Bridge heavy-load vehicle identification method and system based on multi-source data fusion, and storage medium

Abstract

The invention relates to the technical field of bridge safety monitoring, in particular to a bridge heavy-load vehicle identification method and system based on multi-source data fusion and a storage medium. According to the bridge heavy-load vehicle identification method based on multi-source data fusion, firstly, a plurality of strain sensors are arranged on a bridge, at least two strain sensors are distributed on a straight line parallel to the length direction of the bridge, and at least two strain sensors are distributed on a straight line parallel to the width direction of the bridge; tracking the vehicle, collecting strain data of the strain sensor when the vehicle passes through the strain sensor, and extracting a strain peak value sp and a strain valley value st; and inputting the vehicle speed vp, the strain peak value sp and the strain valley value st into the prediction model, and outputting a load prediction value by the prediction model. According to the method, the vehicle load is reversely deduced by using the response generated by the bridge when the heavy-load vehicle passes through and combining with a machine learning algorithm, so that the accuracy and timeliness of bridge safety monitoring are effectively improved, and the operation and maintenance cost is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge safety monitoring, and in particular to a method, system and storage medium for identifying heavy-duty vehicles on bridges based on multi-source data fusion. Background Art

[0002] With the rapid development of the bridge construction industry, the importance of bridges in the national economy has become increasingly prominent. However, during the use of bridges, they are subjected to various loads, and among them, vehicle loads are the main factors causing the reduction of the reliability of bridge structures and fatigue damage. Therefore, accurately identifying and monitoring vehicle loads on bridges, especially heavy-duty vehicle loads, is of great significance for ensuring the safe operation of bridges and preventing structural damage. Traditional bridge load identification methods often rely on single monitoring means, such as dynamic weighing systems or bridge structure response monitoring. Although these methods can reflect the stress conditions of bridges to a certain extent, they have problems such as high costs of monitoring equipment, limited monitoring ranges, and low data fusion degrees. Through single monitoring means, only specific areas of bridges can often be covered, and the overall stress conditions of bridges cannot be comprehensively reflected. The data generated by different monitoring means are often independent of each other, lacking effective data fusion and comprehensive analysis means. Summary of the Invention

[0003] In order to overcome the defect in the prior art that there are no effective data fusion and analysis means for bridge stress analysis, the present invention proposes a method for identifying heavy-duty vehicles on bridges based on multi-source data fusion. By using the responses generated by bridges when heavy-duty vehicles pass through and combining machine learning algorithms, vehicle loads are inversely deduced, effectively improving the accuracy and timeliness of bridge safety monitoring and greatly reducing operation and maintenance costs.

[0004] A method for identifying heavy-duty vehicles on bridges based on multi-source data fusion proposed by the present invention first arranges a plurality of strain sensors on the bridge, where at least two strain sensors are distributed on a straight line parallel to the bridge length direction, and at least two strain sensors are distributed on a straight line parallel to the bridge width direction;

[0005] Track the vehicle, collect the strain data of the strain sensors when the vehicle passes through the strain sensors, and extract the strain peak value sp and the strain valley value st;

[0006] Input the vehicle speed vp, the strain peak value sp, and the strain valley value st into the prediction model, and the prediction model outputs the load prediction value.

[0007] Preferably, analyze the monitoring video through the target detection algorithm, obtain the time when the vehicle passes through the strain sensor, obtain the strain data of the strain sensor at this time, and extract the strain peak value and the strain valley value from its fluctuation region.

[0008] Preferably, the fluctuation region corresponding to the passing time of the vehicle in the strain data is recorded as the vehicle vibration data. After preprocessing the vehicle vibration data, the strain peak value and the strain valley value are extracted.

[0009] Preferably, the preprocessing of the vehicle vibration data includes deburring. The deburring identification method is as follows: First, divide the vehicle vibration data into multiple data intervals with equal numerical widths; Let the mean value of the data interval containing the largest number of numerical values be used as the reference value; Calculate the difference between the numerical value at each time point of the vehicle vibration data and the reference value as the fluctuation at that time point, and statistically identify the continuous time points with fluctuations greater than the set deburring threshold as burrs.

