A method and system for visualizing road disease distribution

Abstract

The present application provides a kind of road disease distribution visualization method and system, including data acquisition module through mobile detection vehicle equipped with industrial camera, three-dimensional laser scanner and combined navigation device, real-time collection disease data, and carry out space-time synchronization and pretreatment;Disease analysis module realizes the space alignment of multi-source data based on position data, identifies the disease boundary in image through semantic segmentation and maps to scanning data, extracts disease point set and calculates disease area, depth and volume parameters;Decision module integrates historical and real-time disease parameters, weather data, generates disease evolution trend by combining time series prediction model, and marks high-risk areas by risk index;Visualization module dynamically renders disease markers on three-dimensional map, supports interactive display of historical changes on time axis, and dynamically renders high-risk areas on heat map.The present application realizes multi-dimensional dynamic monitoring and risk warning of road disease, and improves road maintenance decision efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation infrastructure monitoring, and in particular to a method and system for visualizing road hazard distribution. Background Art

[0002] Current road defect detection relies primarily on manual inspections or single sensor devices, which present significant limitations. Traditional manual inspections are inefficient, making it difficult to achieve high-frequency coverage over large areas, and subjective judgments can easily lead to omissions or misjudgments of defect records. Automated inspections using on-board sensors often lack effective spatiotemporal synchronization mechanisms for multi-source data, making it difficult to accurately correlate images, point clouds, and location information, and thus to construct a unified spatial defect model. Furthermore, existing methods lack quantitative analysis of defect characteristics, often remaining at the level of two-dimensional image recognition. Accurately extracting three-dimensional geometric parameters (such as pothole volume and crack depth) hinders objective assessment of defect severity.

[0003] In terms of disease visualization and decision-making, existing technologies generally suffer from insufficient dynamics and a lack of predictive capabilities. Most systems can only statically display the historical distribution of diseases and are unable to integrate environmental factors such as meteorology to predict evolutionary trends, making it difficult to timely mark high-risk areas. The visualization format is single, usually presented as two-dimensional icons or tables, lacking a three-dimensional spatial correlation display, making it difficult for maintenance departments to intuitively understand the spatial distribution patterns and development trends of diseases. At the same time, the historical data backtracking function is weak, lacking interactive timeline analysis, and unable to support disease life cycle tracking and maintenance measures effectiveness evaluation, which reduces the scientific nature and timeliness of road maintenance decisions. Summary of the Invention

[0004] In order to solve the technical problems mentioned in the current background technology, the purpose of the present invention is to provide a method and system for visualizing road damage distribution.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A road damage distribution visualization system, comprising:

[0007] Data acquisition module: collects disease data in real time, including image data, scanning data and location data, and performs spatiotemporal synchronization and preprocessing of disease data;

[0008] The disease analysis module is configured to spatially align the image data and the scan data based on the position data; identify the diseased area in the image data to obtain the disease boundary; map the disease boundary to the scan data to extract a set of disease points; and obtain disease parameters based on the set of disease points, the disease parameters including the disease area, disease depth, and disease volume.

[0009] Decision module: Constructs a first data set based on the disease parameters and the location data, and obtains future weather data as a second data set; inputs the first and second data sets into a preset time series prediction model to generate a set of predicted values ​​for the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate;

[0010] The disease parameters include historical disease parameters and current disease parameters;

[0011] The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer;

[0012] The input layer inputs a first data set and a second data set;

[0013] The three types of gate layers are the forget gate, input gate and output gate;

[0014] The forget gate determines the retention ratio of the first data set based on the second data set;

[0015] The input gate filters new features based on the second data set;

[0016] The output gate generates a hidden state feature;

[0017] The first output layer outputs the predicted value set based on the hidden state feature;

[0018] The second output layer outputs a risk index based on the hidden state feature;

[0019] Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold;

[0020] Visualization module: Generates disease markers on a three-dimensional map, where the size and shape of the disease markers are determined by the disease parameters; displays changes in disease parameters through a timeline control; renders the high-risk areas using a heat map gradient, where the color and transparency of the high-risk areas are dynamically adjusted based on the risk index.

