Road detection method and system based on unmanned aerial vehicle

Abstract

The invention provides a road detection method and system based on an unmanned aerial vehicle, and the method comprises the steps: carrying out the analysis of an original three-dimensional model of a road, and obtaining a cruise path; receiving first data collected when the unmanned aerial vehicle carrying the laser radar cruises according to the cruise path, and constructing a three-dimensional real scene model; and comparing and analyzing the three-dimensional live-action model and the original three-dimensional model to obtain an analysis result. According to the road detection method based on the unmanned aerial vehicle, rapid and accurate road detection is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of road detection, and in particular to a road detection method and system based on an unmanned aerial vehicle. Background Art

[0002] As an important part of the transportation system, the safety and reliability of roads are directly related to the overall operation of the transportation system. In addition, during the long-term use of roads, they will be affected by various factors such as the natural environment and vehicle loads, leading to structural damage and performance degradation. Through road health monitoring, abnormal conditions and potential risks of roads can be discovered in time to avoid road accidents, thereby ensuring traffic safety. In addition, damage and aging of roads can be discovered in time, and corresponding maintenance measures can be taken to extend the service life of roads and reduce maintenance costs.

[0003] Existing monitoring is basically achieved through manual methods, and how to achieve rapid road detection is a technical problem that needs to be solved urgently. Summary of the invention

[0004] One of the purposes of the present invention is to provide a road detection method and system based on drone to achieve fast and accurate road detection.

[0005] An embodiment of the present invention provides a road detection method based on an unmanned aerial vehicle, comprising: Analyze the original three-dimensional model of the road to obtain the cruising path; Receiving first data collected by the UAV equipped with a laser radar when cruising along a cruising path and constructing a three-dimensional real scene model; Compare and analyze the 3D real scene model and the original 3D model to obtain the analysis results.

[0006] Preferably, the original three-dimensional model is constructed based on design data or historical surveying and mapping data of the road.

[0007] Preferably, the analysis steps for the original three-dimensional model of the road are as follows: Perform region segmentation and remove areas that do not need to be detected; Filling the area so that the area boundary is a regular shape to serve as the target to be scanned; The scanning frame determined by the scanning boundary of the laser radar is used to scan the target to be scanned to obtain a scanning path; According to the distribution of the area to be scanned within the area scanned by the scanning frame, the scanning path is corrected to obtain the cruising path.

[0008] Preferably, the comparative analysis of the 3D real scene model and the original 3D model includes: The original 3D model and the 3D real scene model are dismantled and reduced based on the area to be inspected; The original three-dimensional model and the three-dimensional real scene model after demolition are segmented by segmentation grids and the segmented units are associated; The associated units are compared and statistics are calculated.

[0009] Preferably, the comparative analysis of the 3D real scene model and the original 3D model includes: Performing point sampling on the original three-dimensional model and the three-dimensional real scene model respectively to obtain a plurality of associated sampling points; The associated sampling points are compared and counted.

[0010] The present invention also provides a road detection system based on an unmanned aerial vehicle, comprising: a path analysis module, a real scene generation module and a comparative analysis module; wherein the path analysis module analyzes the original three-dimensional model of the road to obtain a cruising path; the real scene generation module receives first data collected by an unmanned aerial vehicle equipped with a laser radar when cruising along the cruising path and constructs a three-dimensional real scene model; the comparative analysis module compares and analyzes the three-dimensional real scene model and the original three-dimensional model to obtain an analysis result.

[0011] Preferably, the original three-dimensional model is constructed based on design data or historical surveying and mapping data of the road.

[0012] Preferably, the path analysis module includes: a segmentation and filling unit, a scanning unit and a correction unit; wherein the segmentation and filling unit performs area segmentation and removes areas that do not need to be detected, and performs area filling to make the area boundary a regular shape as a target to be scanned; the scanning unit scans the target to be scanned with a scanning frame determined by the scanning boundary of the laser radar to obtain a scanning path; the correction unit corrects the scanning path according to the distribution of the area to be scanned within the area scanned by the scanning frame to obtain a cruising path.

[0013] Preferably, the comparison and analysis module includes: a disassembly and association unit and a comparison and statistics unit; wherein the disassembly and association unit disassembles the original three-dimensional model and the three-dimensional real scene model based on the area to be detected, respectively, and divides the original three-dimensional model and the three-dimensional real scene model after disassembly with a segmentation grid and associates the segmented units; the comparison and statistics unit compares and performs statistics on the associated units.

