Method and system for locating spatial distribution of pavement load through multi-source heterogeneous data fusion

Abstract

The present invention belongs to the technical field of pavement load positioning. In order to solve the problem that the strain distribution of the entire scale space of the pavement cannot be comprehensively monitored at present, a method and system for positioning the spatial distribution of pavement loads by fusing multi-source heterogeneous data are provided. Among them, the method includes obtaining pavement lidar data and pavement mechanical data at each moment of a set section; fitting and removing the ground from the pavement lidar data, and based on the pavement lidar data after removing the fitted ground, clustering to identify vehicle types and their corresponding loads, and then tracking the vehicle trajectories; using the finite element mechanical analysis method to decouple the vehicle position and pavement stress from the pavement mechanical data, and then combining the tracked vehicle trajectories to generate a pavement stress and load distribution map in three-dimensional space, ensuring the accurate and real-time identification of the spatial distribution of pavement vehicle loads.

Description

Technical Field

[0001] The present invention belongs to the technical field of pavement load positioning, and particularly relates to a method and system for positioning the spatial distribution of pavement loads through multi-source heterogeneous data fusion. Background Art

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] With the increase in the service life of highways, existing pavement structures face problems such as a decline in bearing capacity and difficulty in timely detection of diseases. Especially under the action of vehicle loads, the risk of structural failure increases, seriously threatening traffic safety. The core of these problems lies in the difficult conversion and connection of various stage states in the life cycle of existing pavement structures, which cannot dynamically reflect the changes in traffic operation status and pavement performance, resulting in a lack of refined and real-time basis for maintenance management.

[0004] At present, lidar based on the speed of light has made progress in traffic flow recognition, but there are still problems of insufficient comprehensiveness and real-time in vehicle load distribution recognition and pavement mechanics perception. Although distributed fiber optic sensing technology can detect vibration signals in the environment, limited by construction conditions and equipment costs, it cannot comprehensively monitor the strain distribution of the entire pavement scale space and accurately locate and real-time monitor pavement loads. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for positioning the spatial distribution of pavement loads through multi-source heterogeneous data fusion, which is based on the multi-source heterogeneous data fusion method to achieve accurate positioning and real-time monitoring of pavement loads.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The first aspect of the present invention provides a method for positioning the spatial distribution of pavement loads through multi-source heterogeneous data fusion.

[0008] In one or more embodiments, a method for positioning the spatial distribution of pavement loads through multi-source heterogeneous data fusion is provided, including:

[0009] Obtain the pavement lidar data and pavement mechanics data at each moment of a set section; wherein, the pavement mechanics data is the pavement stress when a vehicle passes by.

[0010] Fit and remove the ground from the pavement lidar data, and based on the pavement lidar data after removing the fitted ground, cluster and identify the vehicle type and its position, and dynamically track the vehicle trajectory.

[0011] Use the finite element mechanics analysis method to decouple the vehicle position and pavement stress from the pavement mechanical data, and then combine the tracked vehicle trajectory to generate the pavement stress and load distribution map in three-dimensional space.

[0012] As an implementation, according to the pavement lidar data after removing the fitted ground, divide the set section into several continuous cubes, and calculate the point density of each cube; by comparing the point density of each cube with the dynamic threshold of the point density, distinguish the background and moving vehicles.

[0013] As an implementation, when the point density of any cube is higher than the average value of the point densities of all cubes, increase the point density threshold; where the adjusted point density threshold = α × the initial point density threshold; α > 1, which is a preset first adjustment coefficient.

[0014] As an implementation, when the point density of any cube is lower than the average value of the point densities of all cubes and the point density variance of the current cube is less than the preset variance threshold, decrease the point density threshold; where the adjusted point density threshold = β × the initial point density threshold; β < 1, which is a preset second adjustment coefficient.

[0015] As an implementation, by analyzing the change of the point density of each cube, adaptively adjust the upper and lower limits of the point density threshold based on the histogram.

[0016] As an implementation, in the process of generating the pavement stress and load distribution map in three-dimensional space, ensure the alignment of the pavement lidar data and the pavement mechanical data on the same time axis through time synchronization and spatial correlation, and match the point cloud in the pavement lidar data with the stress data in the pavement mechanical data through the vehicle position information.

