the database according to the license plate information through a processor, determines the prescribed limits corresponding to the vehicle type, and then determines whether the height, width, length and weight of the current vehicle exceed the prescribed limits. If they do not exceed the prescribed limits, no operation is performed. Otherwise, it is prompted that the vehicle is over-limited and the current volume data of the vehicle is returned to the visual interface.

[0011] Preferably, the detection point initial setting unit includes a regional division module and an indicator weight calculation module. The regional division module divides the designated area into multiple areas, and numbers the roads involved in each area, and uses the gross domestic product, the total population and the total number of enterprises as three first-class indicators, and counts the values ​​of the corresponding first-class indicators in each area, and transmits these data to the database. The indicator weight calculation module uses the numbered roads as nodes, determines the number of first-class indicators and the total number of nodes, and stores the first-class indicator values ​​corresponding to each node in the form of a matrix, and calculates the proportion of each node in each first-class indicator in the sum of the corresponding indicator values. These data are used to analyze the entropy value corresponding to each first-class indicator, and the corresponding weight is calculated according to the entropy value of each first-class indicator through a weight analysis algorithm. The weight analysis algorithm is specifically as follows:

[0012] ;

[0013] In the formula, Indicates the weight corresponding to the indicator, Indicates the entropy value corresponding to the indicator, Indicates the parameters, Indicates the number of indicators.

[0014] Preferably, the detection point initial setting unit also includes a first importance analysis module, a second importance analysis module and a total importance determination module. The first importance analysis module uses the node number to determine the area to which the node belongs, queries the database according to the area name through the database query device, obtains the first-class indicator value corresponding to the area, uses the node importance analysis algorithm to calculate according to the weight corresponding to each first-class indicator and the indicator value of the area to which it belongs, and obtains the first importance value corresponding to the current node. After determining the first importance values ​​of all nodes, they are stored in the database. The second importance analysis module uses the road driving speed, road length, highway grade and number of bridges as four second-class indicators, and after counting the second-class indicator values ​​corresponding to each numbered highway, determines the weight corresponding to each indicator, and uses the node importance analysis algorithm to calculate the second importance value corresponding to each node. The total importance determination module analyzes the total importance corresponding to each node according to the first importance value and the second importance value, and sets the initial number of detection points corresponding to each node according to the total importance.

[0015] Preferably, the layout scheme generating unit comprises a mark bit setting module, a target building module, a constraint adding module and a scheme executing module. The mark bit setting module sets a mark bit for each node. If the mark bit is 0, it indicates that the initial number of detection points corresponding to the node is zero. If the mark bit is 1, it indicates that the initial number of detection points corresponding to the node is not zero. The target building module determines the target formula for maximizing the sum of the total importance of all roads after setting the detection point according to the total importance value corresponding to each node and the mark bit, extracts the number of bridges contained in each node and the mark bit, thereby obtaining the target formula for maximizing the sum of the number of bridges on all roads after setting the detection point, obtains the sum of the cost of the detection points in each node, and calculates the target formula for minimizing the total cost of all detection points according to the mark bit of the node. The constraint adding module The constraint conditions are added to the block, where the constraint conditions include that the total cost is not higher than the preset value and the coverage of the detection points is not lower than the preset range. After determining the total number of nodes and the number of model iterations, the multi-objective optimization model is used to analyze according to the target formula and the constraint conditions. The number of detection points corresponding to each node is continuously adjusted through the layout optimization algorithm, and the target formula corresponding to each layout scheme is calculated. When the number of executions of the layout optimization algorithm is higher than the number of model iterations, the optimal layout scheme is output through the multi-objective optimization model. The scheme execution module sets detection points for the highway corresponding to each node according to the optimal layout scheme. Multiple laser radars, IMU sensors, gravity sensors and cameras are arranged at each detection point. The specific model of the laser radar is RS_LiDAR_16 laser radar. The layout optimization algorithm is specifically as follows:

[0016] ;

[0017] In the formula, represents the difference between the optimal layout solution and the current iteration, Indicates the optimization of the layout of highway overload detection points The optimal solution that satisfies the objectives and constraints is obtained in the iteration. express The optimal layout of the detection points is express The optimal layout of the detection points is Indicates the current iteration number, and represents the coefficient vector.

