Emergency lane flatness monitoring and early warning system and control method thereof

Abstract

The invention provides an emergency lane flatness monitoring and early warning system and a control method thereof. The system comprises a road surface flatness module, a cloud platform data analysis module, a user interaction interface module and a dynamic safety early warning module which are connected in sequence. Firstly, the system carries out initialization and self-inspection, and then enters a vehicle monitoring stage. Then, the system scans the road surface, and uploads and analyzes the collected data. And according to a data analysis result, the system formulates a security and maintenance strategy and updates user information at the same time. In a construction area, the system can give out safety early warning, the system is continuously monitored to ensure normal operation of the system, and the early warning system is reset. The performance of the emergency lane flatness monitoring and early warning system is remarkably improved through a plane optimization algorithm, the safety management process is optimized, and efficient and accurate data support is provided for road maintenance.

Description

Technical Field

[0001] The invention belongs to the technical field of road safety, and in particular relates to a risk avoidance lane flatness monitoring and early warning system and a control method thereof. Background Art

[0002] The flatness of the escape lane is particularly critical for preventing traffic accidents. If the escape lane is not leveled after use or after a period of time, the aggregate in the brake slope is uneven. When a vehicle with brake failure rushes into the escape lane, the forces on both sides of the axle are uneven, and the overturning moment borne by the vehicle is greater than the stabilizing moment, causing the vehicle to roll over, which may cause casualties in severe cases. In the traditional maintenance and management of the escape lane, the monitoring of flatness relies on manually operated 3-meter rulers and precision levels. This method is not only labor-intensive, but also slow in detection speed, and it is difficult to adapt to the requirements of modern road traffic for immediacy and accuracy. Due to the limitations of manual detection, it can often only cover some key areas, and it is difficult to achieve continuous and fine monitoring of the flatness of the entire road section, resulting in some potential uneven areas not being discovered and repaired in time, and unable to achieve rapid warning, which increases the risk of vehicles using the escape lane in an emergency.

[0003] In addition, traditional detection methods are limited by weather and light conditions. It is difficult to carry out work in bad weather or at night, which affects the continuity and accuracy of the safety warning of the safe lane. At the same time, the subjective judgment factors of manual detection will also affect the consistency and reliability of the data. There may be differences in judgment standards between different inspectors, which further reduces the accuracy of the safety warning of the safe lane. Summary of the invention

[0004] The purpose of the embodiment of the present invention is to provide a safe lane flatness monitoring and early warning system to solve the limitations of manual monitoring of safe lane flatness and improve the efficiency and accuracy of safe lane safety early warning;

[0005] A second objective of an embodiment of the present invention is to provide a control method for a dangerous lane flatness monitoring and warning system.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is a risk avoidance lane flatness monitoring and early warning system, comprising a road surface flatness module, a cloud platform data analysis module, a user interaction interface module, and a dynamic safety early warning module connected in sequence; wherein the dynamic safety early warning module is also connected to the road surface flatness module signal;

[0007] The road surface flatness module is used to automatically collect and pre-process the road surface point cloud data of the safe lane, and upload the pre-processed data to the cloud platform data analysis module through data transmission;

[0008] The cloud platform data analysis module includes a vehicle detection sensor, a ground three-dimensional laser scanner, and a data transmission interface. The vehicle detection sensor is connected to the ground three-dimensional laser scanner by signal, the ground three-dimensional laser scanner is connected to the data transmission interface by signal, and the data transmission interface is connected to the cloud platform data analysis module through Internet data.

[0009] Furthermore, the cloud platform data analysis module is installed in a computer with storage and data processing functions, including

[0010] A data storage unit, used for securely storing and backing up the pre-processed point cloud data;

[0011] The data processing unit performs data filtering on the pre-processed point cloud data, and after the data filtering is completed, determines the flatness index of the road surface of the danger avoidance lane;

[0012] A standard comparison unit is used to compare and analyze the actual road surface roughness index with the improved safe lane roughness detection standard;

[0013] The maintenance decision unit is used to determine whether repaving is necessary based on the data comparison and analysis results.

[0014] Furthermore, the process of digital filtering performed by the data processing unit is specifically as follows:

[0015] S1, slope-based filtering;

[0016] S101. Use the slope calculation formula to calculate the slope S of each pair of adjacent points in the point cloud:

[0017]

[0018] Among them, x1, x2, y1, y2, z1, z2 represent the three-dimensional coordinates of two adjacent sampling points;

[0019] S102, preliminary filtering: adaptive parameter preset, eliminating non-ground points with a slope greater than a preset threshold;

[0020] S2, refinement of triangulated network based on cloth model optimization;

[0021] S201, constructing the point cloud data after preliminary filtering into an irregular triangulated network model by using Delaunay triangulation;

[0022] S202, simulating a virtual cloth model, and covering the virtual cloth on the irregular triangulated network model constructed in S201; wherein the elastic characteristics of the virtual cloth are:

[0023] F=k×(Δx-x0)

[0024] F is the elastic force, k is the spring constant, Δx is the deformation of the spring, and x0 is the natural length of the spring without force;

[0025] S203, simulating the sagging process of the cloth, using gravity to simulate the natural compliance of the cloth to the terrain, to identify the non-ground points that are misidentified. This process updates the position through Verlet integration:

[0026]

[0027] x(t+Δt) represents the position of the cloth at time t+Δt, v(t) represents the velocity of the cloth at time t, a represents the acceleration of the cloth at time t, Δt is the time interval, and x(t) represents the position of the cloth at time t;

[0028] S204, integrating the virtual cloth model with the TIN model: the cloth updates and adjusts the TIN grid according to the final state of the drooping process, and the ground point cloud data is optimized according to the result of the cloth drooping; the optimization objective function is:

[0029]

[0030] Among them, p i Indicates the position of the i-th vertex in the TIN model. It is a three-dimensional vector containing the X, Y, and Z coordinates of the vertex. i Represents the relationship between the i-th TIN vertex p i The three-dimensional coordinates of the corresponding point on the cloth model. The position of this point is in the process of cloth drooping, and it is consistent with the vertex p of the TIN model. i The closest point; i represents the sequence number of the vertex in the TIN model, and n represents the total number of vertices in the TIN model;

[0031] S205, using the cloth simulation results, re-screening and refining the ground point cloud, and removing those non-ground points that fail to pass the cloth simulation verification;

[0032] S206. Reconstruct the TIN model using the finally screened and adjusted ground point cloud data.