[0017] As an implementation, use the data fusion algorithm combining Kalman filter and particle filter to process the decoupled vehicle position and pavement stress and the vehicle trajectory, and generate the pavement stress and load distribution map in three-dimensional space.

[0018] The second aspect of the present invention provides a pavement load spatial distribution positioning system for multi-source heterogeneous data fusion.

[0019] In one or more embodiments, a pavement load spatial distribution positioning system for multi-source heterogeneous data fusion includes:

[0020] A multi-source heterogeneous data acquisition module, which is used to acquire the pavement lidar data and pavement mechanical data at each moment of the set section; where the pavement mechanical data is the pavement stress when the vehicle passes by.

[0021] A vehicle recognition and tracking module, which is used to fit the ground from the road surface lidar data and remove it. Based on the road surface lidar data after removing the fitted ground, it clusters and identifies the vehicle type and its position, and dynamically tracks the vehicle trajectory;

[0022] A stress and load distribution map generation module, which uses the finite element mechanics analysis method to decouple the vehicle position and road surface stress from the road surface mechanics data, and then combines the tracked vehicle trajectory to generate a road surface stress and load distribution map in three-dimensional space.

[0023] As an implementation, in the vehicle recognition and tracking module, according to the road surface lidar data after removing the fitted ground, a set section is divided into several continuous cubes, and the point density of each cube is calculated; by comparing the point density of each cube with the dynamic threshold of the point density, the background and moving vehicles are distinguished.

[0024] In another embodiment, a pavement load spatial distribution positioning system for multi-source heterogeneous data fusion includes: a lidar, a distributed fiber optic sensor, and a data processor;

[0025] The lidar is used to collect the road surface lidar data at each moment of a set section and transmit it to the data processor;

[0026] The distributed fiber optic sensor is buried under the roadbed of the set section and is used to collect the road surface mechanics data at each moment of the set section; wherein, the road surface mechanics data is the road surface stress when the vehicle passes by.

[0027] The data processor is configured to execute the steps in the pavement load spatial distribution positioning method for multi-source heterogeneous data fusion as described above.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] (1) Based on the road surface lidar data after removing the fitted ground, the present invention clusters and identifies the vehicle type and its position, dynamically tracks the vehicle trajectory, and then combines the vehicle position and road surface stress decoupled from the road surface mechanics data to generate a road surface stress and load distribution map in three-dimensional space. By combining lidar data and distributed fiber optic sensing data and performing mechanical analysis and decoupling processing, it makes up for the deficiencies of a single data source in identifying the vehicle load position and intensity, and significantly improves the identification accuracy.

[0030] (2) The present invention automatically learns the cube point density threshold, and by analyzing the density of the lidar scan data, compares the point density of each cube of the road surface lidar data with the dynamic threshold of the point density, ensuring the efficiency and accuracy of real-time data processing, excluding background noise points, reducing the interference of redundant data, and achieving the effect of accurately distinguishing the background and moving vehicles. Brief Description of the Drawings

[0031] The accompanying drawings forming a part of this invention are used to provide a further understanding of the invention. The schematic embodiments and descriptions thereof of the invention are used to explain the invention and shall not unduly limit the invention.

[0032] Figure 1 It is a flowchart of the method for locating the spatial distribution of pavement loads by multi-source heterogeneous data fusion according to an embodiment of the invention;

[0033] Figure 2 It is a schematic plan view of lidar point cloud data when processing lidar data according to an embodiment of the invention;

[0034] Figure 3 It is the working principle of the distributed optical fiber sensor used in an embodiment of the invention;

[0035] Figure 4 It is a schematic diagram of the sensor layout of the road structure cross-section according to an embodiment of the invention;

[0036] Figure 5 It is a schematic diagram of decoupling the spatial distribution of vehicle loads by mechanical derivation according to an embodiment of the invention. Detailed Embodiments

[0037] The present invention will be further described below in conjunction with the drawings and embodiments.