[0021] Furthermore, the disease data is collected by a mobile inspection vehicle equipped with an industrial camera, a three-dimensional laser scanner and a combined navigation device;

[0022] The industrial camera is used to collect the image data, the three-dimensional laser scanner is used to collect the scanning data, and the combined navigation device is used to collect the position data.

[0023] Furthermore, the spatiotemporal synchronization includes time synchronization and space calibration;

[0024] Time synchronization uses a precise time protocol to establish a master-slave clock architecture;

[0025] The master clock is an atomic clock built into the integrated navigation device, which outputs the coordinated universal time as a time reference;

[0026] The slave clock is the clock built into the industrial camera and the 3D laser scanner, and synchronizes the time to the master clock.

[0027] The spatial calibration obtains the conversion parameters from the industrial camera to the 3D laser scanner through the checkerboard calibration method, including the rotation matrix R and the first translation vector T , the industrial camera coordinates are converted to the 3D laser scanner coordinate system, and the conversion formula is:

[0028] ;

[0029] in, 、 and is the coordinate of the converted 3D laser scanner, 、 and is the industrial camera coordinate;

[0030] The first translation vector T is a column vector.

[0031] Furthermore, the pre-processing steps are as follows:

[0032] 1) Using Gaussian filtering to reduce noise on the image data, and setting the filter kernel size and standard deviation;

[0033] 2) Use statistical outlier removal algorithm to filter abnormal scan data,

[0034] Abnormal scan data is judged by setting a standard deviation threshold. Scan data whose mean exceeds the standard deviation threshold is judged as abnormal scan data;

[0035] 3) The position data is smoothed by Kalman filtering, and the attitude angle is used as a state variable to suppress signal jitter of the integrated navigation device.

[0036] Furthermore, the location data includes latitude and longitude, elevation and attitude angle;

[0037] Convert image data and scan data to a unified world coordinate system using position data:

[0038] ;

[0039] in, 、 and is the coordinate of the world coordinate system, is the rotation matrix, is the second translation vector;

[0040] The second translation vector is a column vector;

[0041] The rotation matrix The calculation formula is as follows:

[0042] ;

[0043] in, 、 and Respectively around axis, Axis and The rotation matrix of the axis, 、 and is the attitude angle.

[0044] Furthermore, the diseased area identification uses a semantic segmentation algorithm to identify the diseased area in the image data to obtain a binary mask of the diseased area;

[0045] Apply morphological edge detection to the binary mask to extract the coordinate set of the disease boundary:

[0046] The defect boundary is mapped to the scan data, and the coordinates in the coordinate set are solved by the projection equation to obtain the position of the defect in the 3D laser scanner coordinate system:

[0047] ;

[0048] in, and are coordinates in the coordinate set, is the 3×3 dimensional intrinsic parameter matrix of the industrial camera, is the preset third translation vector, is the preset 3×3 dimensional rotation matrix, is the third translation vector The 3×4 dimensional external parameter matrix of the industrial camera spliced ​​with the rotation matrix, is the scale factor;

[0049] The third translation vector is a 3×1 dimensional column vector;

[0050] Perform nearest neighbor search based on the space partition tree to extract the set of diseased points;

[0051] The defect points in the defect point set are projected onto the road plane, and the defect parameters are calculated.

[0052] Furthermore, the diseased area The calculation formula is as follows:

[0053] ;

[0054] in, and For the The coordinates of the projected disease points on the road plane, and is the coordinate of the next disease point projected on the road plane, is the total number of disease points;

[0055] The disease depth includes the maximum disease depth and the average disease depth:

[0056] Maximum disease depth The calculation formula is as follows:

[0057] ;

[0058] in, For the The elevation value of each disease point, is the elevation of the theoretical road plane,

[0059] Average disease depth The calculation formula is as follows:

[0060] ;

[0061] The set of defect points is divided into several triangles based on Delaunay triangulation. For each triangle, its projected area on the road plane and the average vertex depth are calculated. The total volume enclosed by the defect area and the road plane is obtained by multiplying and adding the projected area of ​​each triangle with the average vertex depth:

[0062] ;

[0063] in, is the disease volume, is the total number of triangles, For the b The projected area of ​​the triangle, is the average depth of the triangle vertices.

