Over-limit truck early warning method and system based on positioning data

Abstract

The invention provides an over-limit truck early warning method based on positioning data, and the method comprises the steps: (1) collecting existing truck positioning and speed data, truck overweight detection data and truck driving attribute data, adding the data into a database, and continuously adding real-time data; (2) performing linear interpolation completion, road section matching and abnormal data elimination on the data, forming standardized speed sequence data from speed data according to road section speed limitation, and dividing the standardized speed sequence data into fixed-length standardized speed sequences according to a specified data length; (3) taking the fixed-length standardized speed sequence data and the truck driving attribute data as input, and taking truck overweight detection data as labels of the input data to train a support vector machine model and a Logistic regression model respectively, and taking the models as a first model and a second model respectively; and (4) determining an early warning period length T, inputting the acquired fixed-length standardized speed sequence data and the truck driving attribute data into the first model and the second model, and periodically performing early warning according to the early warning period length.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic management and intelligent monitoring, and in particular to an overload truck early warning method and system based on positioning data. Background Art

[0002] With the rapid development of the logistics industry, the volume of truck transportation has increased dramatically, and the problem of overweight and overloaded transportation has become increasingly serious. Overweight trucks not only cause serious damage to road infrastructure, but also cause traffic accidents and threaten public safety. At present, the traditional overweight truck detection method mainly relies on fixed weighing equipment and manual inspection, which has problems such as low detection efficiency, insufficient monitoring coverage, and high equipment installation costs. Especially in complex road conditions and highway scenarios, traditional methods are difficult to detect overweight behavior in real time and effectively.

[0003] In recent years, intelligent early warning technology based on vehicle trajectory and speed data has gradually attracted attention. By analyzing the satellite positioning data of trucks and analyzing and learning the samples of vehicle overweight violation data, it is possible to realize dynamic monitoring and analysis of vehicle positioning, and identify and warn overweight vehicles through the characteristics of positioning data. However, existing technologies mostly use simple rule judgments or single machine learning models, fail to fully utilize the advantages of multiple algorithms, and have some limitations on detection accuracy and real-time performance. Therefore, there is an urgent need for a data analysis method that combines multiple models to improve the accuracy and timeliness of overweight truck detection. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide an overweight truck warning method and system based on positioning data, which realizes real-time analysis of truck trajectory and speed data and overweight warning by constructing a database module, a data processing module, a model module and a feedback module. The system can dynamically select the optimal model for overweight detection to avoid overfitting problems, and continuously optimize the accuracy and adaptability of the warning model by continuously updating and learning new data, thereby improving the detection efficiency of overweight truck behavior and the intelligent level of traffic management.

[0005] Technical solution, in order to solve the above technical problems, the present invention proposes an overload truck warning method based on positioning data, the method comprising the following steps:

[0006] (1) Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data and add them to the database, and continuously add real-time data;

[0007] (2) Perform linear interpolation to fill the data, match the road sections, and remove abnormal data. The speed data is converted into standardized speed sequence data according to the road section speed limit, and divided into fixed-length standardized speed sequences according to the specified data length;

[0008] (3) Using the fixed-length standardized speed sequence data and truck driving attribute data as input, and the truck overweight detection data as the label of the input data, the support vector machine model and the logistic regression model are trained respectively, and used as the first model and the second model respectively;

[0009] (4) Determine the length of the warning cycle T, input the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issue warnings periodically according to the length of the warning cycle.

[0010] Furthermore, the specific method of step (1) is as follows:

[0011] Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time;

[0012] Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

[0013] Furthermore, the specific method of step (2) is as follows:

[0014] The method of linear interpolation filling of data is as follows:

[0015] The truck positioning and speed data are linearly interpolated to complete. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i ,Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 ,Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1):

[0016]

[0017] To match the data to a road segment, we add a road segment label to each location point in the truck location and speed data through the following four steps. The specific method is as follows:

[0018] 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points;

[0019] 2) Divide the road sections in the trajectory map into four categories: expressways, expressways, trunk roads and other roads, extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope, and determine the truck speed limit V for each type of road section. limit , in kilometers per hour;

[0020] 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively;

[0021] Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lat α ) and (Lon β ,Lat β ) is calculated by formula (2):

[0022]

[0023] 4) Calculate the distance between the truck location point C and the line connecting A and B using equations (3), (4), and (5), and use this distance as the distance between the road section and the truck location point. Take the road section with the smallest distance to the truck location point as the road section label for the truck location point.