[0014] Preferably, the comparison and analysis module also includes: a sampling association unit and a comparison statistics unit, wherein the sampling association unit performs point sampling on the original three-dimensional model and the three-dimensional real scene model respectively to obtain multiple associated sampling points; the comparison statistics unit compares and performs statistics on the associated sampling points.

[0015] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0016] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Schematic diagram of a road detection method based on a drone in an embodiment of the present invention; Figure 2 Schematic diagram of a road detection system based on a drone in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0019] The embodiment of the present invention provides a road detection method based on an unmanned aerial vehicle, such as Figure 1 As shown, including: Step 1: Analyze the original three-dimensional model of the road to obtain the cruising path; The original 3D model is constructed based on the design data or historical surveying data of the road; that is, the 3D modeling is performed according to the road size to generate the original 3D model; and the analysis steps for the original 3D model are as follows: performing area segmentation and removing the area that does not need to be detected; performing area filling so that the area boundary is a regular shape as the target to be scanned; scanning the target to be scanned with the scanning frame determined by the scanning boundary of the laser radar to obtain the scanning path; and correcting the scanning path according to the distribution of the area to be scanned within the area scanned by the scanning frame to obtain the cruising path; In the process of analyzing the original 3D model, the main purpose of regional segmentation is to separate the green area, water area, road area and guardrail area. In general, only the road area needs to be detected, so the green area, water area, guardrail area, etc. can be removed; only the road area is retained; the movement of the drone is to facilitate the laser radar to scan the target, so it is necessary to fill in the area to be detected in a regular shape, and then scan to determine the scanning path, and then determine the cruising path of the drone according to the scanning path; the scanning method is to align the two aligned sides of the scanning frame (the side of the rear end of the scanning direction and any adjacent side) with the two adjacent sides of the target to be scanned, and then translate in the scanning direction. After translating out of the boundary of the target to be scanned, flip 180 degrees and continue translating, and so on, until the entire target to be scanned is scanned, and the center points before and after the flipping position are connected with a preset arc, and the trajectory of the center point of the scanning frame is used as the scanning path; finally, the scanning path needs to be corrected, which is to determine the maximum span area of ​​the area to be detected in the scanning frame corresponding to each point on the scanning path (that is, the distance between the two farthest boundaries of the first area to be detected and the last area to be detected in the scanning frame); determine the midpoint of the span area, and replace the corresponding point on the scanning path with the midpoint; ensure that the area to be detected is located in a better scanning position through correction; improve the scanning effect, and thus improve the accuracy of model construction; Step 2: receiving first data collected when the UAV equipped with a laser radar cruises along a cruise path and constructing a three-dimensional real scene model; The laser radar carried by drones can be used to quickly obtain the required data for road mapping. The 3D point cloud data obtained during mapping can be used to build a 3D real scene model. The main steps of building a 3D real scene model include: data preprocessing, modeling and post-processing; data preprocessing includes: denoising, simplification, registration and other processing; modeling can rely on Rhino, BIM, 3Dmax, ContextCapture, and post-processing includes adjustment and optimization. Building a 3D real scene model from 3D point cloud data is an existing mature technology, so I won’t explain it here. Step 3: Compare and analyze the 3D real scene model and the original 3D model to obtain the analysis results.

[0020] The above are all data collection. The main purpose of detection is to obtain an analysis result. By comparing the 3D real scene model with the original 3D model, the current road situation can be determined. The specific comparative analysis includes: taking the area to be detected as the benchmark, the original 3D model and the 3D real scene model are disassembled and reduced respectively; the original 3D model and the 3D real scene model after disassembly are segmented with a segmentation grid and the segmented units are associated (that is, the original 3D model and the units on the 3D real scene model at the same position of the road are associated); the associated units are compared and counted. The above scheme is a unit segmentation analysis method. For the comparative analysis of the associated units, it can be achieved by calculating the similarity of the unit surface images, that is, when the similarity is less than or equal to the preset similarity threshold, the area on the road corresponding to the unit is judged to be abnormal; A point sampling analysis method may also be used, specifically: point sampling is performed on the original three-dimensional model and the three-dimensional real scene model respectively (a point is sampled at every preset interval) to obtain a plurality of associated sampling points (one of the associated sampling points is a sampling point on the original three-dimensional model, and the other is a sampling point at the same position on the three-dimensional real scene model); the associated sampling points are compared and statistically analyzed; the coordinate difference between the sampling points is calculated, and when the coordinate difference is less than or equal to a preset difference threshold, the sampling point is used as an abnormal positioning point, and the adjacent abnormal positioning points are connected to form an abnormal area; The accuracy of the analysis can also be further improved by combining points and surfaces. That is, when segmenting, the segmented units may have the same degree of surface wear. When performing unit detection, there will be misjudgments, but the comprehensive point sampling method can effectively avoid misjudgments. Similarly, for the point sampling analysis method, there may be defects and omissions, which can be supplemented by the unit analysis method. That is, when reanalyzing the results, the abnormal areas of the point sampling analysis and the abnormal areas of the unit analysis are combined to obtain a more accurate abnormal area.