[0018] Preferably, the point cloud data extraction unit includes a data acquisition module and a scanning angle analysis module. The data acquisition module uses an IMU sensor to collect data related to the position and posture of the vehicle to be detected, transmits the collected data to a storage device, uses a laser radar to scan the vehicle to be detected, analyzes the collected point cloud data through a coordinate conversion formula, converts the polar coordinates of each point cloud data into the corresponding three-dimensional coordinates in the vehicle coordinate system, and transmits them to the storage device. After the scanning angle analysis module determines the horizontal angle corresponding to each point cloud data, it calculates the difference between the horizontal angle of the earliest data point in the collection time and the horizontal angle of the latest data point in the collection time, and uses the difference as the actual scanning angle of the current laser radar, and extracts the IMU data of the previous moment and the next moment closest to the collection moment corresponding to each point cloud data.

[0019] Preferably, the point cloud data extraction unit also includes a coordinate correction module and a data grouping module. The coordinate correction module analyzes the corresponding vehicle position, speed and posture according to the IMU data corresponding to each point cloud data, calculates these data through a vehicle position correction algorithm, and obtains the actual vehicle position, speed and posture related data at each point cloud data collection moment, and determines the actual three-dimensional coordinates of all point cloud data based on these data. The data grouping module groups the point cloud data collected by the laser radars with relative installation positions, sets the number of data corresponding to each group, and thus obtains multiple groups of point cloud data.

[0020] Preferably, the size analysis unit includes a point cloud data analysis module and a vehicle length determination module. The point cloud data analysis module selects a whole set of point cloud data corresponding to different laser radars for analysis to obtain the actual width and height corresponding to the vehicle to be tested. The vehicle length determination module determines the average number of point cloud data collected at any time when the vehicle to be tested has not entered the scanning range, and counts the number of point cloud data collected by the laser radar at each moment. If there is a moment when the number of point cloud data collected is higher than the average value and the difference is higher than or equal to a threshold, it is determined that the vehicle to be tested enters the scanning area at this moment, and the data collected at this moment and thereafter are transmitted to the processor until the difference between the number of point cloud data and the average value is lower than the threshold. The transmission is stopped. After receiving the point cloud data, the processor calculates the actual length corresponding to the vehicle to be tested based on these data.

[0021] Preferably, the vehicle volume calculation unit includes a data slicing module and a reference point generation module. After the data slicing module extracts the actual three-dimensional coordinates of all point cloud data, it sets a slicing distance and slices the point cloud data longitudinally according to the slicing distance to obtain multiple groups of point cloud data. After the reference point generation module selects any point in each group of point cloud data as the first reference point, it sets a threshold. If the distance between the same group of point cloud data and the reference point is less than or equal to the threshold, the point cloud data is stored in a point cloud set, and any point in the set except the first reference point is selected as the second reference point. After determining the center point of a circle that passes through the first reference point and the second reference point at the same time, the two center points generated are used as the third reference point and the fourth reference point, respectively.

[0022] Preferably, the vehicle volume calculation unit also includes a boundary point judgment module and a volume analysis module. The boundary point judgment module deletes the first reference point and the second reference point in the set, calculates the distances between all points and the third reference point and the fourth reference point, and if there is a point whose distances to the third reference point and the fourth reference point are both greater than half of a threshold, the point is determined to be a boundary point, otherwise the point is determined not to be a boundary point. The volume analysis module connects all boundary points to obtain contour information corresponding to the point cloud slice, calculates the area corresponding to each point cloud slice using Green's formula, analyzes the volume corresponding to each slice according to the slice surface and the slice distance, and accumulates the volumes of these slices to obtain the total volume of the vehicle to be tested.

[0023] Preferably, the license plate recognition unit includes an information positioning module and a license plate analysis module. The information positioning module uses a camera to shoot the vehicle to be tested, and transmits the captured image information to the image processing software. After the image information is preprocessed by the image processing software, the image information is screened by using Blob analysis technology to determine the specific location of the license plate information. After the license plate analysis module performs character segmentation on the image information, a license plate recognition model is constructed using convolutional neural network technology, and the segmented image information is transmitted to the model. After analyzing the image information using the license plate recognition model, the license plate information is obtained and transmitted to the processor. The image processing software is specifically Halcon image processing software.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. The present invention numbers all highways involved in the designated area through the detection point initial setting unit, and analyzes the numbered highways as nodes, clarifies the first importance value and the second importance value corresponding to each node, and increases the number of detection points corresponding to highways involved in areas with a large gross domestic product, a large traffic volume, and a high possibility of overload, so as to ensure that the initial establishment of detection points is more scientific and reasonable, and the layout plan generation unit further optimizes the detection point layout in the detection point initial setting unit, taking into account multiple aspects such as detection coverage, the number of bridges included, and detection point cost, so as to ensure that each detection point can play the best effect when it is put into use;