[0024] Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lon B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5):

[0025]

[0026] Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max , then the positioning point is regarded as abnormal data and removed;

[0027] The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows:

[0028] The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be v i and V limit , then the normalized velocity v of the positioning point i,std The calculation formula is:

[0029]

[0030] The standardized speed sequence data is converted into a data length T data Segmentation forms fixed-length standardized speed sequence data.

[0031] Furthermore, the specific method for establishing the first model and the second model in step (3) is as follows:

[0032] The first model is trained in the following way:

[0033] The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is an overweight high-frequency truck type are combined into a one-dimensional vector as input, the truck overweight detection data is used as the label of the input data, and whether it is overweight is used as the output of the model. It is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model.

[0034] The second model is trained in the following way:

[0035] The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7);

[0036]

[0037] Where P(y=1|(v mean,std ,x1,x2,x3,x4)) represents the probability of truck overload; v mean,stdrepresents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2,0 ,x 2,1 ,x 2,2 ,…,x 2,23 ) T , any element x in the vector 2,i are all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,…,23, x 2,i =1 means the truck is running during this time period, x 2,i =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. mean,std , regression coefficients of x1, x2, x3 and x4.

[0038] Furthermore, the method in step (4) is as follows:

[0039] (4.1) In the first warning cycle, the second model is selected as the warning model, and the truck driving attribute data is input to obtain the optimal AUC area cutoff point of the Logistic regression model as the optimal standardized speed mean threshold. It is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; the standardized speed sequence data and the truck driving attributes of this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results;

[0040] (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model are compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and an output warning of whether the truck is overweight is output; if the second model is selected, after the truck driving attribute data is input, the AUC area optimal cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed series data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; and the standardized speed series data and the truck driving attributes of this warning cycle are input into another model to output whether it is overweight, and the overweight warning accuracy of the other model is obtained by comparing the actual overweight detection results;

[0041] (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

[0042] In addition, the present invention proposes an overload truck warning system based on positioning data, which includes the following modules:

[0043] The database module collects existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data into the database, and continuously adds real-time data;

[0044] The data processing module performs linear interpolation, road section matching, and abnormal data elimination on the data, and forms standardized speed sequence data according to the speed limit of the road section, and divides the speed data into fixed-length standardized speed sequences according to the specified data length;

[0045] The model module uses the fixed-length standardized speed sequence data and the truck driving attribute data as inputs, and uses the truck overweight detection data as the label of the input data to train the support vector machine model and the logistic regression model, and uses them as the first model and the second model respectively;

[0046] The feedback module determines the warning cycle length T, inputs the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issues warnings periodically according to the warning cycle length.

[0047] Furthermore, the functions of the database module are as follows:

[0048] Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time;

[0049] Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

[0050] Furthermore, the functions of the data processing module are as follows:

[0051] Linear interpolation of data refers to linear interpolation of truck positioning and speed data. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i ,Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 ,Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1):

[0052]

[0053] Matching the data to a road segment involves adding a road segment label to each location point in the truck location and speed data through the following four steps:

[0054] 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points;

[0055] 2) Divide the road sections near the track in the track map into four categories: expressways, expressways, trunk roads and other roads, and extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope. For each category, determine its truck speed limit V limit , in kilometers per hour;

[0056] 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively;

[0057] Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lat α ) and (Lon β ,Lat β ) is calculated by formula (2):

[0058]

[0059] 4) Calculate the distance between the truck location point C and the line connecting A and B by using equations (3), (4), and (5) as the distance between the road section and the truck location point, and take the road section with the smallest distance to the truck location point as the road section label of the truck location point;

[0060] Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lon B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5):

[0061]

[0062] Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max , then the positioning point is regarded as abnormal data and removed;

[0063] The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows:

[0064] The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be vi and V limit , then the normalized velocity v of the positioning point i,std The calculation formula is:

[0065]

[0066] After standardization, the standardized speed sequence data is divided into two parts according to the specified data length T. data Segmentation forms fixed-length standardized speed sequence data.