[0038] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0039] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0040] Embodiment 1

[0041] Figure 1 It is a flowchart of the method for locating the spatial distribution of pavement loads by multi-source heterogeneous data fusion according to an embodiment of the present invention. According to Figure 1 , a method for locating the spatial distribution of pavement loads by multi-source heterogeneous data fusion in this embodiment includes:

[0042] S100: Obtain the road surface lidar data and road surface mechanical data at each moment of the set section; among them, the road surface mechanical data is the road surface stress when the vehicle passes by;

[0043] S200: Fit the ground from the road surface lidar data and remove it. Based on the road surface lidar data after removing the fitted ground, cluster and identify the vehicle types and their positions, and dynamically track the vehicle trajectories;

[0044] S300: Use the finite element mechanics analysis method to decouple the vehicle position and the road surface stress from the road surface mechanical data, and then combine the tracked vehicle trajectories to generate the road surface stress and load distribution map in the three-dimensional space.

[0045] This embodiment uses the multi-source heterogeneous data fusion of the road surface lidar data and the road surface mechanical data, can sense the physical states of the road surface and the vehicle load, map the spatial distribution of the vehicle load, not only can dynamically feedback the operation state of the road surface structure, but also can deduce the service state of the road surface in real time, and provide accurate decision support.

[0046] In step S100, the road surface lidar data is measured by a lidar. Among them, the lidar is installed at a fixed position on the road, bridge or tunnel, such as the central green belt or the tunnel vault. In the road area with street lights, the street lights can be used as support poles, and the lidar can be installed on top of them to save costs. In the tunnel, the lidar can be fixed on the vault, and the distributed optical fiber sensors are buried along the bottom of the tunnel to achieve omnidirectional monitoring. According to the traffic environment, the number of frames of the 16-beam lidar can be set between 1500 and 3000 frames. If it is 32 lines or 64 lines, the number of frames can be reduced to ensure the efficiency and accuracy of real-time data processing.

[0047] The road surface lidar data contains a large number of three-dimensional coordinate points, and these points represent the surface of the vehicle's surrounding environment. Figure 2 The schematic diagram of the radar point cloud data plane in this embodiment when processing the lidar data is given.

[0048] The road surface mechanical data of this embodiment is detected by the distributed optical fiber sensors buried under the roadbed.

[0049] The distributed optical fiber sensors adopt the fiber Bragg grating technology, can sense the tiny strain and temperature changes. When the vehicle passes by, the distributed optical fiber sensors monitor the strain situation of the road surface in real time and record the strain data, and convert the strain data into an electrical signal through the optical reflection principle, as Figure 3 shown.

[0050] Figure 4The schematic diagram of the sensor layout of the road structure cross-section in this embodiment is given. A distributed optical fiber sensor is buried in the road surface to monitor physical changes such as strain and vibration in real time when a vehicle passes by, and detect the road surface load intensity. The data processor is responsible for analyzing the sensor signals, determining the load magnitude and its acting point. A Brillouin device is used to perform spectral analysis on the scattering signals at different positions of the optical fiber, and Gaussian fitting is used to determine the Brillouin frequency shift. :

[0051] ;

[0052] Among them, is the strain, is the temperature change, and are the sensitivity coefficients of strain and temperature respectively.

[0053] Preferably, through the dual-wavelength method, strain and temperature are measured respectively, and (strain) is decoupled.

[0054] According to the strain in the optical fiber, the load magnitude is estimated using mechanical formulas.

[0055] The relationship between strain and load is:

[0056] ; Among them, is the elastic modulus of the optical fiber material, is the cross-sectional area of the optical fiber.

[0057] Assume that the optical fiber is installed on the road surface where the vehicle passes. Thus, the estimated load can be approximately considered as part of the vehicle load. Through the strain distribution, the area with the maximum strain in the optical fiber is determined, and then the load acting point is determined.

[0058] In step S200, the ground segmentation algorithm RANSAC (Random Sample Consensus) can be used to fit the ground and remove it. In other embodiments, other existing segmentation algorithms can also be used to fit the ground and remove it.

[0059] In this embodiment, based on the road surface lidar data after removing the fitted ground, Euclidean clustering (EuclideanCluster Extraction) is used to cluster according to the distance between points. Each group of clustered points represents an object, and the vehicle type and its location are identified by clustering.

[0060] In the process of clustering to identify the vehicle type, according to the road surface lidar data after removing the fitted ground, a set section is divided into several consecutive cubes, and the point density of each cube is calculated; by comparing the point density of each cube with the dynamic threshold of the point density, the background and moving vehicles are distinguished.