[0033] Furthermore, the step of determining the roughness index of the safe lane by the data processing unit is specifically as follows:

[0034] S301, based on the RANSAC algorithm, randomly sample the digitally filtered point cloud data, increase the sampling probability of the historical model neighboring points during the sampling process; use the weighted average method to integrate the current data point p c and historical data point p h , forming a new comprehensive point p n , p n =w×pc +(1-w)×p h , where w is the weight based on the similarity metric;

[0035] S302, adjusting the RANSAC algorithm parameters, using the least squares method, combined with the improved RANSAC algorithm to iterate to find the most suitable plane parameters so that the plane equation is most suitable for the current point cloud data set;

[0036] S303, according to the point cloud data characteristics extracted by the improved RANSAC algorithm, using the law of XY coordinate projection, extracting the road boundary in the shape of a straight line or a regular curve; for the straight line in the extracted road boundary line data, using the slope distance straight line formula to fit, for the curve, using a high-order polynomial to fit;

[0037] S304, extracting the road point cloud data using the fitted boundary line to extract the complete road point cloud data, and removing noise from the boundary noise data using the least squares plane fitting method;

[0038] S305. Maintain and optimize the historical data set, including regularly filtering out data that is no longer applicable and updating new successfully fitted model data, to ensure the timeliness and applicability of the historical data set, and substitute the extracted valid data into the predetermined international roughness index calculation formula to obtain the international roughness index actually measured on the road surface.

[0039] Furthermore, the adjustment of the RANSAC algorithm parameters in S302 is specifically as follows:

[0040] S3021. Establish a standard reference image library;

[0041] S3022, using the video surveillance at the entrance of the escape lane to capture the image of the vehicle exiting, and importing the image at this moment into the Matlab processing module;

[0042] S3023, evaluating the clarity of the real-time image;

[0043] S3024, using a similarity metric to compare the weighted average of the input image with the representative values ​​of each level in the database, find the clarity level with the smallest difference, and associate the input image with the level;

[0044] S3025. Transmit the numerical value corresponding to the real-time image clarity to the RANSAC algorithm calculation module to complete the adjustment of the RANSAC algorithm.

[0045] Furthermore, the user interaction interface module is used to provide real-time information to maintenance personnel and drivers, including a maintenance personnel mobile app and a driving personnel information system.

[0046] Furthermore, the dynamic safety warning module is used to ensure the safety of construction workers and issue road condition warnings to drivers in a timely manner, including facial recognition monitoring, a construction safety management system, and a warning information release system; the facial recognition monitoring is installed at the entrance of the safe lane and is signal-connected to the construction safety management system and the warning information release system; the construction safety management system is installed in a computer with a storage medium and a processor; the warning information release system is installed at the entrance of the safe lane and is signal-connected to the construction safety management system, including a display screen and a horn alarm system.

[0047] A control method for a safe lane flatness monitoring and early warning system comprises the following steps:

[0048] Step 1: After the system is started, it will conduct self-test, including the status test of laser scanner, vehicle detection sensor, data transmission interface, cloud platform connection and dynamic safety warning module;

[0049] Step 2: Use the vehicle detection sensor at the intersection to monitor in real time whether there are vehicles passing through the safe lane;

[0050] Step 3: When the sensor detects that the vehicle has exited the safe lane and no vehicle is detected, the ground 3D laser scanner is started to perform a comprehensive scan of the safe lane to generate 3D point cloud data of the road surface;

[0051] Step 4: The ground 3D laser scanner sends the pre-processed data to the data transmission interface, which is encoded by the module and uploaded to the cloud platform data analysis module; the data processing unit of the cloud platform data analysis module filters and analyzes the uploaded data and compares it with the standard data;

[0052] Step 5: According to the analysis results of the cloud platform data analysis module standard comparison unit, determine whether the road surface is safe or needs maintenance;

[0053] Step 6: Maintenance personnel receive maintenance task notifications sent via the mobile app, confirm and prepare for construction; the driver information system updates the status of the safe lane, providing real-time road conditions and safety warnings for passing vehicles;

[0054] Step 7. When the cloud platform data analysis module determines that the road needs maintenance, the facial recognition monitoring and warning information release system is automatically activated. The maintenance personnel perform facial recognition punch-in when starting maintenance, publish information about construction hazards to the display screen of the warning information release system, and activate the horn system when the vehicle enters the safe lane. The maintenance personnel perform facial recognition punch-in when ending maintenance.

[0055] Furthermore, the step 3 of scanning the road surface with a ground-based three-dimensional laser scanner comprises the following steps:

[0056] Step 301, the ground three-dimensional laser scanner is automatically started to perform a three-dimensional scan of the danger lane;

[0057] Step 302, the ground 3D laser scanner records the 3D position information of the road surface by measuring the time delay and intensity reflection of the laser; the data is recorded in the form of a point cloud, and each point represents a specific coordinate position on the road surface;

[0058] Step 303, the ground three-dimensional laser scanner generates a preliminary analysis map of the road surface in real time to detect possible uneven road surfaces;

[0059] Step 304, after the preliminary analysis is completed, the road surface flatness detection module performs preprocessing of the point cloud data, including point cloud denoising, point cloud registration and point cloud sampling;

[0060] Step 305, the pre-processed point cloud data is uploaded to the cloud platform data analysis module through the data transmission interface after encryption protection;

[0061] Step 306. After completing the scanning of the entire road, the ground 3D laser scanner automatically turns off and enters the standby state, waiting for the next scanning cycle or system command.

[0062] The beneficial effects of the present invention are as follows: the present invention can instantly detect potential safety hazards and issue warnings through the safe lane flatness monitoring and early warning system to prevent traffic accidents. Compared with manual detection, the present invention improves detection efficiency and data reliability, and automatically processes data through a cloud platform to optimize the safety management process. In addition, the present invention provides a real-time information feedback channel and has a modular design, which can be flexibly adjusted according to different needs to meet a variety of application scenarios.

[0063] The present invention also proposes an innovative point cloud data filtering algorithm, which combines slope-based filtering technology with triangulated network (TIN)-based optimization processing. The two-stage processing method significantly improves the accuracy of ground point screening and can adapt to various complex terrain conditions. In the first stage, the non-ground points are preliminarily filtered by adaptively adjusting the slope threshold, ensuring the flexibility and versatility of the algorithm; in the second stage, Delaunay triangulation technology and virtual cloth simulation are used to optimize point cloud data, further improving the precision and accuracy of terrain modeling. This unique filtering algorithm can efficiently and accurately extract ground point cloud data and construct a high-precision digital elevation model (DEM), providing reliable basic data for road flatness assessment, thereby significantly improving the performance of the safe lane flatness monitoring and early warning system, and providing strong technical support for road safety management and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 Schematic diagram of the system structure of this embodiment;

[0066] Figure 2 is a schematic diagram of a road surface flatness detection module in this embodiment;

[0067] Figure 3 This is a schematic diagram of the cloud platform data analysis module in this implementation mode;

[0068] Figure 4 is a schematic diagram of a user interaction interface module of this embodiment;

[0069] Figure 5 is a schematic diagram of a dynamic safety warning module in this implementation mode;

[0070] Figure 6 This is a schematic diagram of equipment installation in this embodiment;

[0071] Figure 7 It is a flow chart of the control method steps of this embodiment. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] like Figure 1 This embodiment provides a safe lane flatness monitoring and early warning system, which mainly includes a road surface flatness module, a cloud platform data analysis module, a user interaction interface module, and a dynamic safety early warning module, which are connected in sequence, wherein the dynamic safety early warning module is also connected to the road surface flatness module.