[0067] Furthermore, the model module functions as follows:

[0068] The first model is trained in the following way:

[0069] The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are combined into a one-dimensional vector as input, and the overweight detection data of the truck is used as the label of the input data. Whether it is overweight is used as the output of the model, which is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model;

[0070] The second model is trained in the following way:

[0071] The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7);

[0072]

[0073] Where P(y=1|(v mean,std ,x1,x2,x3,x4)) represents the probability of truck overload; v mean,std represents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2,0 ,x 2,1 ,x 2,2 ,…,x 2,23 ) T , any element x in the vector 2,iare all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,…,23, x 2,i =1 means the truck is running during this time period, x 2,i =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. mean,std , regression coefficients of x1, x2, x3 and x4.

[0074] Furthermore, the specific functions of the feedback module are as follows:

[0075] (4.1) In the first warning cycle, the second model is selected as the warning model, and the truck driving attribute data is input to obtain the optimal AUC area cutoff point of the Logistic regression model as the optimal standardized speed mean threshold. It is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; the standardized speed sequence data and the truck driving attributes of this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results;

[0076] (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model are compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and an output warning of whether the truck is overweight is output; if the second model is selected, after the truck driving attribute data is input, the AUC area optimal cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed series data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; and the standardized speed series data and the truck driving attributes of this warning cycle are input into another model to output whether it is overweight, and the overweight warning accuracy of the other model is obtained by comparing the actual overweight detection results;

[0077] (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

[0078] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0079] (1) The present invention uses two models, support vector machine and logistic regression, for early warning in different data sample size scenarios. When there are fewer samples in the database module, the logistic model is used first to avoid the overfitting problem of the support vector machine model; as the sample size increases, it dynamically switches to the support vector machine model with higher accuracy. Through the collaborative optimization of the two models, the accuracy of overweight truck detection can be greatly improved.

[0080] (2) The system has the ability to dynamically switch models based on the real-time warning effect. The feedback module compares the warning accuracy of each model in the current time period and automatically selects the model with the higher accuracy as the main warning model for the next time period. This dynamic switching mechanism improves the system's adaptability to diverse road conditions and data changes, ensuring that the system always uses the optimal model for judgment. The system uses the location and speed data of the truck and uses a machine learning algorithm to determine whether the truck is overweight, aiming to improve road traffic safety and management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A structural diagram of an over-gauge truck early warning method and system based on positioning data of the present invention;

[0082] Figure 2 The present invention is a flowchart of an overload truck early warning method and system based on positioning data. DETAILED DESCRIPTION

[0083] The present invention is further described in detail below with reference to the accompanying drawings.

[0084] like Figure 2 As shown, the present invention proposes an overload truck warning method based on positioning data, which includes the following steps:

[0085] (1) Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data and add them to the database, and continuously add real-time data;

[0086] (2) Perform linear interpolation to fill the data, match the road sections, and remove abnormal data. The speed data is converted into standardized speed sequence data according to the road section speed limit, and divided into fixed-length standardized speed sequences according to the specified data length;

[0087] (3) Using the fixed-length standardized speed sequence data and truck driving attribute data as input, and the truck overweight detection data as the label of the input data, the support vector machine model and the logistic regression model are trained respectively, and used as the first model and the second model respectively;

[0088] (4) Determine the length of the warning cycle T, input the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issue warnings periodically according to the length of the warning cycle.