[0064] Furthermore, the disease mark is rendered in a three-dimensional polygonal form, and the size and shape of the disease mark are directly determined by the disease parameter;

[0065] The damage area is used to define the base size of the damage mark, and the height of the damage mark is calculated based on the linear mapping of the maximum damage depth. The average damage depth is used to control the color change of the damage mark. A depth threshold is set. When the average damage depth is greater than the depth threshold, the damage mark is displayed in red, otherwise it is displayed in yellow. The damage volume is used to adjust the overall transparency of the damage mark. The larger the damage volume, the lower the transparency.

[0066] Based on the first data set and the predicted value set, the changes of the disease parameters are displayed through a timeline control, and a user is supported to interactively browse the evolution of the disease;

[0067] The data driving mechanism of the timeline control is that when the user drags the timeline to a time point When loading If the data points are discontinuous, linear interpolation is used to ensure smooth transition.

[0068] Rendering the high-risk area as a semi-transparent overlay on a three-dimensional map, where the range of the overlay is determined by the boundary of the set of disease points, and achieving smooth rendering through triangular mesh subdivision;

[0069] The color attribute of the overlay is driven by the risk index, and the color value is calculated by an RGB interpolation formula; the transparency of the overlay is dynamically adjusted according to the risk index.

[0070] A road damage distribution visualization method, comprising:

[0071] Step 1: Real-time collection of disease data, including image data, scanning data, and location data, and spatiotemporal synchronization and preprocessing of the disease data;

[0072] Step 2: spatially aligning the image data and the scan data based on the position data;

[0073] Step 3: Identify the damaged area in the image data to obtain the damage boundary; map the damage boundary to the scan data to extract a set of damage points; and obtain damage parameters based on the set of damage points, wherein the damage parameters include damage area, damage depth, and damage volume;

[0074] Step 4: Construct a first data set based on the disease parameters and the location data, and obtain future weather data as a second data set; input the first data set and the second data set into a preset time series prediction model to generate a set of predicted values ​​of the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate;

[0075] The disease parameters include historical disease parameters and current disease parameters;

[0076] The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer;

[0077] The input layer inputs a first data set and a second data set;

[0078] The three types of gate layers are the forget gate, input gate and output gate;

[0079] The forget gate determines the retention ratio of the first data set based on the second data set;

[0080] The input gate filters new features based on the second data set;

[0081] The output gate generates a hidden state feature;

[0082] The first output layer outputs the predicted value set based on the hidden state feature;

[0083] The second output layer outputs a risk index based on the hidden state feature;

[0084] Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold;

[0085] Step 5: Generate disease markers on the three-dimensional map, where the size and shape of the disease markers are determined by the disease parameters; display changes in the disease parameters through a timeline control; render the high-risk areas using a heat map gradient, and dynamically adjust the color and transparency of the high-risk areas based on the risk index.

[0086] Compared with the prior art, the advantages of the present invention are:

[0087] 1. This invention uses a master-slave clock synchronization mechanism and spatial calibration technology to achieve millisecond-level spatiotemporal alignment of image, scan, and position data, building a unified spatial reference framework. Combining semantic segmentation and point cloud mapping technology, it breaks through the limitations of traditional two-dimensional detection and accurately extracts the three-dimensional geometric parameters of the disease, providing an objective quantitative basis for the severity of the disease.

[0088] 2. This invention integrates historical disease parameters with meteorological data to construct a time-series prediction model, generating real-time disease evolution trends. It also innovatively introduces a risk index algorithm to dynamically mark high-risk areas based on volume change rate, depth, and environmental factors, achieving a shift from passive monitoring to active early warning, significantly improving the timeliness of maintenance responses.

[0089] 3. The present invention uses disease parameters to drive dynamic rendering of three-dimensional maps, supports parametric expression of disease markers and interactive timeline backtracking; high-risk areas use heat map gradient rendering and adaptive transparency adjustment to achieve a dual intuitive presentation of spatial distribution and evolution laws, providing a multi-dimensional diagnostic basis for maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0091] Figure 1 Schematic diagram of the system workflow of the present invention;

[0092] Figure 2 A schematic diagram of the execution flow of the time series prediction model of the present invention;

[0093] Figure 3 Schematic diagram of the pretreatment of the present invention. DETAILED DESCRIPTION

[0094] To achieve the above objectives, the present invention is implemented through the following technical solutions: the present invention provides a road disease distribution visualization system, please refer to Figures 1 to 3 , the system comprises:

[0095] Data acquisition module: collects disease data in real time, including image data, scanning data and location data, and performs spatiotemporal synchronization and preprocessing of disease data.