[0026] 2. The present invention corrects the three-dimensional coordinates of the point cloud data collected by the laser radar through a point cloud data extraction unit to prevent erroneous data from affecting the accuracy of subsequent analysis of the size and volume of the vehicle to be tested. The system uses a vehicle volume calculation unit to obtain the total volume of the vehicle to be tested. If the size or weight of the vehicle to be tested exceeds the specified limit, the operator can obtain the current volume data of the vehicle in the first place, so as to take relevant measures to intervene in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Provides a schematic diagram of the overall system flow for an embodiment of the present invention;

[0028] Figure 2 An internal module block diagram of a detection point initial setting unit provided in an embodiment of the present invention;

[0029] Figure 3 An internal module block diagram of a layout scheme generating unit provided in an embodiment of the present invention;

[0030] Figure 4 An internal module block diagram of a point cloud data extraction unit provided in an embodiment of the present invention;

[0031] Figure 5 This is a block diagram of the internal modules of a vehicle volume calculation unit provided in an embodiment of the present invention.

[0032] In the figure: 1. Detection point initial setting unit; 101. Area division module; 102. Index weight calculation module; 103. First importance analysis module; 104. Second importance analysis module; 105. Total importance determination module; 2. Layout plan generation unit; 201. Marker position setting module; 202. Target construction module; 203. Constraint addition module; 204. Plan execution module; 3. Point cloud data extraction unit; 301. Data acquisition module; 302. Scanning angle analysis module; 303. Coordinate correction module; 304. Data grouping module; 4. Dimension analysis unit; 401. Point cloud data analysis module; 402. Vehicle length determination module; 5. Vehicle volume calculation unit; 501. Data slicing module; 502. Reference point generation module; 503. Boundary point judgment module; 504. Volume analysis module; 6. License plate recognition unit; 601. Information positioning module; 602. License plate analysis module; 7. Result output unit. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] See also Figure 1-5 ,The present invention provides a technical solution: an accurate detection system for over-limit vehicles based on image analysis, comprising a vehicle volume calculation unit 5 and a license plate recognition unit 6;

[0035] The detection point initial setting unit 1 divides the designated area into multiple areas, counts the values ​​of the corresponding first-class indicators in each area, takes the numbered roads as nodes, determines the number of the first-class indicators and the total number of nodes, counts the values ​​of the second-class indicators corresponding to each numbered road, and uses the node importance analysis algorithm to calculate the weights and indicator values ​​corresponding to each first-class indicator and the second-class indicator to obtain the first importance value and the second importance value corresponding to the current node, and analyzes the total importance value corresponding to each node based on these values;

[0036] Layout scheme generating unit 2, layout scheme generating unit 2 sets a mark bit for each node, determines the relevant target formula after setting the detection point according to the total importance value corresponding to each node, the number of bridges included, the cost of the detection point and the mark bit, adds constraints, continuously adjusts the number of detection points corresponding to each node through the layout optimization algorithm, and calculates the target formula corresponding to each layout scheme, outputs the optimal layout scheme through the multi-objective optimization model, and sets the detection point for the highway corresponding to each node according to the optimal layout scheme;

[0037] Point cloud data extraction unit 3, point cloud data extraction unit 3 uses IMU sensor to collect position and posture related data of the vehicle to be detected, uses laser radar to scan the vehicle to be detected, calculates the difference between the horizontal angle of the earliest data point collected and the horizontal angle of the latest data point collected, uses the difference as the actual scanning angle of the current laser radar, analyzes the corresponding vehicle position, speed and posture according to the IMU data corresponding to each point cloud data, calculates these data through the vehicle position correction algorithm, obtains the actual data at each point cloud data collection time, determines the actual three-dimensional coordinates of all point cloud data according to these data, and groups these data;

[0038] The dimension analysis unit 4 calculates the point cloud data to obtain the actual width and height of the vehicle to be tested. If it is determined that the vehicle to be tested enters the scanning area at the time of collection, the data collected at that time and thereafter are transmitted to the processor, and the actual length of the vehicle to be tested is calculated based on these data;

[0039] The result output unit 7 uses a gravity sensor to collect the weight data of the vehicle to be tested, and then transmits it to a storage device. The processor queries the database according to the license plate information, determines the prescribed limits corresponding to the vehicle type, and then determines whether the height, width, length and weight of the current vehicle exceed the prescribed limits. If they do not exceed the prescribed limits, no operation is performed. Otherwise, it is prompted that the vehicle is over-limited and the current volume data of the vehicle is returned to the visual interface.