[0061] When the point density of any cube is higher than the average value of the point densities of all cubes, the point density threshold is increased; where the adjusted point density threshold = α × the initial point density threshold; α > 1, which is a preset first adjustment coefficient.

[0062] When the point density of any cube is lower than the average value of the point densities of all cubes, and the point density variance of the current cube is less than the preset variance threshold, the point density threshold is decreased; where the adjusted point density threshold = β × the initial point density threshold; β < 1, which is a preset second adjustment coefficient.

[0063] By analyzing the change of the point density of each cube, the upper and lower limits of the point density threshold are adaptively adjusted based on the histogram.

[0064] By analyzing the local density change, the upper and lower limits of the threshold are dynamically adjusted to ensure an adaptive response to the density change in different regions and avoid excessive background noise interference. A common strategy is to use an adaptive threshold based on the histogram, that is, to perform a histogram statistics on the point density of each local region and select the inflection point in the histogram distribution as the new threshold.

[0065] In step S300, the process of decoupling the vehicle position and the road surface stress from the road surface mechanical data by using the finite element mechanics analysis method is the prior art.

[0066] In step S300, in the process of generating the road surface stress and load distribution map in the three-dimensional space, through time synchronization and spatial correlation, it is ensured that the road surface lidar data and the road surface mechanical data are aligned on the same time axis, and through the vehicle position information, the point cloud in the road surface lidar data is matched with the stress data in the road surface mechanical data. Figure 5 The schematic diagram of using mechanical derivation to decouple the vehicle load spatial distribution in the embodiment is given.

[0067] Specifically, a data fusion algorithm combining Kalman filtering and particle filtering is used to fuse the decoupled vehicle position and road surface stress with the tracked vehicle trajectory to generate a road surface stress and load distribution map in the three-dimensional space.

[0068] Let the system state be , including: vehicle position, vehicle speed and acceleration, and load distribution.

[0069] State transition equation:

[0070] ;

[0071] wherein, is the state transition function, is the system noise, assumed to be Gaussian distributed.

[0072] wherein, the state transition function describes calculating the state at the current moment from the state at the previous moment.

[0073] In the vehicle and road surface stress analysis, the state transition function is usually given by the vehicle's dynamic model or load propagation model. For the modeling of vehicle position, speed and acceleration, the vehicle's motion can be described by classical motion equations, for example:

[0074] ;

[0075] wherein, is the time interval; is the position of the vehicle at moment; is the position of the vehicle at moment; is the speed of the vehicle at moment ; is the speed of the vehicle at moment; is the acceleration of the vehicle at moment. For the load distribution, it may be affected by factors such as road surface type, vehicle speed, vehicle load mass, etc.

[0076] Assume that the position of the vehicle in the two-dimensional plane is , the position of the fixed object is , then the distance measured by the lidar can be obtained through the following observation function as follows:

[0077] ;

[0078] Assume that the load stress is proportional to the acceleration of the vehicle, and the distributed fiber optic sensor represents the observation equation in the following form:

[0079] ;

[0080] ;

[0081] wherein, is the observation state quantity matrix, including vehicle position, speed, acceleration and stress; is the proportionality coefficient, representing the relationship between acceleration and load stress, is the observation noise; and are the observation functions of lidar and distributed fiber optic sensors respectively, and are the corresponding noises.

[0082] Kalman filtering is suitable for processing linear, Gaussian systems and is used to estimate the dynamic information (position, speed, etc.) of vehicles. Its process includes two steps:

[0083] Prediction stage: Based on the state at the previous moment , use the state transition equation to predict the state at the current moment :

[0084] ;

[0085] wherein, is the state transition function;

[0086] Prediction error covariance :

[0087] ;

[0088] wherein, is the state transition matrix, is the covariance matrix of the process noise; represents the transpose of the matrix; represents the error covariance at the moment.

[0089] Use the observation value of the lidar for update:

[0090] ;

[0091] Update the state estimate :

[0092] ;

[0093] The updated error covariance :

[0094] ;

[0095] wherein, is the state estimate before update; is the Kalman gain, is the observation matrix, is the observation noise covariance.

[0096] Particle filtering is used to process non-linear and non-Gaussian systems and is applicable to the load distribution information fed back by fiber optic sensors. Particle filtering represents the probability distribution of the system state through multiple particles.