[0096] Disease data is collected by mobile inspection vehicles equipped with industrial cameras, 3D laser scanners, and integrated navigation devices:

[0097] In this embodiment, image data is collected by an industrial camera, which collects road surface images (RGB images) with a resolution of ≥4096×2160, a frame rate of 30fps, and a lane width of 4.5m.

[0098] Scanning data is collected by a 3D laser scanner, which obtains point cloud data of the road with a scanning frequency of ≥100Hz and an accuracy of ±2mm (such as LiDAR sensor);

[0099] Position data is collected through a combined navigation device that integrates GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit) to record position coordinates (latitude, longitude, elevation) and attitude angles in real time, with a positioning accuracy of ≤5cm.

[0100] Spatiotemporal synchronization includes time synchronization and space calibration;

[0101] Time synchronization uses the Precision Time Protocol (PTP) to establish a master-slave clock architecture:

[0102] Master clock: The atomic clock of the integrated navigation unit (GNSS-IMU) serves as the master clock, which outputs Coordinated Universal Time (UTC) as the time reference.

[0103] Slave clock: The built-in clock in industrial cameras and 3D laser scanners is synchronized to the master clock, meeting the following requirements:

[0104] ;

[0105] in, is the maximum time deviation between the slave clock and the master clock, is the master clock time, and Local time for industrial cameras and 3D laser scanners;

[0106] Spatial calibration uses the checkerboard calibration method to obtain the conversion parameters from the industrial camera to the 3D laser scanner, including the rotation matrix R (3×3 matrix, representing the coordinate axis rotation relationship) and translation vector T (3×1 vector, indicating the coordinate origin offset), industrial camera coordinates ( , , ) is converted into the 3D laser scanner coordinate system, and the conversion formula is:

[0107] ;

[0108] in, 、 and is the coordinate of the converted 3D laser scanner;

[0109] The first translation vector T is a column vector.

[0110] The steps of data preprocessing are as follows:

[0111] 1) The image data is denoised by Gaussian filtering, with the filter kernel size set to 3×3 pixels and a standard deviation of σ =1.0 to reduce interference caused by lighting changes;

[0112] 2) The scan data is processed using a statistical outlier removal algorithm to remove abnormal scan data that deviates from the mean by more than two standard deviations;

[0113] 3) The position data is smoothed using Kalman filtering, using attitude angles as state variables to suppress Global Navigation Satellite System (GNSS) signal jitter.

[0114] The disease analysis module spatially aligns the image data and the scan data based on the position data; identifies the disease area in the image data to obtain the disease boundary; maps the disease boundary to the scan data and extracts the disease point set; obtains disease parameters based on the disease point set, and the disease parameters include disease area, disease depth and disease volume.

[0115] Position data includes latitude, longitude, elevation and attitude angle;

[0116] The image data (camera coordinate system) and scan data (3D laser scanner coordinate system) are converted to a unified world coordinate system through position data:

[0117] ;

[0118] in, 、 and is the coordinate of the world coordinate system, is the rotation matrix, is the second translation vector, the translation amount required to move the origin of the local coordinate system to the origin of the world coordinate system;

[0119] Second translation vector is a column vector;

[0120] Second translation vector The calculation formula is as follows:

[0121] ;

[0122] in, 、 and are latitude, longitude, and altitude, respectively. is the radius of curvature of the Maoyou circle, is the eccentricity of the ellipsoid;

[0123] Second translation vector The value of is expressed as ;

[0124] It is the curvature radius of the y-axis circle (parallel circle) at a certain point on the Earth's ellipsoid. The calculation formula is:

[0125] ;

[0126] in, is the equatorial radius;

[0127] Rotation Matrix The calculation formula is as follows:

[0128] ;

[0129] in, 、 and Respectively around axis, Axis and The rotation matrix of the axis, 、 and is the attitude angle.