[0040] The detection point initial setting unit 1 includes a regional division module 101 and an indicator weight calculation module 102. The regional division module 101 divides the designated area into multiple areas, and numbers the roads involved in each area, and uses the gross domestic product, the total population and the total number of enterprises as three first-class indicators, and counts the values ​​of the corresponding first-class indicators in each area, and transmits these data to the database. The indicator weight calculation module 102 uses the numbered roads as nodes, determines the number of first-class indicators and the total number of nodes, and stores the first-class indicator values ​​corresponding to each node in the form of a matrix, and calculates the proportion of each node in each first-class indicator in the sum of the corresponding indicator values. These data are used to analyze the entropy values ​​corresponding to each first-class indicator, and the corresponding weight is calculated according to the entropy value of each first-class indicator through a weight analysis algorithm. The weight analysis algorithm is specifically as follows:

[0041] ;

[0042] In the formula, Indicates the weight corresponding to the indicator, Indicates the entropy value corresponding to the indicator, Indicates the parameters, Indicates the number of indicators;

[0043] The detection point initial setting unit 1 also includes a first importance analysis module 103, a second importance analysis module 104 and a total importance determination module 105. The first importance analysis module 103 uses the node number to determine the area to which the node belongs, and queries the database according to the area name through the database query device to obtain the first-class indicator value corresponding to the area, and uses the node importance analysis algorithm to calculate according to the weight corresponding to each first-class indicator and the indicator value of the area to which it belongs, to obtain the first importance value corresponding to the current node, and after determining the first importance values ​​of all nodes, store them in the database. The second importance analysis module 104 uses the road driving speed, road length, highway grade and number of bridges as four second-class indicators, and after counting the second-class indicator values ​​corresponding to each numbered highway, determines the weight corresponding to each indicator, and uses the node importance analysis algorithm to calculate the second importance value corresponding to each node. The total importance determination module 105 analyzes the total importance corresponding to each node according to the first importance value and the second importance value, and sets the initial number of detection points corresponding to each node according to the total importance. The node importance analysis algorithm is specifically as follows:

[0044] ;

[0045] In the formula, Represents a regional node The importance of Indicates The weight of the evaluation index, Indicates The weight of the evaluation index, Indicates The weight of the evaluation index, Representative The value of the regional indicator, Representative The average value of the regional indicators, Representative The value of the regional indicator, Representative The average value of the regional indicators, Representative The value of the regional indicator, Representative The average value of the regional indicators;

[0046] The layout scheme generating unit 2 comprises a mark bit setting module 201, a target building module 202, a constraint adding module 203 and a scheme executing module 204. The mark bit setting module 201 sets a mark bit for each node. If the mark bit is 0, it indicates that the initial number of detection points corresponding to the node is zero. If the mark bit is 1, it indicates that the initial number of detection points corresponding to the node is not zero. The target building module 202 determines the target formula for maximizing the sum of the total importance of all roads after setting the detection point according to the total importance value corresponding to each node and the mark bit, extracts the number of bridges and the mark bit contained in each node, thereby obtaining the target formula for maximizing the sum of the number of bridges on all roads after setting the detection point, obtains the sum of the cost of the detection points in each node, and calculates the target formula for minimizing the total cost of all detection points according to the mark bit of the node. The constraint The adding module 203 adds constraints, wherein the constraints include that the total cost is not higher than the preset value and the coverage of the detection points is not lower than the preset range. After determining the total number of nodes and the number of model iterations, the multi-objective optimization model is used to analyze according to the target formula and the constraints. The number of detection points corresponding to each node is continuously adjusted through the layout optimization algorithm, and the target formula corresponding to each layout scheme is calculated. When the number of executions of the layout optimization algorithm is higher than the number of model iterations, the optimal layout scheme is output through the multi-objective optimization model. The scheme execution module 204 sets detection points for the highway corresponding to each node according to the optimal layout scheme. Multiple laser radars, IMU sensors, gravity sensors and cameras are arranged at each detection point. The specific model of the laser radar is RS_LiDAR_16 laser radar. The layout optimization algorithm is specifically as follows:

[0047] ;

[0048] In the formula, represents the difference between the optimal layout solution and the current iteration, Indicates the optimization of the layout of highway overload detection points The optimal solution that satisfies the objectives and constraints is obtained in the iteration. express The optimal layout of the detection points is express The optimal layout of the detection points is Indicates the current iteration number, and represents the coefficient vector;

[0049] The point cloud data extraction unit 3 includes a data acquisition module 301 and a scanning angle analysis module 302. The data acquisition module 301 uses an IMU sensor to collect data related to the position and posture of the vehicle to be detected, transmits the collected data to a storage device, uses a laser radar to scan the vehicle to be detected, analyzes the collected point cloud data through a coordinate conversion formula, converts the polar coordinates of each point cloud data into the corresponding three-dimensional coordinates in the vehicle coordinate system, and transmits it to a storage device. After the scanning angle analysis module 302 determines the horizontal angle corresponding to each point cloud data, it calculates the difference between the horizontal angle of the earliest data point collected and the horizontal angle of the latest data point collected, and uses the difference as the actual scanning angle of the current laser radar. The IMU data of the previous moment and the next moment closest to the collection moment corresponding to each point cloud data are extracted;

[0050] The point cloud data extraction unit 3 also includes a coordinate correction module 303 and a data grouping module 304. The coordinate correction module 303 analyzes the corresponding vehicle position, speed and posture according to the IMU data corresponding to each point cloud data, calculates these data through the vehicle position correction algorithm, and obtains the actual vehicle position, speed and posture related data at each point cloud data collection moment. The actual three-dimensional coordinates of all point cloud data are determined based on these data. The data grouping module 304 groups the point cloud data collected by the laser radars with relative installation positions, sets the number of data corresponding to each group, and thus obtains multiple groups of point cloud data. The vehicle position correction algorithm is specifically as follows:

[0051] ;

[0052] In the formula, Indicates the vehicle position corresponding to the data point, Indicates the vehicle position corresponding to the IMU data at the previous moment, Indicates the vehicle position corresponding to the IMU data at the next moment, Indicates the vehicle speed corresponding to the data point, Indicates the vehicle speed corresponding to the IMU data at the previous moment, Indicates the vehicle speed corresponding to the IMU data at the next moment, Indicates the vehicle posture corresponding to the data point, Indicates the vehicle posture corresponding to the IMU data at the previous moment, Indicates the vehicle posture corresponding to the IMU data at the next moment, represents the interpolation factor;

[0053] The size analysis unit 4 includes a point cloud data analysis module 401 and a vehicle length determination module 402. The point cloud data analysis module 401 selects a whole set of point cloud data corresponding to different laser radars for analysis to obtain the actual width and height corresponding to the vehicle to be tested. The vehicle length determination module 402 determines the average number of point cloud data collected at any time when the vehicle to be tested does not enter the scanning range, and counts the number of point cloud data collected by the laser radar at each moment. If there is a moment when the number of point cloud data collected is higher than the average value and the difference is higher than or equal to the threshold, it is determined that the vehicle to be tested enters the scanning area at this moment, and the data collected at this moment and thereafter are transmitted to the processor until the difference between the number of point cloud data and the average value is lower than the threshold. The transmission is stopped. After the processor receives the point cloud data, it calculates the actual length corresponding to the vehicle to be tested based on these data;

[0054] The vehicle volume calculation unit 5 includes a data slicing module 501 and a reference point generation module 502. After the data slicing module 501 extracts the actual three-dimensional coordinates of all point cloud data, it sets a slicing distance, and slices the point cloud data in the longitudinal direction according to the slicing distance to obtain multiple groups of point cloud data. After the reference point generation module 502 selects any point in each group of point cloud data as the first reference point, it sets a threshold. If the distance between the same group of point cloud data and the reference point is less than or equal to the threshold, the point cloud data is stored in a point cloud set, and any point in the set except the first reference point is selected as the second reference point. After determining the center point of a circle passing through the first reference point and the second reference point at the same time, the two center points generated are respectively used as the third reference point and the fourth reference point.