[0097] Initialize particles: Generate particles , where each particle represents a possible state, and the initial particles are generated based on the prior information of the load distribution.

[0098] Update particle weights: According to the observations of the distributed fiber optic sensor and the observation prediction corresponding to each particle , update the weight

[0099] of each particle:

[0100] where is the observation likelihood function of the given particle state, representing the matching degree between the observation data and the particle state; is the weight of each particle before update.

[0101] Resampling: To avoid the particle degeneracy phenomenon (i.e., the weights of most particles approach zero), it is necessary to resample the particle set. Particles with high weights are retained, and particles with low weights are eliminated to generate a new particle set.

[0102] State estimation: The weighted average of the particles is used to estimate the load distribution state at the current moment :

[0103] .

[0104] In this embodiment, Kalman filtering is responsible for the linear part (such as vehicle dynamics), and particle filtering processes the non-linear part (such as load distribution). The predicted output of Kalman filtering can provide prior information for particle filtering, and the result of particle filtering is used to further correct the estimation of Kalman filtering. Through the fusion of Kalman filtering and particle filtering, the system can obtain an accurate estimate of the vehicle load distribution and dynamically track the movement trajectory of the vehicle. These data are further used for the construction of the load distribution model.

[0105] Embodiment 2

[0106] In one or more embodiments, a pavement load spatial distribution positioning system for multi-source heterogeneous data fusion includes:

[0107] A multi-source heterogeneous data acquisition module, which is used to acquire the pavement lidar data and pavement mechanics data at each moment of a set section; wherein, the pavement mechanics data is the pavement stress when a vehicle passes by.

[0108] A vehicle recognition and tracking module, which is used to fit the ground from the road surface lidar data and remove it. Based on the road surface lidar data after removing the fitted ground, it clusters and identifies the vehicle type and its position, and dynamically tracks the vehicle trajectory;

[0109] A stress and load distribution map generation module, which uses the finite element mechanics analysis method to decouple the vehicle position and road surface stress from the road surface mechanics data, and then combines the tracked vehicle trajectory to generate a road surface stress and load distribution map in three-dimensional space.

[0110] In the vehicle recognition and tracking module, according to the road surface lidar data after removing the fitted ground, a set section is divided into several consecutive cubes, and the point density of each cube is calculated; by comparing the point density of each cube with the dynamic threshold of the point density, the background and moving vehicles are distinguished.

[0111] It should be noted here that each module in the road surface load spatial distribution positioning system with multi-source heterogeneous data fusion corresponds to each step in the road surface load spatial distribution positioning method with multi-source heterogeneous data fusion in the first embodiment above, and its specific implementation process is the same, so it will not be elaborated here.

[0112] Embodiment III

[0113] In this embodiment, a road surface load spatial distribution positioning system with multi-source heterogeneous data fusion is provided, which includes: a lidar, a distributed optical fiber sensor, and a data processor;

[0114] The lidar is used to collect the road surface lidar data at each moment of a set section and transmit it to the data processor;

[0115] The distributed optical fiber sensor is buried under the roadbed of the set section and is used to collect the road surface mechanics data at each moment of the set section; wherein, the road surface mechanics data is the road surface stress when a vehicle passes by;

[0116] The data processor is configured to execute the steps in the road surface load spatial distribution positioning method with multi-source heterogeneous data fusion as described in the first embodiment above.