[0130] For example:

[0131] ;

[0132] The semantic segmentation algorithm is used to identify the diseased area in the image data to obtain a binary mask of the diseased area (0 represents background, 1 represents disease);

[0133] Apply morphological edge detection (such as Sobel operator) to the binary mask to extract the coordinate set of the disease boundary :

[0134] ;

[0135] in, are the coordinates of the boundary points.

[0136] Map the disease boundary to the scan data and The coordinates in the 3D laser scanner coordinate system are used to find the position of the defect through the projection equation:

[0137] ;

[0138] in, and are coordinates in the coordinate set, is the 3×3 dimensional intrinsic parameter matrix of the industrial camera, is the preset third translation vector, is the preset 3×3 dimensional rotation matrix, is the third translation vector The 3×4 dimensional external parameter matrix of the industrial camera spliced ​​with the rotation matrix, is the scale factor;

[0139] The third translation vector is a 3×1-dimensional column vector;

[0140] Perform nearest neighbor search based on KD-Tree to extract the set of disease points :

[0141] ;

[0142] Project the damage points onto the road plane to calculate the damage parameters;

[0143] Diseased area The calculation formula is as follows:

[0144] ;

[0145] in, and For the The coordinates of the projected disease points on the road plane, and is the coordinate of the next disease point projected on the road plane, is the total number of disease points;

[0146] The disease depth includes the maximum disease depth and the average disease depth:

[0147] Maximum disease depth The calculation formula is as follows:

[0148] ;

[0149] in, For the The elevation value of each disease point, is the elevation of the theoretical road plane,

[0150] Average disease depth The calculation formula is as follows:

[0151] ;

[0152] The set of defect points is divided into several triangles based on Delaunay triangulation. For each triangle, its projected area on the road plane and the average vertex depth are calculated. The total volume enclosed by the defect area and the road plane is obtained by multiplying and adding the projected area of ​​each triangle with the average vertex depth:

[0153] ;

[0154] in, is the disease volume, is the total number of triangles, For the b The projected area of ​​the triangle, is the average depth of the triangle vertices;

[0155] Decision module: Constructs a first data set based on the disease parameters and the location data, and obtains future weather data as a second data set; inputs the first and second data sets into a preset time series prediction model to generate a set of predicted values ​​for the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate;

[0156] The disease parameters include historical disease parameters and current disease parameters;

[0157] The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer;

[0158] The input layer inputs a first data set and a second data set;

[0159] The three types of gate layers are the forget gate, input gate and output gate;

[0160] The forget gate determines the retention ratio of the first data set based on the second data set;

[0161] The input gate filters new features based on the second data set;

[0162] The output gate generates a hidden state feature;

[0163] The first output layer outputs the predicted value set based on the hidden state feature;

[0164] The second output layer outputs a risk index based on the hidden state feature;

[0165] Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold;

[0166] Future weather data is obtained from the weather station, including future rainfall C (unit: mm), temperature D (unit: °C) and number of freeze-thaw cycles E (The lowest temperature of the day is ≤0℃ and the highest temperature is ≥5℃, which is counted as 1 time). The future weather data and historical disease data are aligned in time and space to ensure the correlation in the same location and time period.

[0167] In this embodiment: the high risk threshold is set to ,when When , it is marked as a high-risk area;

[0168] Visualization module: Generates disease markers on a three-dimensional map, where the size and shape of the disease markers are determined by the disease parameters; displays changes in disease parameters through a timeline control; renders the high-risk areas using a heat map gradient, where the color and transparency of the high-risk areas are dynamically adjusted based on the risk index.

[0169] Disease markers are rendered in 3D polygonal form, and their positions are determined by the world coordinates provided by the disease analysis module ( and and ), the size and shape of the disease mark are directly determined by the disease parameters:

[0170] The size and shape of the disease mark are directly determined by the disease parameters:

[0171] The disease area F is used to define the base size of the mark: the horizontal projection size of the disease mark on the three-dimensional map S (Unit: pixels) by scaling factor k Mapping calculation:

[0172] ;

[0173] Maximum disease depth and average disease depth Affects the height and color of the disease mark, the vertical height of the disease mark H (Unit: meter) Based on the maximum disease depth Linear mapping:

[0174] ;

[0175] in, is the preset height scaling factor;

[0176] At the same time, the average disease depth Used to control the color change of the disease mark and set the depth threshold. When the average disease depth is greater than the depth threshold, it will display red, otherwise it will display yellow.