[0055] The vehicle volume calculation unit 5 also includes a boundary point judgment module 503 and a volume analysis module 504. The boundary point judgment module 503 deletes the first reference point and the second reference point in the set, calculates the distances between all points and the third reference point and the fourth reference point, and if there is a point whose distances to the third reference point and the fourth reference point are both greater than half of the threshold, the point is determined to be a boundary point, otherwise the point is determined not to be a boundary point. The volume analysis module 504 connects all boundary points to obtain the contour information corresponding to the point cloud slice, calculates the area corresponding to each point cloud slice using the Green's formula, analyzes the volume corresponding to each slice according to the slice surface and the slice distance, and accumulates the volumes of these slices to obtain the total volume of the vehicle to be tested;

[0056] The license plate recognition unit 6 includes an information positioning module 601 and a license plate analysis module 602. The information positioning module 601 uses a camera to shoot the vehicle to be tested, and transmits the captured image information to the image processing software. After the image information is preprocessed by the image processing software, the image information is screened using the Blob analysis technology to determine the specific location of the license plate information. The license plate analysis module 602 performs character segmentation on the image information, and then uses the convolutional neural network technology to construct a license plate recognition model, and transmits the segmented image information to the model. After analyzing the image information using the license plate recognition model, the license plate information is obtained and transmitted to the processor. The image processing software is specifically the Halcon image processing software.

[0057] Working principle: The present invention obtains the first-class indicator value corresponding to each numbered highway through the regional division module 101 in the detection point initial setting unit 1, and calculates the weight corresponding to each first-class indicator through the indicator weight calculation module 102. Then, the first importance analysis module 103 calculates the first importance value corresponding to the current node according to the weight corresponding to each first-class indicator and the indicator value of the corresponding area. The second importance analysis module 104 uses the node importance analysis algorithm to calculate the second importance value corresponding to each node according to the weight corresponding to each second-class indicator and the indicator value. The total importance determination module 105 analyzes the total importance corresponding to each node, uses the mark bit setting module 201 in the layout plan generation unit 2 to set the mark bit for each node, determines the target formula corresponding to all nodes through the target construction module 202, adds constraints by the constraint adding module 203, and outputs the optimal layout plan according to the target formula and the constraints. The plan execution module 204 sets a detection point on the road corresponding to each node according to the optimal layout plan. The data acquisition module 301 in the point cloud data extraction unit 3 converts the polar coordinates of the collected point cloud data into the corresponding three-dimensional coordinates in the vehicle coordinate system. The scanning angle analysis module 303 02 determines the actual scanning angle of the current laser radar, uses the coordinate correction module 303 to analyze the actual three-dimensional coordinates of all point cloud data, the data grouping module 304 groups the data collected by different laser radars, calculates the actual width and height of the vehicle to be tested by the point cloud data analysis module 401 in the size analysis unit 4, and analyzes the actual length of the vehicle to be tested by the vehicle length determination module 402. The point cloud data is sliced ​​by the data slicing module 501 in the vehicle volume calculation unit 5, and the reference point generation module 502 determines the first reference point, the second reference point, the third reference point and the fourth reference point. The specific position of the vehicle is determined by the boundary point judgment module 503, which analyzes all the boundary points based on these reference points. The volume analysis module 504 accumulates the volumes of these slices to obtain the total volume of the vehicle to be tested. The information positioning module 601 in the license plate recognition unit 6 determines the specific position of the license plate information. The license plate analysis module 602 transmits the obtained license plate information to the processor. The result output unit 7 determines whether the height, width, length and weight of the current vehicle exceed the specified limits. If they do not exceed the specified limits, no operation is performed. Otherwise, it is prompted that the vehicle is over-limited and the current volume data of the vehicle is returned to the visual interface.