[0117] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for locating the spatial distribution of road loads by fusion of multi-source heterogeneous data, characterized in that: include: Obtaining road surface laser radar data and road surface mechanical data at each time of a set road section; wherein the road surface mechanical data is the road surface stress when a vehicle passes by; Fit the ground from the road LiDAR data and remove it, cluster and identify the vehicle type and its location based on the road LiDAR data after removing the fitted ground, and dynamically track the trajectory of the vehicle; Finite element mechanics analysis is used to decouple vehicle position and pavement stress from pavement mechanics data, and then combined with tracked vehicle trajectories to generate pavement stress and load distribution maps in three-dimensional space; Based on the road surface lidar data after removing the fitted ground, the set road section is divided into several continuous cubes, and the point density of each cube is calculated; the point density of each cube is compared with the point density dynamic threshold to distinguish the background and the moving vehicle; The data fusion algorithm combining Kalman filter and particle filter is used to process the decoupled vehicle position, road stress and vehicle trajectory to generate the road stress and load distribution map in three-dimensional space; the process is as follows: Assume the system status is , including: vehicle position, vehicle speed and acceleration, and load distribution; State transfer equation: ; in, is the state transfer function, is the system noise, which is assumed to be Gaussian distributed; The state transfer function describes the state at the previous moment and calculates the current state; the state transfer function is given by the vehicle's dynamic model or load propagation model; for the modeling of vehicle position, velocity and acceleration, the vehicle's motion is described by the classical motion equation: ； in, is the time interval; The vehicle is The location at the moment; The vehicle is The location at the moment; The vehicle is at time speed; The vehicle is The speed of the moment; The vehicle is The acceleration of the moment; Assume that the position of the vehicle on the two-dimensional plane is , the position of the fixed object is , the distance measured by the lidar is obtained through the following observation function get: ； Assuming that the load stress is proportional to the acceleration of the vehicle, the observation equation of the distributed optical fiber sensor is expressed in the following form: ； ； in, is the observed state matrix, including vehicle position, velocity, acceleration and stress; is the proportionality coefficient, which expresses the relationship between acceleration and load stress, is the observation noise; and are the observation functions of LiDAR and distributed fiber optic sensor, and is the corresponding noise; Kalman filtering is used to estimate the dynamic information of the vehicle, and its process includes two steps: Prediction stage: based on the state of the previous moment , use the state transition equation to predict the current state : ; in, is the state transfer function; Forecast Error Covariance : ; in, is the state transition matrix, is the covariance matrix of the process noise; Represents the transpose of a matrix; express The error covariance at time Using LiDAR observations To update: ; Update state estimate : ; Updated error covariance : ; in, is the state estimate before the update; is the Kalman gain, is the observation matrix, is the observation noise covariance; Particle filtering is used to process the load distribution information fed back by the optical fiber sensor; particle filtering uses multiple particles to represent the probability distribution of the system state, and the process is as follows; Initialize particles: Generate Particles , each particle represents a possible state, and the initial particles are generated based on the prior information of the load distribution; Particle weight update: based on the observations of distributed fiber optic sensors The observation prediction corresponding to each particle , update the weight of each particle : ; in, is the observation likelihood function for a given particle state, indicating the degree of match between the observed data and the particle state; is the weight of each particle before updating; Resample the particle set; State estimation: The weighted average of particles is used to estimate the load distribution state at the current moment : ； The predicted output of the Kalman filter provides prior information for the particle filter, and the result of the particle filter is used to further correct the estimation of the Kalman filter; through the fusion of the Kalman filter and the particle filter, an accurate estimation of the vehicle load distribution is obtained, and the movement trajectory of the vehicle is dynamically tracked, which is further used for the construction of the load distribution model.

2. The method for spatial distribution positioning of road load by fusion of multi-source heterogeneous data as claimed in claim 1 is characterized in that: When the point density of any cube is higher than the average point density of all cubes, the point density threshold is increased; wherein, when the adjusted point density threshold = α×initial point density threshold; α>1, it is the preset first adjustment coefficient.

3. The method for spatial distribution positioning of road load by fusion of multi-source heterogeneous data as claimed in claim 1, characterized in that: When the point density of any cube is lower than the average point density of all cubes, and the point density variance of the current cube is less than the preset variance threshold, the point density threshold is lowered; wherein, the adjusted point density threshold = β × initial point density threshold; β<1, which is the preset second adjustment coefficient.

4. The method for spatial distribution positioning of road load by fusion of multi-source heterogeneous data as claimed in claim 1, characterized in that: By analyzing the changes in the point density of each cube, the upper and lower limits of the point density threshold are adaptively adjusted based on the histogram.

5. The method for spatial distribution positioning of road load by fusion of multi-source heterogeneous data as claimed in claim 1, characterized in that: In the process of generating the pavement stress and load distribution diagram in three-dimensional space, time synchronization and spatial association are used to ensure that the pavement lidar data and the pavement mechanics data are aligned on the same time axis, and the point cloud in the pavement lidar data is matched with the stress data in the pavement mechanics data through the vehicle's position information.