[0177] The disease volume is used to determine the overall transparency of the mark. The larger the disease volume, the lower the transparency.

[0178] Based on the first dataset and the predicted value set, a timeline control is used to display the changes in disease parameters and allow users to interactively browse the evolution of the disease. The timeline control covers the time span. The specific steps are as follows:

[0179] The data-driven mechanism of the timeline, when the user drags the timeline to a time point When the disease parameters at that moment are loaded, if the data points are discontinuous (such as there are gaps between weekly data points), linear interpolation is used to ensure a smooth transition.

[0180] For the disease volume In time Interpolation calculation at:

[0181] ;

[0182] in, and is the timestamp of the adjacent historical data points, and is the volume of the disease at the corresponding time point;

[0183] After interpolation, the disease marks are dynamically updated on the 3D map.

[0184] The timeline control combines the changes in the disease markers to show the expansion or repair of the disease through animation effects.

[0185] For example, the historical trend of the diseased area F is represented by the gradual change of the marker size S: when the user slides the time axis, the marker size changes according to F( ) Real-time scaling to create a "growing" or "shrinking" visual effect.

[0186] At the same time, the color coding is associated with the disease depth, and the increase in disease depth will cause the color of the disease mark to transition from yellow to red;

[0187] The timeline control also supports playback mode, which automatically plays historical sequences to help users identify long-term trends (such as the exacerbation of diseases caused by winter freeze-thaw).

[0188] Dynamic rendering of high-risk areas, the specific steps are as follows:

[0189] Geometry generation of high-risk areas: High-risk areas are rendered as semi-transparent overlays on the 3D map. The overlay range is determined by the boundaries of the disease point set and smooth rendering is achieved through triangular mesh subdivision.

[0190] Dynamic color mapping: The color attribute of the overlay is driven by the risk index, and the color value is calculated using the RGB interpolation formula:

[0191] ;

[0192] in, Z is the risk index, and is the preset minimum and maximum risk index, ensuring Z Higher values ​​make the overlay redder (high warning), while low Z values ​​make it appear green (safe).

[0193] In addition, the transparency of the overlay (Range: 0.3~0.7) Dynamically adjusted according to risk index:

[0194] ;

[0195] This ensures that high-risk areas are both visible and do not completely obscure the underlying map.

[0196] Real-time update mechanism: The rendering of high-risk areas is dynamic and updated according to the risk index;

[0197] When weather data changes (e.g., rainfall C increases), the risk index is recalculated and the overlay color and transparency are adjusted in real time. For example, if heavy rainfall is predicted, the area will flash red as a warning.

[0198] A road damage distribution visualization method, comprising:

[0199] Step 1: Real-time collection of disease data, including image data, scanning data, and location data, and spatiotemporal synchronization and preprocessing of the disease data;

[0200] Step 2: spatially aligning the image data and the scan data based on the position data;

[0201] Step 3: Identify the damaged area in the image data to obtain the damage boundary; map the damage boundary to the scan data to extract a set of damage points; and obtain damage parameters based on the set of damage points, wherein the damage parameters include damage area, damage depth, and damage volume;

[0202] Step 4: Construct a first data set based on the disease parameters and the location data, and obtain future weather data as a second data set; input the first data set and the second data set into a preset time series prediction model to generate a set of predicted values ​​of the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate;

[0203] The disease parameters include historical disease parameters and current disease parameters;

[0204] The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer;

[0205] The input layer inputs a first data set and a second data set;

[0206] The three types of gate layers are the forget gate, input gate and output gate;

[0207] The forget gate determines the retention ratio of the first data set based on the second data set;

[0208] The input gate filters new features based on the second data set;

[0209] The output gate generates a hidden state feature;

[0210] The first output layer outputs the predicted value set based on the hidden state feature;

[0211] The second output layer outputs a risk index based on the hidden state feature;

[0212] Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold;

[0213] Step 5: Generate disease markers on the three-dimensional map, where the size and shape of the disease markers are determined by the disease parameters; display changes in the disease parameters through a timeline control; render the high-risk areas using a heat map gradient, and dynamically adjust the color and transparency of the high-risk areas based on the risk index.