[0058] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0059] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An accurate overload vehicle detection system based on image analysis, comprising a vehicle volume calculation unit (5) and a license plate recognition unit (6), characterized in that: A detection point initialization unit (1), the detection point initialization unit (1) comprising a region division module (101), an indicator weight calculation module (102), a first importance analysis module (103), a second importance analysis module (104) and a total importance determination module (105), the region division module (101) divides a designated area into a plurality of regions, and numbers the roads involved in each region, takes the gross domestic product, the total population and the total number of enterprises as three first-class indicators, counts the values ​​of the first-class indicators corresponding to each region, and transmits these data to a database, the indicator weight calculation module (102) takes the numbered roads as nodes, determines the number of first-class indicators and the total number of nodes, stores the first-class indicator values ​​corresponding to each node in the form of a matrix, and calculates the proportion of each node in each first-class indicator in the total value of the corresponding indicator, uses these data to analyze the entropy value corresponding to each first-class indicator, and calculates the entropy value of each first-class indicator according to the entropy value of each first-class indicator through a weight analysis algorithm. The first importance analysis module (103) determines the region to which the node belongs by using the node number, queries the database according to the region name through the database query device, obtains the first-class indicator value corresponding to the region, calculates according to the weight corresponding to each first-class indicator and the indicator value of the region by using the node importance analysis algorithm, obtains the first importance value corresponding to the current node, and after determining the first importance values ​​of all nodes, stores them in the database. The second importance analysis module (104) uses the road speed, road length, highway grade and number of bridges as four second-class indicators, counts the second-class indicator values ​​corresponding to each numbered highway, determines the weight corresponding to each indicator, and calculates the second importance value corresponding to each node by using the node importance analysis algorithm. The total importance determination module (105) analyzes the total importance corresponding to each node according to the first importance value and the second importance value, and sets the initial number of detection points corresponding to each node according to the total importance. A layout scheme generating unit (2), the layout scheme generating unit (2) comprising a mark bit setting module (201), a target construction module (202), a constraint adding module (203) and a scheme executing module (204), the mark bit setting module (201) setting a mark bit for each node, if the mark bit is 0, it means that the initial number of detection points corresponding to the node is zero, if the mark bit is 1, it means that the initial number of detection points corresponding to the node is not zero, the target construction module (202) determines the target formula for maximizing the sum of the total importance of all roads after setting the detection point according to the total importance value corresponding to each node and the mark bit, extracts the number of bridges contained in each node and the mark bit, thereby obtaining the target formula for maximizing the sum of the number of bridges on all roads after setting the detection point, and obtains the total cost of the detection points in each node. , the target formula for minimizing the total cost of all detection points is calculated according to the mark position of the node, the constraint adding module (203) adds constraint conditions, wherein the constraint conditions include that the total cost is not higher than a preset value and the detection point coverage is not lower than a preset range, after determining the total number of nodes and the number of model iterations, the multi-objective optimization model is used to analyze according to the target formula and the constraint conditions, the number of detection points corresponding to each node is continuously adjusted through the layout optimization algorithm, and the target formula corresponding to each layout scheme is calculated, when the number of executions of the layout optimization algorithm is higher than the number of model iterations, the optimal layout scheme is output through the multi-objective optimization model, and the scheme execution module (204) sets detection points on the highway corresponding to each node according to the optimal layout scheme, and multiple laser radars, IMU sensors, gravity sensors and cameras are arranged at each detection point; A point cloud data extraction unit (3), wherein the point cloud data extraction unit (3) uses an IMU sensor to collect data related to the position and posture of the vehicle to be detected, uses a laser radar to scan the vehicle to be detected, calculates the difference between the horizontal angle of the earliest data point collected and the horizontal angle of the latest data point collected, uses the difference as the actual scanning angle of the current laser radar, analyzes the corresponding vehicle position, speed and posture based on the IMU data corresponding to each point cloud data, calculates these data through a vehicle position correction algorithm, obtains actual data at each point cloud data collection time, determines the actual three-dimensional coordinates of all point cloud data based on these data, and groups these data; A dimension analysis unit (4), wherein the dimension analysis unit (4) calculates the point cloud data to obtain the actual width and height corresponding to the vehicle to be tested. If it is determined that the vehicle to be tested enters the scanning area at the time of collection, the data collected at that time and thereafter are transmitted to the processor, and the actual length corresponding to the vehicle to be tested is calculated based on the data; A result output unit (7), wherein the result output unit (7) collects weight data of the vehicle to be tested using a gravity sensor, transmits the data to a storage device, queries a database based on the license plate information through a processor, determines the prescribed limits corresponding to the vehicle type, and determines whether the height, width, length and weight of the current vehicle exceed the prescribed limits. If the prescribed limits are not exceeded, no operation is performed; otherwise, a prompt is given that the vehicle is over-limited, and the current volume data of the vehicle is returned to the visual interface.