[0021] During actual detection, the detection timing is controlled to avoid waste of manpower and material resources caused by frequent detection. In one embodiment, the road detection method based on drones also includes: statistical analysis of the traffic data of each section of the road, when the detection is triggered, the traffic data of the section is predicted, and the time suitable for detection in the predicted traffic data is selected for detection; wherein the statistical analysis is specifically to assign values ​​to each type of vehicle in the traffic data, and accumulate the assigned values, and when the accumulated value reaches a preset trigger threshold, the trigger detection is determined; wherein the assignment is implemented through a pre-configured score table; the score table is configured with scores for each type of vehicle (for example: 1 point for a car, 3 points for a bus, 4 points for a truck, 5 points for an engineering vehicle, etc.); for the prediction of traffic data Specifically, a pre-trained neural network model can be used to screen historical driving data, obtain data to be analyzed, extract features from the data to be analyzed, and input the extracted features into the neural network model to obtain future traffic trend data; when screening historical driving data, the data to be analyzed is mainly extracted from the historical driving data, and the extraction rules refer to the pre-configured requirements of the neural network model, for example, the neural network model requires that the data of each time period of each day is taken as a separate sampling data, and a preset number (24 or an integer multiple of 24) of continuous sampling data is taken as the data to be analyzed; feature extraction of the data to be analyzed is to extract features from each sampling data in the data to be analyzed, and the extracted features include: the number of vehicles of each type, etc. The future traffic data is essentially composed of the forecast data of each time period in the next N days; first, compare the future weather forecast data (weather type, visibility, etc.), and according to the pre-configured unsuitable situation table, eliminate some forecast data; then calculate the scores of other forecast data, and take the time corresponding to the forecast data with the lowest score and within the pre-configured time range (2 hours to 10 hours after triggering) as the detection time.

[0022] In order to deal with the detection of road damage in sudden road situations, in one embodiment, the road detection method based on drones also includes: obtaining event data on the road (events include: traffic accidents, road maintenance, etc.); evaluating and analyzing the event data to determine the road interference range; accurately positioning and scanning the road interference range to construct a three-dimensional real scene model corresponding to the road interference range; comparing and analyzing the three-dimensional real scene model and the corresponding original three-dimensional model. Among them, the evaluation and analysis of the event data specifically includes: determining the type of event, the rating after the severity evaluation of the event (these are all recorded in the event data), calling the corresponding road interference range parameters (radius of the interference range) from the pre-configured event interference correspondence table, and accurately locating the road interference range based on the location and parameters of the event, wherein the type of event in the event interference correspondence table, the rating after the severity evaluation of the event and the radius value are correspondingly associated.

[0023] In order to accurately capture and analyze the abnormal area, in one embodiment, the bridge detection method based on a drone further includes: The second detection result is determined by analyzing the image of the abnormal area taken by the image acquisition device carried by the drone.

[0024] The image analysis steps are as follows: Matching the image with a standard abnormal image in a pre-configured abnormal database, and when a match exists, obtaining the annotation data associated with the standard abnormal image as a second detection result; The second detection result includes whether there is cracking, the length and width of the crack when there is cracking, whether there is subsidence, the size of the subsidence area, etc.