[0010] Preferably, after detecting burrs in the vehicle vibration data, the burrs are deleted, and the interpolation algorithm is used to fill the breakpoints.

[0011] Preferably, the preprocessing of the vehicle vibration data includes detrending. The method is as follows: Discretize the vehicle vibration data. The discrete sequence is divided into G arrays, and the base point value of each array is calculated; Define the trend arrays of the first G - 1 arrays, and the numerical values of each array correspond one by one to the numbers in the corresponding trend arrays; The trend array is an arithmetic sequence from the base point value of the current array to the base point value of the next array;

[0012] For the first G - 1 arrays, let their numerical values be updated to the difference between the current numerical value and the numerical value in the corresponding trend array; Combine the updated first G - 1 arrays to form the detrending processing result of the vehicle vibration data.

[0013] Preferably, the calculation method of the base point value of each array is: the mean value of the 1 / 4 quantile, the 3 / 4 quantile, and the two 1 / 2 quantiles.

[0014] Preferably, the prediction model uses the Transformer model.

[0015] A bridge heavy - load vehicle identification system based on multi - source data fusion proposed by the present invention includes a memory and a processor. A computer program is stored in the memory. The processor is connected to the memory, and the processor is used to execute the computer program to implement the above - mentioned bridge heavy - load vehicle identification method based on multi - source data fusion.

[0016] A storage medium proposed by the present invention stores a computer program, and when the computer program is executed, it is used to implement the above - mentioned bridge heavy - load vehicle identification method based on multi - source data fusion.

[0017] The advantages of the present invention are as follows:

[0018] (1) A method for identifying heavy - loaded vehicles on bridges based on multi - source data fusion proposed by the present invention estimates the vehicle weight by combining bridge strain data and vehicle speed. In the present invention, first, through the time correlation between vehicle tracking and the fluctuation of strain data, the strain amplitude of the bridge caused by the vehicle is obtained, and then the peak value and valley value are extracted to characterize the strain characteristics. The present invention performs data fusion on the strain peak value, valley value, and vehicle speed, and predicts the vehicle weight, simplifying the data fusion and analysis means, and achieving higher prediction accuracy in combination with the prediction model.

[0019] (2) The present invention introduces multi - source data fusion technology into the field of bridge load identification. By integrating data generated by various monitoring means (such as vehicle load detection data, bridge structure response data, etc.) and combining artificial intelligence algorithms, the complementarity and enhancement of data are realized, so as to more accurately identify and monitor the vehicle load on the bridge. Through the fusion and correlation matching of multi - source data, the vehicle load on the bridge can be more accurately identified and monitored.

[0020] (3) The present invention uses the data of existing monitoring equipment for fusion analysis, without the need to install additional high - precision monitoring equipment, reducing the monitoring cost. And it realizes the effective fusion and comprehensive analysis of data generated by different monitoring means, improving the utilization value of data.

[0021] (4) The present invention has broad application prospects in the fields of bridge health monitoring, structural damage warning, etc. By real - time monitoring the vehicle load situation on the bridge, potential safety hazards can be discovered in time and effective measures can be taken for disposal, thus ensuring the safe operation of the bridge and extending its service life.

[0022] (5) The invention solves the core problems in traditional technologies such as low data decoupling accuracy, sensitivity to environmental interference, and high deployment cost through multi - source data spatio - temporal fusion, high - precision synchronization mechanism, and lightweight Transformer model, achieving a comprehensive breakthrough in accuracy, efficiency, and cost in the field of bridge heavy - loaded vehicle identification, and having significant technological advancement and commercial application value. Brief Description of the Drawings

[0023] Figure 1 is a flow chart of a method for identifying heavy - loaded vehicles on bridges based on multi - source data fusion proposed by the present invention;

[0024] Figure 2(a) shows the layout scenario of strain sensors in the embodiment;

[0025] Figure 2(b) is a schematic diagram of the layout of strain sensors in the embodiment;

[0026] Figure 3(a) shows the distribution scenario of strain sensors in Layout Scheme 1;

[0027] Figure 3(b) is a schematic diagram of the distribution of strain sensors in Layout Scheme 1;

[0028] Figure 4(a) shows the distribution scenario of strain sensors in the second layout plan;