6. A road load spatial distribution positioning system based on multi-source heterogeneous data fusion, characterized in that: include: A multi-source heterogeneous data acquisition module, which is used to obtain road surface lidar data and road surface mechanical data at each time of a set road section; wherein the road surface mechanical data is the road surface stress when a vehicle passes by; The vehicle identification and tracking module is used to fit the ground from the road LiDAR data and remove it, cluster and identify the vehicle type and its location based on the road LiDAR data after removing the fitted ground, and dynamically track the trajectory of the vehicle; Stress and load distribution diagram generation module, which uses finite element mechanics analysis to decouple vehicle position and road stress from road mechanics data, and then combines the tracked vehicle trajectory to generate road stress and load distribution diagrams in three-dimensional space; In the vehicle identification and tracking module, the set road section is divided into a number of continuous cubes according to the road surface laser radar data after removing the fitted ground, and the point density of each cube is calculated; the point density of each cube is compared with the point density dynamic threshold to distinguish the background and the moving vehicle; The data fusion algorithm combining Kalman filter and particle filter is used to process the decoupled vehicle position, road stress and vehicle trajectory to generate the road stress and load distribution map in three-dimensional space; the process is as follows: Assume the system status is , including: vehicle position, vehicle speed and acceleration, and load distribution; State transfer equation: ; in, is the state transfer function, is the system noise, which is assumed to be Gaussian distributed; The state transfer function describes the state at the previous moment and calculates the current state; the state transfer function is given by the vehicle's dynamic model or load propagation model; for the modeling of vehicle position, velocity and acceleration, the vehicle's motion is described by the classical motion equation: ； in, is the time interval; The vehicle is The location at the moment; The vehicle is The location at the moment; The vehicle is at time speed; The vehicle is The speed of the moment; The vehicle is The acceleration of the moment; Assume that the position of the vehicle on the two-dimensional plane is , the position of the fixed object is , then the distance measured by the lidar is obtained through the following observation function get: ； Assuming that the load stress is proportional to the acceleration of the vehicle, the observation equation of the distributed optical fiber sensor is expressed in the following form: ； ； in, is the observed state matrix, including vehicle position, velocity, acceleration and stress; is the proportionality coefficient, which expresses the relationship between acceleration and load stress, is the observation noise; and are the observation functions of LiDAR and distributed fiber optic sensor, and is the corresponding noise; Kalman filtering is used to estimate the dynamic information of the vehicle, and its process includes two steps: Prediction stage: based on the state of the previous moment , use the state transition equation to predict the current state : ; in, is the state transfer function; Forecast Error Covariance : ; in, is the state transition matrix, is the covariance matrix of the process noise; Represents the transpose of a matrix; express The error covariance at time Using LiDAR observations To update: ; Update state estimate : ; Updated error covariance : ; in, is the state estimate before the update; is the Kalman gain, is the observation matrix, is the observation noise covariance; Particle filtering is used to process the load distribution information fed back by the optical fiber sensor; particle filtering uses multiple particles to represent the probability distribution of the system state, and the process is as follows; Initialize particles: Generate Particles , each particle represents a possible state, and the initial particles are generated based on the prior information of the load distribution; Particle weight update: based on the observations of distributed fiber optic sensors The observation prediction corresponding to each particle , update the weight of each particle : ; in, is the observation likelihood function for a given particle state, indicating the degree of match between the observed data and the particle state; is the weight of each particle before updating; Resample the particle set; State estimation: The weighted average of particles is used to estimate the load distribution state at the current moment : ； The predicted output of the Kalman filter provides prior information for the particle filter, and the result of the particle filter is used to further correct the estimation of the Kalman filter; through the fusion of the Kalman filter and the particle filter, an accurate estimation of the vehicle load distribution is obtained, and the movement trajectory of the vehicle is dynamically tracked, which is further used for the construction of the load distribution model.

7. A road load spatial distribution positioning system based on multi-source heterogeneous data fusion, characterized in that: include: LiDAR, distributed fiber optic sensors, and data processors; The laser radar is used to collect road surface laser radar data at each time of a set road section and transmit it to a data processor; The distributed optical fiber sensor is buried under the roadbed of a set road section and is used to collect pavement mechanical data of the set road section at various times; wherein the pavement mechanical data is the pavement stress when a vehicle passes by; The data processor is configured to execute the steps in the method for locating the spatial distribution of road load by fusing multi-source heterogeneous data as described in any one of claims 1-5.

Images