[0089] Furthermore, the specific method of step (1) is as follows:

[0090] Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time;

[0091] Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

[0092] Furthermore, the specific method of step (2) is as follows:

[0093] The method of linear interpolation filling of data is as follows:

[0094] The truck positioning and speed data are linearly interpolated to complete. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i ,Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 ,Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1):

[0095]

[0096] To match the data to a road segment, we add a road segment label to each location point in the truck location and speed data through the following four steps. The specific method is as follows:

[0097] 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points;

[0098] 2) Divide the road sections in the trajectory map into four categories: expressways, expressways, trunk roads and other roads, extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope, and determine the truck speed limit V for each type of road section. limit , in kilometers per hour;

[0099] 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively;

[0100] Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lon α ) and (Lon β ,Lat β ) is calculated by formula (2):

[0101]

[0102] 4) Calculate the distance between the truck location point C and the line connecting A and B using equations (3), (4), and (5), and use this distance as the distance between the road section and the truck location point. Take the road section with the smallest distance to the truck location point as the road section label for the truck location point.

[0103] Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lon B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5):

[0104]

[0105] Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max, then the positioning point is regarded as abnormal data and removed;

[0106] The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows:

[0107] The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be v i and v limit , then the normalized velocity v of the positioning point i,std The calculation formula is:

[0108]

[0109] The standardized speed sequence data is converted into a data length T data Segmentation forms fixed-length standardized speed sequence data.

[0110] Furthermore, the specific method for establishing the first model and the second model in step (3) is as follows:

[0111] The first model is trained in the following way:

[0112] The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are combined into a one-dimensional vector as input, and the overweight detection data of the truck is used as the label of the input data. Whether it is overweight is used as the output of the model, which is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model;

[0113] The first model uses standardized speed series data and truck driving attribute data as input, and outputs whether the truck is overweight. Considering that the driving speed of overweight trucks is abnormal, inputting standardized speed series data into the model can effectively capture these abnormal characteristics. For example, if the driving speed is less than the set value, it is mainly considered that the truck is driving slowly due to overweight. At the same time, the model comprehensively considers whether the truck is driving on a high-frequency overweight section. If it is driving on a high-frequency overweight section, it will be given priority attention; considers whether the truck is a high-frequency overweight truck type. If the truck is a high-frequency overweight truck type, it will be given priority attention; considers the driving time of the truck. This is because overweight trucks travel in a concentrated manner during specific periods of time, so it needs to be taken into account in the model; considers the weather when the truck travels. This is because the weather when the truck travels will affect the speed of the truck, which will affect the speed characteristics of the truck, so it needs to be taken into account in the model.

[0114] The second model is trained in the following way:

[0115] The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7);

[0116]

[0117] Where P(y=1|(v mean,std ,x1,x2,x3,x4)) represents the probability of truck overload; v mean,std represents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2,0 ,x 2,1 ,x 2,2 ,…,x 2,23 ) T , any element x in the vector 2,i are all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,…,23, x 2,i =1 means the truck is running during this time period, x 2,i =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. mean,std , regression coefficients of x1, x2, x3 and x4.

[0118] Furthermore, the method in step (4) is as follows:

[0119] (4.1) In the first warning cycle, the second model is selected as the warning model, and the truck driving attribute data is input to obtain the optimal AUC area cutoff point of the Logistic regression model as the optimal standardized speed mean threshold. It is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; the standardized speed sequence data and the truck driving attributes of this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results;

[0120] (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model are compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and an output warning of whether the truck is overweight is output; if the second model is selected, after the truck driving attribute data is input, the AUC area optimal cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed series data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; and the standardized speed series data and the truck driving attributes of this warning cycle are input into another model to output whether it is overweight, and the overweight warning accuracy of the other model is obtained by comparing the actual overweight detection results;

[0121] (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

[0122] In addition, if Figure 1 As shown, the present invention proposes an overload truck warning system based on positioning data, which includes the following modules:

[0123] The database module collects existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data into the database, and continuously adds real-time data;

[0124] The data processing module performs linear interpolation, road section matching, and abnormal data elimination on the data, and forms standardized speed sequence data according to the speed limit of the road section, and divides the speed data into fixed-length standardized speed sequences according to the specified data length;

[0125] The model module uses the fixed-length standardized speed sequence data and the truck driving attribute data as inputs, and uses the truck overweight detection data as the label of the input data to train the support vector machine model and the logistic regression model, and uses them as the first model and the second model respectively;

[0126] The feedback module determines the warning cycle length T, inputs the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issues warnings periodically according to the warning cycle length.