[0029] Figure 4(b) is a schematic diagram of the distribution of strain sensors in the second layout plan;

[0030] Figure 5(a) shows the distribution scenario of strain sensors in the third layout plan;

[0031] Figure 5(b) is a schematic diagram of the distribution of strain sensors in the third layout plan;

[0032] Figure 6 are the data collected by the strain sensors. Specific implementation manner

[0033] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0034] Refer to Figure 1 , a method for identifying heavy-duty vehicles on bridges based on multi-source data fusion proposed by the present invention. First, a plurality of strain sensors are arranged on the bridge, where at least two strain sensors are distributed on a straight line parallel to the length direction of the bridge, and at least two strain sensors are distributed on a straight line parallel to the width direction of the bridge;

[0035] Track the vehicle through video monitoring. When the vehicle passes through the strain sensors, obtain the vehicle speed vp, collect the strain data of the strain sensors, and extract the strain peak value sp and strain valley value st corresponding to the strain sensors; input the vehicle speed vp, strain peak value sp, and strain valley value st into the prediction model, and the prediction model outputs the load prediction value;

[0036] Obtain the load prediction values corresponding to each strain sensor and take the average as the final load of the vehicle.

[0037] The prediction model can specifically adopt the Transformer model.

[0038] In specific implementation, analyze the monitoring video through an object detection algorithm such as YOLOv8s to obtain the time when the vehicle passes through the strain sensors. Taking the time as a reference, associate the vehicle and the fluctuation regions of the strain sensor detection data, and extract the strain peak value sp and strain valley value st from the fluctuation regions.

[0039] In this embodiment, the fluctuation region extracted from the detection data of the strain sensor is denoted as vehicle vibration data. After extracting the vehicle vibration data, data preprocessing is first performed, and then the maximum value is extracted as the strain peak value sp, and the minimum value is extracted as the strain trough value st.

[0040] In this embodiment, the preprocessing of the vehicle vibration data includes: deburring and detrending.

[0041] The deburring method is as follows: First, detect and remove the burr anomaly, and then use the interpolation algorithm to fill the breakpoints.

[0042] The burr anomaly detection method proposed in this embodiment divides the vehicle vibration data into multiple data intervals with equal numerical widths, and the numerical width of the data interval is the difference between the maximum value and the minimum value on the data interval;

[0043] Let the mean value of the data interval containing the largest number of numerical values be used as the reference value;

[0044] Calculate the difference between the numerical value at each time point of the vehicle vibration data and the reference value as the fluctuation at that time point, and statistically identify the continuous time points with fluctuations greater than the burr threshold as burrs.

[0045] The burr anomaly detection method specifically includes the following steps:

[0046] S1. Obtain the vehicle vibration data X and divide it into equal-width continuous data intervals;

[0047] In specific implementation, the data interval width can be set to w;

[0048]

[0049] X = {x1, x2, …, x n , …, x N}

[0050] where maxX is the maximum value of the vehicle vibration data X, minX is the minimum value of the vehicle vibration data X; x n is the numerical value at the nth time point in the vehicle vibration data X, 1 ≤ n ≤ N, and N is the number of numerical values in the vehicle vibration data X; that is, x1, x2, x N are the numerical values at the 1st, 2nd, and Nth time points in the vehicle vibration data X respectively;

[0051] k is the number of data intervals, and specifically, k can take the value k = 1 + 3.322log 10 (N).

[0052] Then the numerical range of the data interval i is [minX + (i - 1)w, minX + iw], 1 ≤ i ≤ k.

[0053] S2. Calculate the frequency of each data interval, where the frequency is the number of values contained in the data interval.

[0054] S3. Calculate the average value of the data interval with the largest frequency as the reference value D x ;

[0055] S4. Calculate the fluctuation curve of the vehicle vibration data X, and denote the fluctuation of the value x n as dif n ;

[0056] dif n = |x n - D x |

[0057] S5. Determine whether dif n is greater than the spike threshold;

[0058] If it is less than the threshold, it is defined as normal;

[0059] If it is greater than the threshold, it is defined as a spike point and step S6 is executed;

[0060] S6. Statistically identify spike points that are identified as spike points in continuous time and whose continuous time length is greater than or equal to the set point threshold as spike anomalies; regard the occurrence event of the first spike point in continuous time as the spike occurrence time, and the time span of continuous spike points as the duration of the spike.