[0074] In some specific implementations, the road surface flatness detection module is a core component of the road surface flatness monitoring and safety warning system for the safe lane, and is used to automatically collect road surface data of the safe lane, and upload the pre-processed data to the cloud platform through data transmission. In this implementation, the road surface flatness detection module includes a vehicle detection sensor, a ground-based three-dimensional laser scanner, and a data transmission interface, wherein the vehicle detection sensor is connected to the ground-based three-dimensional laser scanner signal, the ground-based three-dimensional laser scanner is connected to the data transmission interface signal, and the data transmission interface is connected to the cloud platform data analysis module through the Internet.

[0075] In some specific embodiments, the vehicle detection sensor is first responsible for identifying whether there is a vehicle on the emergency lane. When a vehicle is detected, the sensor generates a trigger signal. This trigger signal is transmitted to the ground-based 3D laser scanner via an electrical signal. This signal tells the scanner when to start scanning the road surface flatness. After receiving the trigger signal from the vehicle detection sensor, the ground-based 3D laser scanner begins to scan the road surface and collect data. The data collected by the scanner is transmitted to the data transmission interface via a Wi-Fi wireless connection. The data transmission interface acts as a data relay station and sends the scanned data to the cloud platform for analysis.

[0076] Example 1

[0077] like Figure 2 and Figure 6 This embodiment provides a specific road surface flatness detection module, in which a vehicle monitoring sensor is set at the entrance of the escape lane to detect the entry and exit status of the vehicle; the vehicle detection sensor in this embodiment is based on the working principle of video image processing. The vehicle detection sensor can detect whether the same vehicle leaves the escape lane, and after the vehicle leaves, it sends a signal to the ground three-dimensional laser scanner to trigger the road scanning program. By detecting the entry and exit of the vehicle, it can help the system determine whether the next step needs to be carried out.

[0078] Based on the research on rollover accidents caused by uneven road surface in the past and the setting principle of escape lane, it is found that the brake slope of the escape lane is the key point, so the ground three-dimensional laser scanner is set in the middle of the brake slope of the escape lane; the ground three-dimensional laser scanner is used to automatically scan the escape lane and generate high-precision three-dimensional point cloud data; in this embodiment, based on the working principle of laser ranging, the spatial position information of the scanned object, that is, point cloud data, is obtained, and the massive point cloud data is pre-processed and transmitted to the cloud platform through data for the next step of data analysis; in this embodiment, the form of preprocessing includes: moving least squares method, by fitting the least squares plane for each point and the points in its neighborhood, and then adjusting the position of the point according to the fitting result to achieve the effect of smoothing the data; statistical denoising, noise point removal based on the distance distribution of the neighboring points of each point. For each point, the average distance to the neighboring points is calculated and compared with the global average distance. If the average distance of a point is much larger than the global average (usually using the standard deviation as the threshold), the point is considered to be a noise point and is removed; Gaussian filtering denoising, applying the Gaussian model to smooth the point cloud to reduce the impact of noise on the data. Usually, the position of each point is Gaussian weighted average, and the weight is determined by the distance between the points. By collecting road point cloud data, the original measured data can be provided for the subsequent calculation of international flatness.

[0079] The data transmission interface is used to automatically upload the pre-processed data of the ground 3D laser scanner to the cloud platform. The data transmission interface of this embodiment can select suitable wired technology and wireless technology according to actual needs, as well as the interface protocol with the cloud platform. The stability, security and compatibility of the communication interface need to be considered in the selection of the data transmission interface to ensure the reliability of data transmission and the confidentiality of the data.

[0080] In a specific implementation, the cloud platform data analysis module is used to store and process the measured data, and compare it with the improved safe lane flatness detection standard IRI to determine whether road maintenance is needed. The cloud platform data analysis module in this implementation includes a data storage unit, a data processing unit, a standard comparison unit, and a maintenance decision unit.

[0081] In some specific implementations, the cloud platform data analysis module is mounted on a server system device with computing and storage capabilities, and the server system device is connected to the data transmission interface and user interaction interface module via the Internet.

[0082] Example 2

[0083] like Figure 3, this embodiment provides a specific cloud platform data analysis module, which is installed in a computer with storage and data processing functions, including a data storage unit, a data processing unit, a standard comparison unit and a maintenance decision unit, wherein the data storage unit is used to safely store and back up the pre-processed point cloud data. The data storage unit of this embodiment saves data in a scalable, efficient and reliable manner, supports high-performance data acquisition and processing, and thus provides basic services for large-scale data analysis.

[0084] The data processing unit is used to clean, organize and analyze the pre-processed data through a set data processing algorithm; in this embodiment, the data processing unit uses a combined filtering algorithm to perform ground filtering on the pre-processed point cloud data. After the data filtering is completed, the improved RANSAC algorithm is used, combined with the historical data fusion technology to determine the flatness index of the road surface of the safe lane, so as to improve the accuracy of extracting road point cloud data in rainy and foggy weather in mountainous areas.

[0085] The standard comparison unit is used to compare and analyze the actual road surface roughness index with the improved safe lane roughness detection standard IRI. The standard comparison unit of this embodiment first selects the IRI value of other asphalt highway pavements of 4.2m / km as a reference benchmark, and sets a safety bonus factor considering the traffic pressure and material characteristics that safe lanes usually bear, and then scientifically adjusts it to reduce the IRI upper limit to below 3.5m / km, so as to put forward higher requirements for the road surface responsiveness and stability in emergency braking or safe lane situations. It can compare and analyze the actual calculated road surface roughness index according to the set threshold, and transmit the comparison and analysis results to the maintenance decision unit.

[0086] The maintenance decision unit is used to determine whether re-laying is necessary based on the data comparison and analysis results. When the comparison and analysis finds an abnormal situation, the maintenance decision unit will trigger a maintenance notification to notify the maintenance personnel to take corresponding measures.

[0087] Example 3

[0088] This embodiment provides a specific form of a data processing unit combined filtering algorithm. The algorithm described in this embodiment includes two stages:

[0089] First stage, slope-based filtering:

[0090] S101. Use the slope calculation formula to calculate the slope S of each pair of adjacent points in the point cloud:

[0091]

[0092] Among them, x1, x2, y1, y2, z1, z2 represent the coordinates of two adjacent sampling points;

[0093] S102, preliminary filtering: adaptive parameter preset, remove non-ground points with a slope greater than a preset threshold. Automatically adjust the slope threshold parameter according to the conditions such as the aggregate type of the escape lane. For general gravel roads or unpaved escape lanes, the slope threshold is between 3° and 5°, and for escape lanes paved with large stones or relatively rugged, the slope threshold is between 5° and 8°.