[0127] Furthermore, the functions of the database module are as follows:

[0128] Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time;

[0129] Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

[0130] Furthermore, the functions of the data processing module are as follows:

[0131] Linear interpolation of data refers to linear interpolation of truck positioning and speed data. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i ,Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 ,Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1):

[0132]

[0133] Matching the data to a road segment involves adding a road segment label to each location point in the truck location and speed data through the following four steps:

[0134] 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points;

[0135] 2) Divide the road sections near the track in the track map into four categories: expressways, expressways, trunk roads and other roads, and extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope. For each category, determine its truck speed limit V limit , in kilometers per hour;

[0136] 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively;

[0137] Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lon α ) and (Lon β ,Lat β ) is calculated by formula (2):

[0138]

[0139] 4) Calculate the distance between the truck location point C and the line connecting A and B by using equations (3), (4), and (5) as the distance between the road section and the truck location point, and take the road section with the smallest distance to the truck location point as the road section label of the truck location point;

[0140] Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lon B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5):

[0141]

[0142] Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max , then the positioning point is regarded as abnormal data and removed;

[0143] The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows:

[0144] The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be vi and v limit , then the normalized velocity v of the positioning point i,std The calculation formula is:

[0145]

[0146] After standardization, the standardized speed sequence data is divided into two parts according to the specified data length T. data Segmentation forms fixed-length standardized speed sequence data.

[0147] Furthermore, the model module functions as follows:

[0148] The first model is trained in the following way:

[0149] The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are combined into a one-dimensional vector as input, and the overweight detection data of the truck is used as the label of the input data. Whether it is overweight is used as the output of the model, which is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model;

[0150] The second model is trained in the following way:

[0151] The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7);

[0152]

[0153] Where P(y=1|(v mean,std ,x1,x2,x3,x4)) represents the probability of truck overload; v mean,std represents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2,0 ,x 2,1 ,x 2,2 ,…,x 2,23 ) T , any element x in the vector 2,iare all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,…,23, x 2,i =1 means the truck is running during this time period, x 2,i =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. mean,std , regression coefficients of x1, x2, x3 and x4.

[0154] Furthermore, the specific functions of the feedback module are as follows:

[0155] (4.1) In the first warning cycle, the second model is selected as the warning model, and the truck driving attribute data is input to obtain the optimal AUC area cutoff point of the Logistic regression model as the optimal standardized speed mean threshold. It is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; the standardized speed sequence data and the truck driving attributes of this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results;

[0156] (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model are compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and an output warning of whether the truck is overweight is output; if the second model is selected, after the truck driving attribute data is input, the AUC area optimal cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed series data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight warning of the truck is output; and the standardized speed series data and the truck driving attributes of this warning cycle are input into another model to output whether it is overweight, and the overweight warning accuracy of the other model is obtained by comparing the actual overweight detection results;

[0157] (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

[0158] Example

[0159] Take a data example to illustrate the process of forming a fixed-length standardized speed sequence in step (2). Suppose the truck positioning and speed data are shown in the following table:

[0160]

[0161]

[0162] According to formula (1), the latitude, longitude and speed data of 18:10:33 are supplemented, and the supplemented data are shown in the following table:

[0163] Time (hours: minutes: seconds) longitude latitude Speed ​​(km / h) 18:06:33 118.70950 32.002346 17 18:07:33 118.71073 32.001560 12 18:08:33 118.711678 32.000940 17 18:09:33 118.713235 31.999916 26 18:10:33 118.714059 31.999389 20 18:11:33 118.714883 31.998861 15 18:12:33 118.715798 31.998255 6 18:13:33 118.715870 31.998206 6

[0164] Matching the above data by road section, the road section labels of the first 6 data are trunk roads, and the road section labels of the last 2 data are other roads. The speed limit for trucks on the trunk roads in the area is 60km / h, and the speed limit for trucks on other roads is 40km / h. The standardized speed is calculated by formula (6) as shown in the following table:

[0165] Time (hours: minutes: seconds) longitude latitude Normalized speed 18:06:33 118.70950 32.002346 0.283 18:07:33 118.71073 32.001560 0.200 18:08:33 118.711678 32.000940 0.283 18:09:33 118.713235 31.999916 0.433 18:10:33 118.714059 31.999389 0.333 18:11:33 118.714883 31.998861 0.250 18:12:33 118.715798 31.998255 0.150 18:13:33 118.715870 31.998206 0.150