[0061] The method for detrending vehicle vibration data includes the following steps:

[0062] St1. Discretize the vehicle vibration data X into a discrete sequence, expressed as a set of numerical points Y = {y i , 1 ≤ i ≤ I}, where I is the total number of numerical points; set the array length n and divide the set of numerical points into G arrays;

[0063] If I / n is an integer, or the remainder of I / n is less than 0.5n, then

[0064] If the remainder of I / n is greater than or equal to 0.5n, then

[0065] represents the floor of the quotient of I / n; n can specifically take values between 1 / 30 and 1 / 10 of I, and the value of n affects the accuracy of the trend line type.

[0066] St2. Calculate the base value of the array in combination with the quartiles of the array;

[0067] Denote the numbers at the 25%, 50%, and 75% positions of the g-th array as Q1 g and Q2g and Q3 g , the base value B of the g-th array g =(Q1 g +Q3 g +2Q2 g ) / 4; 1 ≤ g ≤ G;

[0068] St3. Calculate the trend array R corresponding to the g-th array g and the trend data R corresponding to the set of numerical points Y = {y i , 1 ≤ i ≤ I};

[0069] R g =[B g , B g +m g , B g +2m g , ……, B g+1 , g + 1 ≤ G

[0070] m g =(B g+1 -B g ) / n g

[0071] B g+1 is the base value of the (g + 1)-th array, and m g is the numerical interval;

[0072] n g is the data volume of the g-th array, and the data volumes of the first G - 1 arrays are all n;

[0073] R = {R g ; 1 ≤ g ≤ G - 1} = {B g , B g +m g , B g +2m g ……B g+1} 1≤g≤G-1 ={r j , 1 ≤ j ≤ n(G - 1)}

[0074] St4. Subtract the trend data R from the first n(G - 1) data of Y to obtain the data H after removing the temperature effect;

[0075] H = {h j , 1 ≤ j ≤ n(G - 1)}

[0076] h j =y j -r j

[0077] y j ∈Y = {yi , 1 ≤ i ≤ I}

[0078] The data H is the detrended result of the vehicle vibration data.

[0079] It can be seen that after detrending the discrete sequence Y of the vehicle vibration data X with length I, the length of the detrended result obtained is n(G - 1), and n(G - 1) < I.

[0080] In this way, when segmenting the vehicle vibration data, an interval can be extended backward so that the first G - 1 groups of data can fully cover the vehicle vibration, realizing the detrending of the vehicle vibration fluctuation interval.

[0081] The following combines specific embodiments to verify the bridge heavy - vehicle recognition method based on multi - source data fusion proposed by the present invention.

[0082] In this embodiment, for the convenience of the experiment, vehicles with different loads are made to pass through the bridge at a constant speed.

[0083] In this embodiment, the YOLOv8s detection algorithm is used to analyze the monitoring video to obtain the time of the vehicle strain sensor, and the extreme - value time point of the strain sensor is obtained at the elapsed time as the vehicle passing time point; the extreme value of the strain sensor refers to the value with a larger absolute value among the wave crest and wave trough, and the extreme - value time point is the acquisition time of the extreme value.

[0084] In this embodiment, the difference between the vehicle passing time points corresponding to the same vehicle of two adjacent strain sensors is calculated as the vehicle running time, and the ratio of the distance between the two adjacent strain sensors to the vehicle running time is used as the vehicle speed vp of this vehicle.

[0085] In this embodiment, 5 strain sensors shown in Fig. 2(a) and Fig. 2(b) are set on a bridge with a length of L. Along the bridge length direction, the distance between adjacent strain sensors is 4 / L.

[0086] Referring to Fig. 3(a) and Fig. 3(b), let strain sensors 1, 4, and 5 form layout plan one; referring to Fig. 4(a) and Fig. 4(b), let strain sensors 1, 3, and 4 form layout plan two; referring to Fig. 5(a) and Fig. 5(b), let strain sensors 2, 4, and 5 form layout plan three.