[0094] In the second stage, the refined processing of the triangulated network is optimized based on the cloth model:

[0095] S201, establishing a triangulated network: constructing the point cloud data after preliminary filtering into an irregular triangulated network (TIN) using Delaunay triangulation;

[0096] S202, cloth simulation initialization: Cover the virtual cloth on the TIN model constructed in S201. This process uses Hooke's law to simulate the elastic characteristics of the cloth:

[0097] F=k×(Δx-x0)

[0098] F is the elastic force, k is the spring constant, Δx is the deformation of the spring, and x0 is the natural length of the spring without force;

[0099] S203, simulate the process of cloth drooping, use gravity to simulate the natural compliance of cloth to the terrain, and identify the non-ground points that are misidentified. This process updates the position through Verlet integration:

[0100]

[0101] x(t+Δt) represents the position of the cloth at time t+Δt, that is, the new position of each point on the cloth after a time interval Δt; v(t) represents the speed of the cloth at time t, a represents the acceleration of the cloth at time t, Δt is the time interval, x(t) represents the position of the cloth at time t, that is, the current position of each point on the cloth, which is the starting point for position update calculation;

[0102] By simulating the drooping process of cloth, the cloth will only cover the ground points, while the points above the cloth can be identified as non-ground points. By taking advantage of the flexibility and gravity of virtual cloth, a natural fit between cloth and terrain can be achieved through physical simulation. The drooping process of cloth simulates the influence of gravity on cloth, and the cloth's layout position is naturally adjusted according to the undulations of the terrain, which can depict the general outline and characteristics of the terrain. This physical simulation method can effectively simulate physical behavior in the real world and provide an intuitive way to process and analyze terrain data.

[0103] S204, integrating the cloth model with the TIN model: the cloth updates and adjusts the TIN grid according to the final state of the drooping process, and the ground point cloud data is optimized according to the result of the cloth drooping; the optimization objective function is:

[0104]

[0105] Among them, p i Indicates the position of the i-th vertex in the TIN model. It is a three-dimensional vector containing the X, Y, and Z coordinates of the vertex; q i Represents the relationship between the i-th TIN vertex p i The three-dimensional coordinates of the corresponding point on the cloth model. The position of this point is in the process of cloth drooping, and it is consistent with the vertex p of the TIN model. i The closest point; i represents the serial number of the vertex in the TIN model, i = 1, 2, 3, …, n; n represents the total number of vertices in the TIN model.

[0106] The combination of cloth model and TIN model is essentially to use the softness of cloth to obtain ground detail information. By letting the cloth sag in the shape of the terrain, the cloth shape is finally obtained, and then the least squares method is used to transfer the cloth shape information to the TIN model to adjust its vertex position, thereby building a more realistic terrain model.

[0107] S205, ground point cloud refinement: using the results of cloth simulation, re-screen and refine the ground point cloud, and remove those non-ground points that fail to pass the cloth simulation verification.

[0108] S206, final TIN optimization: Reconstruct the TIN model using the final screened and adjusted ground point cloud data.

[0109] The optimization process of the cloth model can redistribute the triangular mesh points according to the needs of the terrain features. In areas with large terrain changes, the mesh can become denser to accurately capture the terrain changes. In areas with relatively flat terrain, the mesh can be relatively sparse, which not only ensures the accuracy of the model, but also optimizes the scale of the data and ensures computational efficiency. The optimization algorithm based on the cloth model utilizes the principle of physical simulation, gradually reduces the energy function through an iterative process, and makes the entire model stable. This method can effectively avoid the problem of local optimal solutions that may occur in traditional optimization algorithms, thereby improving the overall convergence and stability of the algorithm.

[0110] Example 4

[0111] This embodiment provides a specific method for a data processing unit to determine the roughness index of a dangerous lane.

[0112] S301, data sampling and historical data fusion: Randomly sample the filtered point cloud data, and increase the sampling probability of the historical model neighboring points during the sampling process; this embodiment will increase the sampling probability of the historical model neighboring points during the sampling process based on the point cloud data characteristics extracted by the RANSAC algorithm to improve the efficiency of the RANSAC algorithm in fitting the plane. Use the weighted average method to integrate the current data point p c and historical data point p h , forming a new comprehensive point p n , p n =w×p c +(1-w)×p h , where w is a weight based on similarity measurement, which can be dynamically adjusted according to the similarity between the currently collected data conditions and historical data;

[0113] S302, adjusting the RANSAC algorithm parameters, and using the improved RANSAC algorithm for plane fitting: using the least squares method, combined with the improved RANSAC algorithm for iteration to find the most suitable plane parameters a, b, c, d so that the plane equation ax+by+cz+d=0 (this equation is used to describe the mathematical expression of all points in the current point cloud data set relative to the fitting plane. By determining the most suitable plane parameters to minimize the distance between the point cloud data and the fitting plane, it is ensured that the found plane is the equation that best represents the data set) is most suitable for the current point cloud data set, where a, b, c are the normal vector parameters of the plane, representing the direction of the plane normal vector, d is the distance between the plane and the origin, and x, y, z are the coordinates of the three-dimensional coordinate system;

[0114] S303, extract road boundaries and fit boundary lines: according to the characteristics of point cloud data extracted by the improved RANSAC algorithm, the road boundaries in the shape of straight lines or regular curves are extracted by using the law of XY coordinate projection. For straight lines in the extracted road boundary line data, the slope distance linear formula is used for fitting, and for curves, a high-order polynomial is used for fitting;

[0115] S304, extracting road point cloud data: extracting complete road point cloud data using the fitted boundary line, and denoising the noise data of the boundary part using the least square plane fitting;

[0116] S305, historical data optimization, determination of smoothness index: After completing a road extraction process, the historical data set is maintained and optimized, including regular filtering of no longer applicable data and updating of new successfully fitted model data to ensure the timeliness and applicability of the historical data set. Finally, the extracted valid data is substituted into the predetermined international smoothness index calculation formula to obtain the international smoothness index of the road surface measured.

[0117] Example 5

[0118] This embodiment uses the video image data captured in real time to evaluate the weather conditions in the environment by dividing the clarity of the image, and automatically classifies the current weather conditions according to the preset clarity level standard. Finally, the relevant parameters in the RANSAC algorithm are dynamically adjusted according to the clarity of the image under different weather conditions, as follows:

[0119] S3021. Establish a standard reference image library: Each weather condition has a corresponding standard image with a clarity level. The clarity of a sunny day image is 4, the clarity of light rain and fog weather is 3, the clarity of moderate rain and fog weather is 2, and the clarity of heavy rain and fog weather is 1.