[0214] In summary, the present invention innovatively incorporates meteorological factors into the time series prediction model to generate the dynamic evolution trend of diseases. It quantifies the coupling effect of environmental and disease parameters through the risk index algorithm, realizes the active marking of high-risk areas, and transforms passive response into forward-looking warning.

[0215] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A road disease distribution visualization system, characterized by: include: Data acquisition module: collects disease data in real time, including image data, scanning data and location data, and performs spatiotemporal synchronization and preprocessing of disease data; Disease analysis module: spatially aligning the image data and the scan data based on the position data; Identifying a diseased area in the image data to obtain a disease boundary; mapping the disease boundary to the scan data to extract a set of disease points; and obtaining disease parameters based on the set of disease points, the disease parameters including disease area, disease depth, and disease volume; Decision module: Constructs a first data set based on the disease parameters and the location data, and obtains future weather data as a second data set; inputs the first and second data sets into a preset time series prediction model to generate a set of predicted values ​​for the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate; The disease parameters include historical disease parameters and current disease parameters; The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer; The input layer inputs a first data set and a second data set; The three types of gate layers are the forget gate, input gate and output gate; The forget gate determines the retention ratio of the first data set based on the second data set; The input gate filters new features based on the second data set; The output gate generates a hidden state feature; The first output layer outputs the predicted value set based on the hidden state feature; The second output layer outputs a risk index based on the hidden state feature; Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold; Visualization module: Generates disease markers on a three-dimensional map. The size and shape of the disease markers are determined by disease parameters. Changes in disease parameters are displayed through a timeline control. The high-risk area is rendered using a heat map gradient, and the color and transparency of the high-risk area are dynamically adjusted based on the risk index.

2. The system according to claim 1, wherein: The disease data is collected by a mobile inspection vehicle equipped with an industrial camera, a three-dimensional laser scanner and a combined navigation device; The industrial camera is used to collect the image data, the three-dimensional laser scanner is used to collect the scanning data, and the combined navigation device is used to collect the position data.

3. The system according to claim 2, characterized in that The spatiotemporal synchronization includes time synchronization and space calibration; Time synchronization uses a precise time protocol to establish a master-slave clock architecture; The master clock is an atomic clock built into the integrated navigation device, which outputs the coordinated universal time as a time reference; The slave clock is the clock built into the industrial camera and the 3D laser scanner, and synchronizes the time to the master clock. The spatial calibration obtains the conversion parameters from the industrial camera to the 3D laser scanner through the checkerboard calibration method, including the rotation matrix R and the first translation vector T , the industrial camera coordinates are converted to the 3D laser scanner coordinate system, and the conversion formula is: ； in, 、 and is the coordinate of the converted 3D laser scanner, 、 and is the industrial camera coordinate; The first translation vector T is a column vector.

4. The system according to claim 3, characterized in that The steps of the pre-processing are as follows: 1) Using Gaussian filtering to reduce noise on the image data, and setting the filter kernel size and standard deviation; 2) Use statistical outlier removal algorithm to filter abnormal scan data, Abnormal scan data is judged by setting a standard deviation threshold. Scan data whose mean exceeds the standard deviation threshold is judged as abnormal scan data; 3) The position data is smoothed by Kalman filtering, and the attitude angle is used as a state variable to suppress signal jitter of the integrated navigation device.

5. The system according to claim 4, characterized in that The location data includes latitude and longitude, elevation and attitude angle; Convert image data and scan data to a unified world coordinate system using position data: ； in, 、 and is the coordinate of the world coordinate system, is the rotation matrix, is the second translation vector; The second translation vector A is a column vector; The rotation matrix The calculation formula is as follows: ； in, 、 and Respectively around axis, Axis and The rotation matrix of the axis, 、 and is the attitude angle.

6. The system according to claim 5, characterized in that The diseased area identification uses a semantic segmentation algorithm to identify the diseased area in the image data to obtain a binary mask of the diseased area; Apply morphological edge detection to the binary mask to extract the coordinate set of the disease boundary: The defect boundary is mapped to the scan data, and the coordinates in the coordinate set are solved by the projection equation to obtain the position of the defect in the 3D laser scanner coordinate system: ； in, and are coordinates in the coordinate set, is the 3×3 dimensional intrinsic parameter matrix of the industrial camera, is the preset third translation vector, is the preset 3×3 dimensional rotation matrix, is the third translation vector The 3×4 dimensional external parameter matrix of the industrial camera spliced ​​with the rotation matrix, is the scale factor; The third translation vector is a 3×1 dimensional column vector; Perform nearest neighbor search based on the space partition tree to extract the set of diseased points; The defect points in the defect point set are projected onto a road plane, and the defect parameters are calculated.