[0087] Obviously, in layout plan one, strain sensors 1 and 4 are distributed along the bridge width direction; strain sensors 4 and 5 are distributed along the bridge length direction, and the distance between them is 0.5L.

[0088] In this embodiment, multi-condition comparative experiments are designed. Under the first condition, the vehicle load is 35 tons and the vehicle speed is 40 kilometers per hour. Under the second condition, the vehicle load is 55 tons and the vehicle speed is 60 kilometers per hour. Under the third condition, the vehicle load is 65 tons and the vehicle speed is 50 kilometers per hour.

[0089] In this embodiment, in the first layout plan, the peak values sp and valley values st of the vehicle corresponding to the three strain sensors 1, 4, and 5 are respectively extracted. For each strain sensor, the collected peak value sp, valley value st, and the calculated vehicle speed vp are input into the pre-trained prediction model, and then the predicted vehicle load is output by the prediction model. Then, the average value of the three predicted loads is taken as the calculated vehicle weight value.

[0090] In the second layout plan, the peak values sp and valley values st of the vehicle corresponding to the three strain sensors 1, 3, and 4 are respectively extracted. For each strain sensor, the collected peak value sp, valley value st, and the calculated vehicle speed vp are input into the pre-trained prediction model, and then the predicted vehicle load is output by the prediction model. Then, the average value of the three predicted loads is taken as the calculated vehicle weight value.

[0091] In the third layout plan, the peak values sp and valley values st of the vehicle corresponding to the three strain sensors 2, 4, and 5 are respectively extracted. For each strain sensor, the collected peak value sp, valley value st, and the calculated vehicle speed vp are input into the pre-trained prediction model, and then the predicted vehicle load is output by the prediction model. Then, the average value of the three predicted loads is taken as the calculated vehicle weight value.

[0092] In this embodiment, the vehicle weight inversion error = |calculated vehicle weight value - actual vehicle weight value| / actual vehicle weight value;

[0093] The vehicle speed inversion error = |calculated vehicle speed value - actual vehicle speed value| / actual vehicle speed value;

[0094] Under different conditions, the vehicle weight and vehicle speed evaluation indexes obtained under each layout plan are shown in Tables 1, 2, and 3 below.

[0095] Table 1 Statistics of Condition 1

[0096]

[0097] Table 2 Statistics of Condition 2

[0098]

[0099] Table 3 Statistics of Condition 3

[0100]

[0101] Combined with Tables 1 - 3, it can be seen that under various working conditions, the vehicle weight inversion error of the first layout plan is much smaller than that of the second and third layout plans; the vehicle speed inversion error of the first layout plan is also smaller than that of the second and third layout plans.

[0102] Under the three working conditions, the average vehicle weight inversion error of the first layout plan is only 4.1%, significantly lower than 5.8% of the second layout plan and 6.6% of the third layout plan, and the correlation coefficient is close to perfect (R 2 = 0.97).

[0103] Under the three working conditions, the average vehicle speed inversion error of the first layout plan is as low as 3.3%, 4.7% for the second layout plan, and 3.4% for the third layout plan. And the average vehicle speed correlation coefficient R 2 of the first layout plan reaches 0.96, indicating that the vehicle speed prediction is highly consistent with the actual value.

[0104] Under the three working conditions, the standard deviation of the vehicle weight and vehicle speed inversion errors of the first layout plan is 0.2, the standard deviation of the vehicle weight and vehicle speed inversion errors of the second layout plan is 0.5, and the standard deviation of the vehicle weight inversion error of the third layout plan is 0.7. It can be seen that the first layout plan performs more stably under complex working conditions and has high precision, strong correlation, and strong stability in both vehicle weight and vehicle speed inversion, making it the optimal choice for complex load inversion scenarios.

[0105] During specific implementation, it may be due to the vehicles being too close to each other, resulting in the connection of vehicle vibration data. For example, Figure 6 in the strain sensor detection data shown; the fluctuation regions A and C are the fluctuations when a single vehicle passes by; the data B and D are the strain data fluctuations caused by two vehicles passing by one after another.