[0120] S3022, image acquisition: use the video surveillance at the entrance of the escape lane to capture the image of the vehicle exiting, and import the image at this moment into the Matlab processing module.

[0121] S3023, clarity calculation: Use the standard reference image library to evaluate the clarity method to evaluate the clarity of the real-time image. The specific evaluation method is as follows: First, the image edge information is obtained through the Canny algorithm, and then the average intensity of the edge pixels is calculated through the image edge information, and then the image contrast is calculated to obtain the mean of the histogram contrast, and then the weighted average of the average intensity of the edge pixels and the mean of the contrast is calculated.

[0122] Specifically, to calculate the average intensity of edge pixels, the Canny algorithm is first used for edge detection. First, the image is converted into a grayscale image, and then the "gradient" technology is used to find areas with drastic brightness changes on the entire image. These areas are usually the edges of objects. After being processed by the Canny algorithm, a black and white image is obtained, in which white pixels represent edges. The "intensity" of the edge is obtained by looking at the brightness of these white pixels in the original grayscale image. The larger the brightness value, the greater the intensity of this edge pixel. In order to obtain the global edge intensity, the intensity of all edge pixels is averaged. Specifically, the brightness values ​​of all edge pixels are added up and then divided by the total number of edge pixels (that is, the number of all white pixels).

[0123] S3024, determining clarity: using a similarity metric to compare the weighted average of the input image with the representative values ​​of each level in the database, finding the clarity level with the smallest difference, and associating the input image with that level.

[0124] S3025. Transmit the numerical value corresponding to the real-time image clarity to the RANSAC algorithm calculation module to complete the adjustment of the RANSAC algorithm.

[0125] In some specific implementations, the weight "w" based on the similarity metric in S301 is adjusted as follows:

[0126] w=w j + Image clarity value × adjustment factor

[0127] w j The base weight value is 0.6, and the adjustment coefficient is 0.1; the sampling probability of the neighboring points of the historical model is: p = 1-w;

[0128] In some specific implementations, the method for determining the number of iterations in S302 is specifically as follows: the maximum number of iterations when the image clarity is 4 is set to 100 times, and the maximum number of iterations in other weather conditions can be determined according to the number of iterations m=100×(1+p);

[0129] In some specific implementations, for weather conditions where the contour is not clear enough in S303, the smoothness of the curve fitting is increased by adjusting the number of control points of the B-spline curve, where the number of control points Q=q×w, where q is the image clarity.

[0130] In some specific embodiments, the user interaction interface module is used to provide real-time information to maintenance personnel and drivers, including a maintenance personnel mobile app and a driving personnel information system;

[0131] The maintenance personnel mobile app is usually set up on the maintenance personnel's mobile devices, such as smartphones or tablets. Maintenance personnel can directly manage tasks and receive relevant notifications through the maintenance personnel mobile app. Driver information system: Directly interact with the vehicle's environmental perception system through vehicle networking technology. The system receives the danger lane status information sent by the cloud platform data analysis module through the vehicle's wireless network. Then this information will be displayed in real time on the driver's navigation system for the driver's reference. The user interaction interface module relies on the wireless network to interact with other modules:

[0132] Interact with the road surface smoothness module: The road surface smoothness module detects the road status and uploads data, and the maintenance personnel mobile app and the vehicle information system display corresponding information based on this data, such as whether the safe lane is occupied by a vehicle; Interact with the cloud platform data analysis module: The cloud platform data analysis module analyzes the received data and transmits the results to the maintenance personnel mobile app and the vehicle information system through the wireless network, such as whether the safe lane needs maintenance; Interact with the dynamic safety warning module: For maintenance personnel who have completed the maintenance task, when they perform face recognition and clock in, the dynamic safety warning module can feedback the task completion information to the maintenance personnel mobile app, so that the maintenance personnel can understand their task status. For the driver information system, when the dynamic safety warning module activates the warning information release system, it can push construction or danger warning information to the driver information system in real time, so that passing drivers can make judgments in advance based on the information on the display screen and avoid entering the construction area.

[0133] Example 6

[0134] like Figure 4 , this embodiment provides a specific user interaction interface module, in which the maintenance personnel mobile app is used for maintenance personnel to receive maintenance notifications and manage maintenance tasks through mobile devices. The maintenance personnel mobile app of this embodiment includes a task management module, which enables maintenance personnel to view maintenance tasks, road conditions, historical data analysis results in real time, and receive and confirm maintenance task notifications; the maintenance personnel mobile app also allows maintenance personnel to update construction status and safety notifications to ensure the accuracy and timeliness of information.

[0135] The vehicle driver information system is used to provide the driver with real-time information about the road status at an appropriate location. The vehicle driver information system of this embodiment uses the vehicle networking technology to directly transmit information to vehicles passing through the area, ensuring that the driver can obtain the most immediate road safety information.

[0136] In some specific implementations, the dynamic safety warning module is used to ensure the safety of construction workers and issue road condition warnings to drivers in a timely manner, including face recognition monitoring, construction safety management system, and warning information release system.

[0137] Example 7

[0138] like Figure 5 and Figure 6This embodiment provides a specific dynamic safety warning module, in which face recognition monitoring is installed at the entrance of the safe lane, and is connected to the construction safety management system and the warning information release system signal to identify construction personnel entering the safe lane. The face recognition monitoring installation uses image capture, facial detection, feature extraction, feature comparison and recognition verification to identify and verify in real time whether the scanned person is a registered maintenance personnel. If the person is a maintenance personnel, he / she will automatically sign in and start the warning information release system. Otherwise, the system will be silent. If the same maintenance personnel scans his / her face again after an interval of more than ten minutes, the system will consider the maintenance task completed and provide real-time feedback.

[0139] In some possible implementations, the construction safety management system is installed in a computer with a storage medium and a processor, and is used to detect and alert construction personnel and vehicles. The construction safety management system is used in conjunction with the warning information release system to achieve the purpose of ensuring the safety of maintenance personnel.

[0140] The warning information release system is installed at the entrance of the escape lane and is connected to the construction safety management system signal. It includes a display screen and a horn alarm system for real-time release of construction or danger warning information. When maintenance personnel start maintenance, they punch in with facial recognition to activate the early warning system, release information about construction hazards to the display screen to prevent vehicles from entering the construction area, and activate the horn system when vehicles enter the escape lane to urge maintenance personnel to leave the escape lane. When maintenance personnel finish maintenance, they punch in with facial recognition, and the warning information release system will return to normal mode and realize information feedback.

[0141] In some specific implementations, the dynamic safety warning module transmits the collected data to the cloud platform data analysis module for analysis. The cloud platform data analysis module processes the data from the road surface flatness detection module, generates analysis results, and then feeds them back to the dynamic safety warning module to provide real-time data support.

[0142] In some specific implementations, the dynamic safety warning module also receives the road surface roughness processed by the cloud platform data analysis module.