7. The system according to claim 6, characterized in that The diseased area The calculation formula is as follows: ； in, and For the The coordinates of the projected disease points on the road plane, and is the coordinate of the next disease point projected on the road plane, is the total number of disease points; The disease depth includes the maximum disease depth and the average disease depth: Maximum disease depth The calculation formula is as follows: ； in, For the The elevation value of each disease point, is the elevation of the theoretical road plane, Average disease depth The calculation formula is as follows: ； The set of defect points is divided into several triangles based on Delaunay triangulation. For each triangle, its projected area on the road plane and the average vertex depth are calculated. The total volume enclosed by the defect area and the road plane is obtained by multiplying and adding the projected area of ​​each triangle with the average vertex depth: ； in, is the disease volume, is the total number of triangles, For the b The projected area of ​​the triangle, is the average depth of the triangle vertices.

8. The system according to claim 7, characterized in that The disease mark is rendered in a three-dimensional polygonal form, and the size and shape of the disease mark are directly determined by the disease parameters; The disease area is used to define the base size of the disease mark, and the height of the disease mark is calculated based on the linear mapping of the maximum disease depth; The average disease depth is used to control the color change of the disease mark. A depth threshold is set. When the average disease depth is greater than the depth threshold, it is displayed in red, otherwise it is displayed in yellow. The disease volume is used to adjust the overall transparency of the disease mark. The larger the disease volume, the lower the transparency. Based on the first data set and the predicted value set, the changes of the disease parameters are displayed through a timeline control, and a user is supported to interactively browse the evolution of the disease; The data driving mechanism of the timeline control is that when the user drags the timeline to a time point When loading If the data points are discontinuous, linear interpolation is used to ensure smooth transition. Rendering the high-risk area as a semi-transparent overlay on a three-dimensional map, where the range of the overlay is determined by the boundary of the set of disease points, and achieving smooth rendering through triangular mesh subdivision; The color attribute of the overlay is driven by the risk index, and the color value is calculated by an RGB interpolation formula; the transparency of the overlay is dynamically adjusted according to the risk index.

9. A method for visualizing road damage distribution, implemented by the system according to any one of claims 1 to 8, characterized in that: include: Step 1: Real-time collection of disease data, including image data, scanning data, and location data, and spatiotemporal synchronization and preprocessing of the disease data; Step 2: spatially aligning the image data and the scan data based on the position data; Step 3: Identify the damaged area in the image data to obtain the damage boundary; map the damage boundary to the scan data to extract a set of damage points; and obtain damage parameters based on the set of damage points, wherein the damage parameters include damage area, damage depth, and damage volume; Step 4: Construct a first data set based on the disease parameters and the location data, and obtain future weather data as a second data set; input the first data set and the second data set into a preset time series prediction model to generate a set of predicted values ​​of the disease parameters and a risk index, wherein the time series prediction model includes a forget gate, an input gate, and an output gate; The disease parameters include historical disease parameters and current disease parameters; The time series prediction model includes an input layer, three types of gate layers, a first output layer and a second output layer; The input layer inputs a first data set and a second data set; The three types of gate layers are the forget gate, input gate and output gate; The forget gate determines the retention ratio of the first data set based on the second data set; The input gate filters new features based on the second data set; The output gate generates a hidden state feature; The first output layer outputs the predicted value set based on the hidden state feature; The second output layer outputs a risk index based on the hidden state feature; Set a high-risk threshold and mark the area as high-risk when the risk index exceeds the high-risk threshold; Step 5: Generate disease markers on the three-dimensional map, where the size and shape of the disease markers are determined by the disease parameters; display changes in the disease parameters through a timeline control; render the high-risk areas using a heat map gradient, and dynamically adjust the color and transparency of the high-risk areas based on the risk index.

Images