[0166] Assuming the data length is 4 minutes, the original data can be converted into a fixed-length standardized speed sequence. After adding the truck driving attribute data, the resulting data format is shown in the following table:

[0167]

[0168] For the first model, the data is formed into an input vector and input into the support vector machine to determine whether the truck is overweight. For the second model, the truck driving attribute data is input into the Logistic regression model to obtain the standardized speed mean threshold and make a judgment. If the threshold is 0.4, for data 1, its standardized speed is not less than 0.4, then it is judged that it is not overweight; and for data 2, its standardized speed is all below 0.4, then it is considered to be overweight and a field overweight detection is performed on it.

Claims

1. An overload truck warning method based on positioning data, characterized in that: The method comprises the following steps: (1) Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data and add them to the database, and continuously add real-time data; (2) Perform linear interpolation to fill the data, match the road sections, and remove abnormal data. The speed data is converted into standardized speed sequence data according to the road section speed limit, and divided into fixed-length standardized speed sequences according to the specified data length; (3) Using the fixed-length standardized speed sequence data and truck driving attribute data as input, and the truck overweight detection data as the label of the input data, the support vector machine model and the logistic regression model are trained respectively, and used as the first model and the second model respectively; (4) Determine the length of the warning cycle T, input the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issue warnings periodically according to the length of the warning cycle.

2. The overload truck warning method based on positioning data according to claim 1 is characterized in that: The specific method of step (1) is as follows: Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time; Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

3. The overload truck warning method based on positioning data according to claim 1 is characterized in that: The specific method of step (2) is as follows: The method of linear interpolation filling of data is as follows: The truck positioning and speed data are linearly interpolated to complete the linear interpolation. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i ,Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 ,Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1): To match the data to a road segment, we add a road segment label to each location point in the truck location and speed data through the following four steps. The specific method is as follows: 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points; 2) Divide the road sections in the trajectory map into four categories: expressways, expressways, trunk roads and other roads, extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope, and determine the truck speed limit V for each type of road section. limit , in kilometers per hour; 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively; Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lat α ) and (Lon β ,Lat β ) is calculated by formula (2): 4) Calculate the distance between the truck location point C and the line connecting A and B using equations (3), (4), and (5), and use this distance as the distance between the road section and the truck location point. Take the road section with the smallest distance to the truck location point as the road section label for the truck location point. Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lat B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5): Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max , then the positioning point is regarded as abnormal data and removed; The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows: The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be v i and V limit , then the normalized velocity v of the positioning point i,std The calculation formula is: The standardized speed sequence data is converted into a data length T data Segmentation forms fixed-length standardized speed sequence data.

4. The overload truck warning method based on positioning data according to claim 1 is characterized in that: The specific method for establishing the first model and the second model in step (3) is as follows: The first model is trained in the following way: The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are combined into a one-dimensional vector as input, and the overweight detection data of the truck is used as the label of the input data. Whether it is overweight is used as the output of the model, which is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model; The second model is trained in the following way: The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7); Where P(y=1|(v mean，std , x1, x2, x3, x4)) represents the probability of truck overload; v mean，std represents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2，0 , x 2，1 , x 2，2 , …, x 2，23 ) T , any element x in the vector 2，i are all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,...,23,x 2，i =1 means the truck is running during this time period, x 2，i =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. mean，std , regression coefficients of x1, x2, x3 and x4.