[0143] In some specific implementations, the workflow of the dynamic safety warning module is:

[0144] Step 1: Start the warning information release system, face recognition monitoring and construction safety management system installed on the computer;

[0145] Step 2: After the facial recognition monitoring detects that the maintenance personnel have arrived at the site, the construction safety management system issues a command to activate the display screen of the warning information release system; the construction safety management system receives the vehicle monitoring information from the road surface flatness detection module (vehicle detection sensor) in real time. When the vehicle enters the monitoring range of the safe lane, the construction safety management system activates the horn system of the warning information release system to urge the maintenance personnel to leave the safe lane;

[0146] Step 3: The maintenance personnel signs out and the system returns to normal;

[0147] In some specific implementations, the system returns to normal means that each submodule in the dynamic safety warning module returns to the initial monitoring state and is ready for the next maintenance cycle or task; including the following aspects:

[0148] Face recognition monitoring module: Continue to monitor whether new maintenance personnel arrive at the site.

[0149] Construction safety management system: stop issuing construction warning instructions and restore the construction status to the non-construction status.

[0150] Warning information release system: turn off the construction warning information on the display screen and stop the alarm of the horn system.

[0151] After the system returns to normal, the construction safety management system remains on standby, ready to restart the corresponding monitoring and early warning processes when new maintenance needs are detected or new maintenance tasks are received.

[0152] Step 4: Upload data to the cloud platform data processing module in real time;

[0153] In some specific implementations, the cloud platform data processing module also processes the following data:

[0154] ① Face recognition data: including maintenance personnel’s check-in and check-out records, including the maintenance personnel’s identity information, arrival time, and departure time.

[0155] ② Vehicle detection data: Data from the road surface roughness detection module (vehicle detection sensor), which records the vehicle entry and exit conditions of the safe lane and real-time traffic flow information.

[0156] ③ Warning information data: record the working status of the warning information release system during construction, including the information content on the display screen and the triggering time of the horn alarm.

[0157] The above data processing process is as follows:

[0158] ① Data preprocessing: Clean the uploaded data, remove redundant and noisy data, and ensure the accuracy and completeness of the data.

[0159] ②Data analysis:

[0160] Behavioral analysis: Analyze facial recognition data to understand the working hours and frequency of maintenance personnel and evaluate the completion of maintenance tasks.

[0161] Traffic flow analysis: Statistical analysis of vehicle detection data to evaluate the use of safe lanes and vehicle flow distribution.

[0162] Warning effect analysis: Analyze warning information data and evaluate the working effect of the warning information release system, including the triggering frequency and duration of warning information.

[0163] Data storage and backup: The cloud platform stores and backs up the pre-processed and analyzed data to ensure data security and traceability.

[0164] In some specific implementations, the cloud platform data processing module analyzes facial recognition data, and the system can generate maintenance personnel's work records and performance evaluation reports to provide decision support for managers. The cloud platform data processing module analyzes vehicle detection data, and the system can monitor the traffic conditions of the escape lanes in real time, detect abnormal traffic or potential traffic hazards, and issue warning information in a timely manner. Through the analysis of warning information data, the system can evaluate and optimize the working effect of the warning information release system to ensure that warning information can be released in a timely and effective manner during construction and in emergency situations to ensure the safety of maintenance personnel and drivers. The data stored in the cloud platform data processing module can be used for long-term historical trend analysis to help road management departments understand the use and maintenance of escape lanes, predict future maintenance needs and traffic flow trends, and formulate more scientific management and maintenance plans.

[0165] This implementation method has carried out algorithm innovation and technology integration for the flatness detection of the safe lane, aiming to improve the detection accuracy and the ability to cope with complex terrain. By combining the cloth simulation algorithm with the filtering technology based on slope and triangulation, the improved filtering algorithm significantly enhances the recognition of subtle terrain features, optimizes the delicacy of data processing, and ensures efficient applicability under variable terrain. At the same time, in response to the impact of weather changes on detection accuracy, the improved RANSAC algorithm uses the real-time data input of the video detection equipment, combined with the intelligent fusion strategy of historical data, to dynamically adjust the algorithm parameters to maintain stability and measurement accuracy under different meteorological conditions, thereby achieving efficient use of resources and effective response to severe weather conditions.

[0166] This implementation method focuses on the functional guarantee of the escape lane as an emergency safety facility. Its core is to ensure that the out-of-control vehicle can effectively slow down and stop safely due to the high flatness of the road surface, which is of decisive significance for improving overall road safety. By integrating information collection, intelligent analysis and real-time monitoring technology, this implementation method promotes the transformation of road surface flatness detection to full automation and intelligence, greatly improves detection efficiency, reduces interference from human factors, and ensures the reliability and accuracy of detection results. This implementation method comprehensively considers the traffic load, material properties and international roughness index (IRI) standard of the escape lane. This implementation method sets a correction factor based on safety considerations, limiting the maximum threshold of IRI to below 3.5m / km, aiming to provide a better driving surface for emergency situations and enhance the safety and stability of the road system. The formulation of this standard rigorously balances the actual application scenarios and safety requirements, and provides a scientific basis for the quantitative evaluation of the flatness of the escape lane.

[0167] like Figure 7 In some possible implementations, the control method of the safe lane flatness monitoring and warning system is as follows:

[0168] S1. System initialization and self-check: After the system is started, it performs self-check, including status check of laser scanner, vehicle detection sensor, data transmission interface, cloud platform connection and dynamic safety warning module;

[0169] S2, vehicle monitoring: The vehicle detection sensor at the intersection monitors in real time whether there are vehicles passing through the safe lane;

[0170] S3, road surface scanning: when the sensor detects that the incoming vehicle has exited the safe lane and no vehicle is detected, the ground 3D laser scanner is started to perform a comprehensive scan of the safe lane to generate 3D point cloud data of the road surface;

[0171] S301, ground 3D laser scanner is installed beside the braking slope section of the escape lane, with a radiation range of 270 meters. If the radiation range is exceeded, another one will be installed. When setting the scanner angle, its center line should be parallel to the vertical plane of the ground;

[0172] S302, the ground three-dimensional laser scanner is automatically started to perform a three-dimensional scan of the danger lane;

[0173] S303, the scanner accurately records the three-dimensional position information of the road surface by measuring the time delay and intensity reflection of the laser. The data is recorded in the form of a point cloud, and each point represents a specific coordinate position on the road surface;

[0174] S304, as the scanning proceeds, the built-in software of the scanner will generate a preliminary analysis map of the road surface in real time, detecting possible depressions, cracks, bulges and other road surface unevenness;

[0175] S305, after the preliminary analysis is completed, the road surface flatness detection module performs preprocessing of the point cloud data, including point cloud denoising, point cloud registration and point cloud sampling;

[0176] S306, the pre-processed point cloud data is wirelessly uploaded to the cloud platform data analysis module through the data transmission interface after encryption protection;

[0177] S307. After completing the scanning of the entire road, the ground 3D laser scanner automatically turns off and enters a standby state, waiting for the next scanning cycle or system command.