2. The overload vehicle precision detection system based on image analysis according to claim 1 is characterized by: The point cloud data extraction unit (3) comprises a data acquisition module (301) and a scanning angle analysis module (302). The data acquisition module (301) uses an IMU sensor to collect data related to the position and posture of the vehicle to be detected, transmits the collected data to a storage device, uses a laser radar to scan the vehicle to be detected, analyzes the collected point cloud data through a coordinate conversion formula, converts the polar coordinates of each point cloud data into corresponding three-dimensional coordinates in the vehicle coordinate system, and transmits them to the storage device. After the scanning angle analysis module (302) determines the horizontal angle corresponding to each point cloud data, it calculates the difference between the horizontal angle of the earliest data point collected and the horizontal angle of the latest data point collected, uses the difference as the actual scanning angle of the current laser radar, and extracts the IMU data of the previous moment and the next moment closest to the collection moment corresponding to each point cloud data.

3. The overload vehicle precision detection system based on image analysis according to claim 2 is characterized by: The point cloud data extraction unit (3) further comprises a coordinate correction module (303) and a data grouping module (304). The coordinate correction module (303) analyzes the corresponding vehicle position, speed and posture according to the IMU data corresponding to each point cloud data, calculates these data by a vehicle position correction algorithm, obtains the actual vehicle position, speed and posture related data at each point cloud data collection moment, and determines the actual three-dimensional coordinates of all point cloud data according to these data. The data grouping module (304) groups the point cloud data collected by the laser radars at the relative installation positions, sets the number of data corresponding to each group, and thereby obtains multiple groups of point cloud data.

4. The overload vehicle precision detection system based on image analysis according to claim 1 is characterized by: The dimension analysis unit (4) comprises a point cloud data analysis module (401) and a vehicle length determination module (402). The point cloud data analysis module (401) selects a whole set of point cloud data corresponding to different laser radars for analysis to obtain the actual width and height corresponding to the vehicle to be tested. The vehicle length determination module (402) determines the average value of the number of point cloud data collected at any time when the vehicle to be tested does not enter the scanning range, and counts the number of point cloud data collected by the laser radar at each time. If there is a time when the number of point cloud data collected is higher than the average value and the difference is higher than or equal to a threshold, it is determined that the vehicle to be tested enters the scanning area at this time, and the data collected at this time and thereafter are transmitted to the processor until the difference between the number of point cloud data and the average value is lower than the threshold. The transmission is stopped. After the processor receives the point cloud data, it calculates the actual length corresponding to the vehicle to be tested based on the data.

5. The overload vehicle precision detection system based on image analysis according to claim 1 is characterized by: The vehicle volume calculation unit (5) comprises a data slicing module (501) and a reference point generation module (502). The data slicing module (501) extracts the actual three-dimensional coordinates of all point cloud data, sets a slicing distance, and slices the point cloud data in the longitudinal direction according to the slicing distance to obtain multiple groups of point cloud data. The reference point generation module (502) selects any point in each group of point cloud data as a first reference point, sets a threshold, and if the distance between the same group of point cloud data and the reference point is less than or equal to the threshold, stores the point cloud data in a point cloud set, selects any point in the set except the first reference point as a second reference point, determines the center point of a circle passing through the first reference point and the second reference point at the same time, and uses the two generated center points as a third reference point and a fourth reference point, respectively.

6. The overload vehicle precision detection system based on image analysis according to claim 5 is characterized by: The vehicle volume calculation unit (5) also includes a boundary point judgment module (503) and a volume analysis module (504). The boundary point judgment module (503) deletes the first reference point and the second reference point in the set, calculates the distances between all points and the third reference point and the fourth reference point, and if there is a point whose distances to the third reference point and the fourth reference point are both greater than half of a threshold, the point is determined to be a boundary point, otherwise the point is determined not to be a boundary point. The volume analysis module (504) connects all boundary points to obtain contour information corresponding to the point cloud slice, calculates the area corresponding to each point cloud slice using the Green's formula, analyzes the volume corresponding to each slice according to the slice surface and the slice distance, and accumulates the volumes of these slices to obtain the total volume of the vehicle to be tested.

7. The overload vehicle precision detection system based on image analysis according to claim 1 is characterized by: The license plate recognition unit (6) comprises an information positioning module (601) and a license plate analysis module (602). The information positioning module (601) uses a camera to shoot a vehicle to be tested, transmits the captured image information to the image processing software, performs preprocessing operations on the image information using the image processing software, and then uses the Blob analysis technology to filter the image information to determine the specific location of the license plate information. The license plate analysis module (602) performs character segmentation on the image information and then uses the convolutional neural network technology to construct a license plate recognition model, transmits the segmented image information to the model, and uses the license plate recognition model to analyze the image information to obtain the license plate information, which is then transmitted to the processor.

Images