[0106] At this time, there are two pairs of peaks and valleys distributed front and back in the fluctuation region B. The first pair of peaks and valleys can be taken as the strain peak value and strain valley value of the previous vehicle, and the second pair of valleys and peaks can be taken as the strain valley value and strain peak value of the following vehicle. In the fluctuation region D, there is only one valley and two peaks. The first peak and valley can be taken as the strain peak value and strain valley value of the previous vehicle, and the valley and the second peak can be taken as the strain valley value and strain peak value of the following vehicle.

[0107] Of course, for those skilled in the art, the present invention is not limited to the details of the above exemplary embodiments, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non - restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0108] In addition, it should be understood that although this specification is described in terms of embodiments, not every embodiment contains only an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0109] The technologies, shapes, and structures not described in detail in the present invention are all well-known technologies.

Claims

1. A method for identifying overloaded vehicles on bridges based on multi-source data fusion, characterized in that First, arrange multiple strain sensors on the bridge, where at least two strain sensors are distributed on a straight line parallel to the length direction of the bridge, and at least two strain sensors are distributed on a straight line parallel to the width direction of the bridge; Track the vehicle, collect the strain data of the strain sensors when the vehicle passes by the strain sensors, and extract the strain peak value sp and the strain valley value st; Input the vehicle speed vp, the strain peak value sp, and the strain valley value st into the prediction model, and the prediction model outputs the load prediction value.

2. The method for identifying overloaded vehicles on bridges based on multi-source data fusion according to claim 1, wherein Analyze the monitoring video through the object detection algorithm to obtain the time when the vehicle passes by the strain sensor, obtain the strain data of the strain sensor at this time, and extract the strain peak value and the strain valley value from its fluctuation region.

3. The method for identifying heavy-load vehicles on bridges based on multi-source data fusion according to claim 2, wherein Obtain the fluctuation region corresponding to the vehicle passing time on the strain data as the vehicle vibration data. After preprocessing the vehicle vibration data, then extract the strain peak value and the strain valley value.

4. The method for identifying overloaded vehicles on bridges based on multi-source data fusion according to claim 3, characterized in that, The preprocessing of the vehicle vibration data includes deburring. The deburring recognition method is as follows: First, divide the vehicle vibration data into multiple data intervals with equal numerical widths; let the mean value of the data interval with the largest number of numerical values be used as the reference value; calculate the difference between the numerical value at each time point of the vehicle vibration data and the reference value as the fluctuation at this time point, and statistically identify the continuous time points with fluctuations greater than the set deburring threshold as burrs.

5. The method for identifying overloaded vehicles on bridges based on multi-source data fusion according to claim 4, characterized in that, After detecting burrs on the vehicle vibration data, delete the burrs and use the interpolation algorithm to fill the breakpoints.

6. The method for identifying heavy-duty vehicles on bridges based on multi-source data fusion according to claim 3, characterized in that, The preprocessing of the vehicle vibration data includes detrending. The method is as follows: Discretize the vehicle vibration data, divide the discrete sequence into G arrays, and calculate the base point values of each array; define the trend arrays of the first G - 1 arrays, and the numerical values of each array correspond one by one to the numbers in the corresponding trend arrays; the trend array is an arithmetic sequence from the base point value of the current array to the base point value of the next array; For the first G - 1 arrays, let their numerical values be updated to the difference between the current numerical value and the numerical value of the corresponding trend array; merge the updated first G - 1 arrays to form the detrending processing result of the vehicle vibration data.

7. The method for identifying overloaded vehicles on bridges based on multi-source data fusion according to claim 6, wherein The calculation method of the base point value of each array is: the mean value of the 1 / 4 quantile, the 3 / 4 quantile, and the two 1 / 2 quantiles.

8. The method for identifying overloaded vehicles on bridges based on multi-source data fusion according to any one of claims 1-7, characterized in that The prediction model uses the Transformer model.

9. A bridge overloaded vehicle recognition system based on multi-source data fusion, characterized in that, It includes a memory and a processor. The memory stores a computer program, and the processor is connected to the memory. The processor is used to execute the computer program to implement the method for identifying heavy-duty vehicles on bridges based on multi-source data fusion as described in any one of claims 1 - 8.

10. A storage medium, characterized in that, Stores a computer program, and when the computer program is executed, it is used to implement the method for identifying heavy-duty vehicles on bridges based on multi-source data fusion as described in any one of claims 1 - 8.

Images