[0178] S4, data upload and analysis: The scanner sends the pre-processed data to the data transmission interface, which is encoded by the module and uploaded to the cloud platform. The data processing unit of the cloud platform filters and analyzes the uploaded data and compares it with the standard data;

[0179] S401, using the least squares quadratic surface fitting method to perform ground filtering on the pre-processed point cloud data to retain ground data and part of near-ground vegetation;

[0180] S402, using an improved RANSAC algorithm to fit a plane to extract road point cloud data, so as to achieve road extraction;

[0181] S403, selecting a method of assigning initial values ​​continuously, selecting a statistical interval of 100 m, and using an international roughness calculation format to calculate the international roughness index of the scanned road surface;

[0182] S404, comparing the international roughness index calculated from the measured data with the predetermined international roughness index IRI of the road surface of the danger lane, if it exceeds the standard, entering the maintenance state, otherwise, the system is silent.

[0183] S5. Safety and maintenance decision: Based on the analysis results of the cloud platform standard comparison unit, determine whether the road surface is safe or needs maintenance. The maintenance decision unit decides whether to issue a road surface maintenance notice based on the analysis results;

[0184] S6, User information update: Maintenance personnel receive maintenance task notifications sent via the mobile app, confirm and prepare for construction. The driver information system updates the status of the safe lane, providing real-time road conditions and safety warnings for passing vehicles;

[0185] S7. Safety warning in construction areas: When the cloud platform determines that a road needs maintenance, the warning system will automatically activate the facial recognition monitoring and warning information release system. Maintenance personnel perform facial recognition clocking in at the beginning of maintenance to activate the dynamic safety warning module, release information about construction hazards to the display screen to prevent vehicles from entering the construction area, and activate the horn system when vehicles enter the safe lane to urge maintenance personnel to leave the safe lane. Maintenance personnel perform facial recognition clocking in at the end of maintenance to achieve information feedback;

[0186] S8. Reset the monitoring system and active warning information release system: After the construction is completed, all systems are reset to prepare for the next monitoring and early warning work.

[0187] To verify the effectiveness of the present invention, this embodiment is tested on the standard point cloud data samples provided by the official website of the International Society for Photogrammetry and Remote Sensing (ISPRS), specifically the Samp51 data sample. The area of ​​this data set is a piece of land with steep slopes, with vegetation on the steep slopes, and some of the flat land on both sides are slope areas, covering the detailed classification of ground points and non-ground points. The data in the experiment have been manually edited and classified on the ISPRS platform to ensure its accuracy and consistency. Such data sets are widely used to verify the accuracy and reliability of different filtering algorithms, and are suitable for scenarios such as terrain flatness assessment and safe lane flatness monitoring.

[0188] In this implementation, an accuracy assessment model is used to evaluate the performance of different filtering algorithms. The model divides the errors into two categories:

[0189] I error: The error in which a ground point is misclassified as a non-ground point, indicating that the actual ground point is not correctly identified.

[0190] Type II error: The error in which non-ground points are misclassified as ground points, indicating that actual non-ground points are mistakenly identified as ground points.

[0191] The test data results are shown in Table 1:

[0192] Table 1 Classification results

[0193] ground points Non-ground points Ground point misclassification Non-ground point misclassification Slope filter algorithm 12900 4800 1400 720 Cloth simulation algorithm 13200 4500 1050 600 Triangulation algorithm 13800 3900 980 500 The present invention 14000 3700 750 650

[0194] The error calculation results of different methods are shown in Table 2:

[0195] Type I Error Type II Error Total error Slope filter algorithm 10.85％ 15.00％ 11.92％ Cloth simulation algorithm 7.95％ 13.33％ 9.17％ Triangulation algorithm 7.10％ 12.82％ 8.45％ The present invention 5.36％ 17.57％ 7.32％

[0196] Through careful analysis of experimental data and evaluation using a specific error model, although the innovative filtering algorithm proposed in the present invention is slightly higher than other comparison algorithms in terms of the second type of error, it shows a significant advantage in terms of the first type of error, and the overall error reaches the lowest level. This result shows that in challenging environments such as hillside terrain like the escape lane, the method of this embodiment can achieve the best performance and effectively balance the accuracy of measurement and reliability in practical applications.

[0197] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0198] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A risk avoidance lane flatness monitoring and early warning system, characterized in that: It includes a road surface smoothness module, a cloud platform data analysis module, a user interaction interface module, and a dynamic safety warning module which are connected in sequence; The dynamic safety warning module is also connected to the road surface smoothness module signal; The road surface flatness module is used to automatically collect and pre-process the road surface point cloud data of the safe lane, and upload the pre-processed data to the cloud platform data analysis module through data transmission; The cloud platform data analysis module includes a vehicle detection sensor, a ground three-dimensional laser scanner, and a data transmission interface. The vehicle detection sensor is connected to the ground three-dimensional laser scanner by signal, the ground three-dimensional laser scanner is connected to the data transmission interface by signal, and the data transmission interface is connected to the cloud platform data analysis module through Internet data.

2. The risk avoidance lane flatness monitoring and early warning system according to claim 1 is characterized in that: The cloud platform data analysis module is installed in a computer with storage and data processing functions, and includes: A data storage unit, used for securely storing and backing up the pre-processed point cloud data; The data processing unit performs data filtering on the pre-processed point cloud data, and after the data filtering is completed, determines the flatness index of the road surface of the danger avoidance lane; A standard comparison unit is used to compare and analyze the actual road surface roughness index with the improved safe lane roughness detection standard; The maintenance decision unit is used to determine whether repaving is necessary based on the data comparison and analysis results.