5. The overload truck warning method based on positioning data according to claim 1 is characterized in that: The method in step (4) is as follows: (4.1) In the first warning cycle, the second model is selected as the warning model, the truck driving attribute data is input, the optimal AUC area cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight truck warning is output; The standardized speed sequence data and the driving attributes of the truck in this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results; (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model is compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and the output is whether the truck is overweight warning; If the second model is selected, after inputting the truck driving attribute data, the optimal AUC area cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight truck warning is output; and the standardized speed sequence data and the truck driving attributes of this warning period are input into another model to output whether it is overweight, and the overweight warning accuracy of another model is obtained by comparing the actual overweight detection results; (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

6. An overload truck warning system based on positioning data, characterized in that: The system includes the following modules: The database module collects existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data into the database, and continuously adds real-time data; The data processing module performs linear interpolation, road section matching, and abnormal data elimination on the data, and forms standardized speed sequence data according to the speed limit of the road section, and divides the speed data into fixed-length standardized speed sequences according to the specified data length; The model module uses the fixed-length standardized speed sequence data and the truck driving attribute data as inputs, and uses the truck overweight detection data as the label of the input data to train the support vector machine model and the logistic regression model, and uses them as the first model and the second model respectively; The feedback module determines the warning cycle length T, inputs the collected fixed-length standardized speed sequence data and truck driving attribute data into the first model and the second model, and issues warnings periodically according to the warning cycle length.

7. The overload truck warning system based on positioning data according to claim 6 is characterized in that: The functions of the database module are as follows: Collect existing truck positioning and speed data, truck overweight detection data, and truck driving attribute data, wherein the truck positioning and speed data include the longitude, latitude, and speed of each positioning point during the truck's driving process; the truck overweight detection data gives a label of whether the truck is overweight; the truck driving attribute data includes whether the truck is driving on a road section with a high frequency of overweight, the time when the truck is driving, the weather type when the truck is driving, and whether the truck is a type of truck with a high frequency of overweight, wherein the road section with a high frequency of overweight is a road section where the number of overweight occurrences exceeds a first threshold within a preset time; the type of truck with a high frequency of overweight is a truck type where the number of overweight occurrences exceeds a second threshold within a preset time; Adding real-time data means adding truck positioning and speed data, truck overweight detection data, and truck driving attribute data to the database based on real-time overweight detection results on the road.

8. The overload truck warning system based on positioning data according to claim 6 is characterized in that: The functions of the data processing module are as follows: Linear interpolation of data refers to linear interpolation of truck positioning and speed data. The specific method of linear interpolation is as follows: let the time of the i-th positioning point in the truck positioning and speed data be t i , the longitude and latitude coordinates are (Lon i , Lat i ), with a velocity of v i , the time of the i+1th positioning point is t i+1 The longitude and latitude coordinates are (Lon i+1 , Lat i+1 ), with a velocity of v i+1 , where t i+1 >t i ; At this time, the positioning point data at time t′ is supplemented, including the longitude and latitude coordinates (Lon, Lat) and the speed v, t i <t′<t i+1 , then the new positioning point data is calculated by formula (1): To match the data to a road segment, we add a road segment label to each location point in the truck location and speed data through the following four steps: 1) Use the Folium library in Python and OSM map to process the truck location and speed data to form a trajectory map with location points; 2) Divide the road sections near the track in the track map into four categories: expressways, expressways, trunk roads and other roads, and extract the longitude and latitude coordinates of the centerline positioning points of each road section within the research scope. For each category, determine its truck speed limit V limit , in kilometers per hour; 3) For each truck positioning point, calculate the distance between it and the longitude and latitude of all positioning points of the center line of each section within the study range through formula (2). For each section within the study range, select the two center line positioning points with the smallest distance to the truck positioning point, and let these two positioning points be A and B respectively; Let the radius of the earth be R, and the two longitude and latitude coordinates (Lon α ,Lat α ) and (Lon β ,Lat β ) is calculated by formula (2): 4) Calculate the distance between the truck location point C and the line connecting A and B by using equations (3), (4), and (5) as the distance between the road section and the truck location point, and take the road section with the smallest distance to the truck location point as the road section label of the truck location point; Let the longitude and latitude coordinates of A and B be (Lon A ,Lat A ) and (Lat B ,Lat B ), the longitude and latitude coordinates of C are (Lon C ,Lat C ), the center of the earth is O, let A′, B′, C′, O′ be the coordinates of points A, B, C, O in three-dimensional Cartesian coordinates, then the distance d′ between the truck positioning point and the line connecting the two centerline positioning points is calculated by formula (3)(4)(5): Abnormal data is eliminated, that is, the distance d′ between the truck positioning point and the road section is judged according to the road section matching: if d′ is greater than the set maximum distance d max , then the positioning point is regarded as abnormal data and removed; The method of converting speed data into standardized speed series data according to the speed limit of the road section is as follows: The speed of each truck location point is calculated by the speed limit of the corresponding road type to form a standardized speed sequence. Let the speed of the truck location point and the speed limit of the corresponding road type be v i and V limit , then the normalized velocity v of the positioning point i,std The calculation formula is: After standardization, the standardized speed sequence data is divided into two parts according to the specified data length T. aata Segmentation forms fixed-length standardized speed sequence data.