3. The system for monitoring and warning the roughness of a dangerous lane according to claim 2 is characterized in that: The process of digital filtering performed by the data processing unit is specifically as follows: S1, slope-based filtering; S101. Use the slope calculation formula to calculate the slope S of each pair of adjacent points in the point cloud: Among them, x1, x2, y1, y2, z1, z2 represent the three-dimensional coordinates of two adjacent sampling points; S102, preliminary filtering: adaptive parameter preset, eliminating non-ground points with a slope greater than a preset threshold; S2, refinement of triangulated network based on cloth model optimization; S201, constructing the point cloud data after preliminary filtering into an irregular triangulated network model by using Delaunay triangulation; S202, simulating a virtual cloth model, and covering the virtual cloth on the irregular triangulated network model constructed in S201; wherein the elastic characteristics of the virtual cloth are: F=k×(Δx-x0) F is the elastic force, k is the spring constant, Δx is the deformation of the spring, and x0 is the natural length of the spring without force; S203, simulating the sagging process of the cloth, using gravity to simulate the natural compliance of the cloth to the terrain, to identify the non-ground points that are misidentified. This process updates the position through Verlet integration: x(t+Δt) represents the position of the cloth at time t+Δt, v(t) represents the velocity of the cloth at time t, a represents the acceleration of the cloth at time t, Δt is the time interval, and x(t) represents the position of the cloth at time t; S204, integrating the virtual cloth model with the TIN model: the cloth updates and adjusts the TIN grid according to the final state of the drooping process, and the ground point cloud data is optimized according to the result of the cloth drooping; the optimization objective function is: Among them, p i Indicates the position of the i-th vertex in the TIN model. It is a three-dimensional vector containing the X, Y, and Z coordinates of the vertex. i Represents the relationship between the i-th TIN vertex p i The three-dimensional coordinates of the corresponding point on the cloth model. The position of this point is in the process of cloth drooping, and it is consistent with the vertex p of the TIN model. i The closest point; i represents the sequence number of the vertex in the TIN model, and n represents the total number of vertices in the TIN model; S205, using the cloth simulation results, re-screening and refining the ground point cloud, and removing those non-ground points that fail to pass the cloth simulation verification; S206. Reconstruct the TIN model using the finally screened and adjusted ground point cloud data.

4. The risk avoidance lane flatness monitoring and early warning system according to claim 2 is characterized in that: The step of determining the roughness index of the danger lane by the data processing unit is specifically as follows: S301, based on the RANSAC algorithm, randomly sample the digitally filtered point cloud data, increase the sampling probability of the historical model neighboring points during the sampling process; use the weighted average method to integrate the current data point p c and historical data point p h , forming a new comprehensive point p n , p n =w×p c +(1-w)×p h , where w is the weight based on the similarity measure; S302, adjusting the RANSAC algorithm parameters, using the least squares method, combined with the improved RANSAC algorithm to iterate to find the most suitable plane parameters so that the plane equation is most suitable for the current point cloud data set; S303, according to the point cloud data characteristics extracted by the improved RANSAC algorithm, using the law of XY coordinate projection, extracting the road boundary in the shape of a straight line or a regular curve; for the straight line in the extracted road boundary line data, using the slope distance straight line formula to fit, for the curve, using a high-order polynomial to fit; S304, extracting the road point cloud data using the fitted boundary line to extract the complete road point cloud data, and removing noise from the boundary noise data using the least squares plane fitting method; S305. Maintain and optimize the historical data set, including regularly filtering out data that is no longer applicable and updating new successfully fitted model data, to ensure the timeliness and applicability of the historical data set, and substitute the extracted valid data into the predetermined international roughness index calculation formula to obtain the international roughness index actually measured on the road surface.

5. The risk avoidance lane flatness monitoring and early warning system according to claim 4 is characterized in that: The adjustment of the RANSAC algorithm parameters in S302 is specifically as follows: S3021. Establish a standard reference image library; S3022, using the video surveillance at the entrance of the escape lane to capture the image of the vehicle exiting, and importing the image at this moment into the Matlab processing module; S3023, evaluating the clarity of the real-time image; S3024, using a similarity metric to compare the weighted average of the input image with the representative values ​​of each level in the database, find the clarity level with the smallest difference, and associate the input image with the level; S3025. Transmit the numerical value corresponding to the real-time image clarity to the RANSAC algorithm calculation module to complete the adjustment of the RANSAC algorithm.

6. The risk avoidance lane flatness monitoring and early warning system according to claim 1 is characterized in that: The user interaction interface module is used to provide real-time information to maintenance personnel and drivers, including a maintenance personnel mobile app and a driving personnel information system.

7. The risk avoidance lane flatness monitoring and early warning system according to claim 1 is characterized in that: The dynamic safety warning module is used to ensure the safety of construction workers and issue road condition warnings to drivers in a timely manner, including facial recognition monitoring, construction safety management system, and warning information release system; the facial recognition monitoring is installed at the entrance of the safe lane and is signal-connected to the construction safety management system and the warning information release system; the construction safety management system is installed in a computer with a storage medium and a processor; the warning information release system is installed at the entrance of the safe lane and is signal-connected to the construction safety management system, including a display screen and a horn alarm system.

8. The control method of the risk avoidance lane flatness monitoring and early warning system according to claims 1 to 7, characterized in that: The following steps are involved: Step 1: After the system is started, it will conduct self-test, including the status test of laser scanner, vehicle detection sensor, data transmission interface, cloud platform connection and dynamic safety warning module; Step 2: Use the vehicle detection sensor at the intersection to monitor in real time whether there are vehicles passing through the safe lane; Step 3: When the sensor detects that the vehicle has exited the safe lane and no vehicle is detected, the ground 3D laser scanner is started to perform a comprehensive scan of the safe lane to generate 3D point cloud data of the road surface; Step 4: The ground 3D laser scanner sends the pre-processed data to the data transmission interface, which is encoded by the module and uploaded to the cloud platform data analysis module; the data processing unit of the cloud platform data analysis module filters and analyzes the uploaded data and compares it with the standard data; Step 5: According to the analysis results of the cloud platform data analysis module standard comparison unit, determine whether the road surface is safe or needs maintenance; Step 6: Maintenance personnel receive maintenance task notifications sent via the mobile app, confirm and prepare for construction; the driver information system updates the status of the safe lane, providing real-time road conditions and safety warnings for passing vehicles; Step 7. When the cloud platform data analysis module determines that the road needs maintenance, the facial recognition monitoring and warning information release system is automatically activated. The maintenance personnel perform facial recognition punch-in when starting maintenance, publish information about construction hazards to the display screen of the warning information release system, and activate the horn system when the vehicle enters the safe lane. The maintenance personnel perform facial recognition punch-in when ending maintenance.

9. The control method of the risk avoidance lane flatness monitoring and early warning system according to claim 8, characterized in that: The step 3 of scanning the road surface with a ground-based 3D laser scanner comprises the following steps: Step 301, the ground three-dimensional laser scanner is automatically started to perform a three-dimensional scan of the danger lane; Step 302, the ground 3D laser scanner records the 3D position information of the road surface by measuring the time delay and intensity reflection of the laser; the data is recorded in the form of a point cloud, and each point represents a specific coordinate position on the road surface; Step 303, the ground three-dimensional laser scanner generates a preliminary analysis map of the road surface in real time to detect possible uneven road surfaces; Step 304, after the preliminary analysis is completed, the road surface flatness detection module performs preprocessing of the point cloud data, including point cloud denoising, point cloud registration and point cloud sampling; Step 305, the pre-processed point cloud data is uploaded to the cloud platform data analysis module through the data transmission interface after encryption protection; Step 306. After completing the scanning of the entire road, the ground 3D laser scanner automatically turns off and enters the standby state, waiting for the next scanning cycle or system command.

Images