9. The overload truck warning system based on positioning data according to claim 6 is characterized in that: The model module functions are as follows: The first model is trained in the following way: The fixed-length standardized speed sequence data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are combined into a one-dimensional vector as input, and the overweight detection data of the truck is used as the label of the input data. Whether it is overweight is used as the output of the model, which is substituted into the model and the sigmoid function is used as the kernel function to train the binary classification support vector machine, so as to obtain the first model; The second model is trained in the following way: The fixed-length standardized speed series data, whether the truck is traveling on a high-frequency overweight section, the time when the truck is traveling, the weather type when the truck is traveling, and whether the truck is a high-frequency overweight truck type are used as input, the overweight detection data of the truck is used as the label of the input data, and the probability of overweight is used as the model output. Substitute it into the Logistic regression model for fitting. The regression formula is shown in formula (7); Where P(y=1|(v mean,std , x1, x2, x3, x4)) represents the probability of truck overload; v mean,std represents the standardized speed mean; x1 represents the dummy variable of whether the truck is traveling on a road section with a high frequency of overweight, x1=1 represents the road section with a high frequency of overweight, and x1=0 represents the road section with a high frequency of overweight; x2 represents the dummy variable vector of the time when the truck is traveling, x2=(x 2，0 , x 2，1 , x 2，2 , …, x 2，23 ) T , any element x in the vector 2，i are all dummy variables, indicating whether the truck travels between hours i and i+1, i=0,1,2,...,23,x 2，i =1 means the truck is running during this time period, x 2，l =0 does not travel in this time period, x2 is a vector with one element being 1 and the rest being 0; x3 is a dummy variable representing the weather type when the truck is traveling, x2=1 represents the truck traveling on a sunny day, and x2=0 represents the truck traveling on a rainy day; x4 is a dummy variable representing whether it is an overweight high-frequency truck type, x4=1 represents the truck is an overweight high-frequency truck type, and x4=0 represents the truck is other types, where β0 is the constant term coefficient, β1, β2, β3, β4 and β5 are the variables v respectively. meanstd , regression coefficients of x1, x2, x3 and x4.

10. The overload truck warning system based on positioning data according to claim 6, characterized in that: The specific functions of the feedback module are as follows: (4.1) In the first warning cycle, the second model is selected as the warning model, the truck driving attribute data is input, the optimal AUC area cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight truck warning is output; The standardized speed sequence data and the driving attributes of the truck in this warning cycle are input into the first model to output whether it is overweight, and the overweight warning accuracy of the second model is obtained by comparing the actual overweight detection results; (4.2) In the next warning cycle, the overweight warning accuracy of the model selected in the previous warning cycle and the other model is compared, and the model with the highest accuracy is selected as the model for the next warning cycle. If the selected model is the first model, the standardized speed series data and the truck driving attributes are input into the support vector machine, and the output is whether the truck is overweight warning; If the second model is selected, after inputting the truck driving attribute data, the optimal AUC area cutoff point of the Logistic regression model is obtained as the optimal standardized speed mean threshold, and it is determined whether the standardized speed sequence data are all less than the standardized speed mean threshold. If they are all less than the standardized speed mean threshold, an overweight truck warning is output; and the standardized speed sequence data and the truck driving attributes of this warning period are input into another model to output whether it is overweight, and the overweight warning accuracy of another model is obtained by comparing the actual overweight detection results; (4.3) Repeat (4.2) and issue warnings periodically according to the length of the warning cycle.

Images