[0025] In addition, due to the different distribution of the positions of various locations on the road, the pressure time that each location bears when the vehicle is driving is different (the road surface at the intersection will be under the influence of gravity for a long time due to the influence of traffic lights), which is more susceptible to wear and tear than the middle road surface. Most of the road surface is the same as the middle road surface. A global scan of the road and the scanning of the middle road surface reduces the scanning efficiency and wastes energy. In one embodiment, the road detection method based on drones also includes: performing statistical analysis on the predicted data and traffic data before the detection time, and querying a pre-configured risk level table (the risk level and the ratio in the table are associated one-to-one) based on the ratio between the statistical analysis score and the time difference between the detection time and the start time of the traffic data, to determine the risk level. risk level; retrieve the area corresponding to the corresponding risk level from the pre-configured position retrieval model as the area to be detected; wherein the position retrieval model is essentially a three-dimensional model corresponding to the road, and multiple block areas are marked on the three-dimensional model, and each block area is marked as a corresponding group according to the probability of possible abnormality. When the block area is the area to be detected, the block area is scanned; for example: according to the probability of abnormality occurrence from low to high, the risk levels corresponding to the configuration groups are 1, 2, 3, and 4 respectively; when the risk level is 1, retrieve the points of group 1; when the risk level is 2, retrieve the points of groups 1 and 2; when the risk level is 3, retrieve the points of groups 1, 2, and 3; when the risk level is 4, retrieve the points of groups 1, 2, 3, and 4; The steps for determining the marking of each point on the location retrieval model are as follows: extract historical detection data of the same or similar roads, perform statistical analysis on the abnormal areas corresponding to the historical detection data, determine the probability of abnormal occurrence of each area on the three-dimensional model, and group and mark the areas according to pre-configured grouping rules; the probability of abnormal occurrence of an area is the ratio of the total number of abnormal data in the area divided by the total number of abnormal data; the grouping rule is: determine the difference between the minimum probability and the maximum probability, determine multiple equal intervals based on the difference, treat each equal interval as a group and number them in sequence to form groups for marking.

[0026] The present invention also provides a road detection system based on an unmanned aerial vehicle, such as Figure 2 As shown, it includes: a path analysis module 1, a real scene generation module 2 and a comparative analysis module 3; wherein the path analysis module 1 analyzes the original three-dimensional model of the road to obtain a cruising path; the real scene generation module 2 receives the first data collected by the drone equipped with a laser radar when cruising along the cruising path and constructs a three-dimensional real scene model; the comparative analysis module 3 compares and analyzes the three-dimensional real scene model and the original three-dimensional model to obtain an analysis result.

[0027] Among them, the original three-dimensional model is constructed based on the design data or historical surveying data of the road.

[0028] Among them, the path analysis module includes: a segmentation and filling unit, a scanning unit and a correction unit; wherein the segmentation and filling unit performs area segmentation and removes the area that does not need to be detected, and performs area filling to make the area boundary a regular shape as a target to be scanned; the scanning unit scans the target to be scanned with a scanning frame determined by the scanning boundary of the laser radar to obtain a scanning path; the correction unit corrects the scanning path according to the distribution of the area to be scanned within the area scanned by the scanning frame to obtain a cruising path.

[0029] The comparison and analysis module includes: a disassembly association unit and a comparison and statistics unit; wherein the disassembly association unit disassembles the original three-dimensional model and the three-dimensional real scene model based on the area to be detected, and divides the original three-dimensional model and the three-dimensional real scene model after disassembly with a segmentation grid and associates the divided units; the comparison and statistics unit compares and counts the associated units. And / or, the comparison and analysis module includes: a sampling association unit and a comparison and statistics unit, wherein the sampling association unit samples the original three-dimensional model and the three-dimensional real scene model respectively to obtain multiple associated sampling points; the comparison and statistics unit compares and counts the associated sampling points.

[0030] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A road detection method based on drone, characterized in that: include: Analyze the original three-dimensional model of the road to obtain the cruising path; Receiving first data collected by the UAV equipped with a laser radar when cruising along a cruising path and constructing a three-dimensional real scene model; Compare and analyze the 3D real scene model and the original 3D model to obtain the analysis results.

2. The road detection method based on drone as claimed in claim 1, characterized in that: The original 3D model is constructed based on the design data or historical surveying data of the road.

3. The road detection method based on drone as claimed in claim 1, characterized in that: The analysis steps for the original 3D model of the road are as follows: Perform region segmentation and remove areas that do not need to be detected; Filling the area so that the area boundary is a regular shape to serve as the target to be scanned; The scanning frame determined by the scanning boundary of the laser radar is used to scan the target to be scanned to obtain a scanning path; According to the distribution of the area to be scanned within the area scanned by the scanning frame, the scanning path is corrected to obtain the cruising path.

4. The road detection method based on drone as claimed in claim 3, characterized in that: The comparative analysis between the 3D reality model and the original 3D model includes: The original 3D model and the 3D real scene model are dismantled and reduced based on the area to be inspected; The original three-dimensional model and the three-dimensional real scene model after demolition are segmented by segmentation grids and the segmented units are associated; The associated units are compared and statistics are calculated.

5. The road detection method based on drone as claimed in claim 3, characterized in that: The comparative analysis between the 3D reality model and the original 3D model includes: Performing point sampling on the original three-dimensional model and the three-dimensional real scene model respectively to obtain a plurality of associated sampling points; The associated sampling points are compared and counted.

6. A road detection system based on drone, characterized in that: include: A path analysis module, a real scene generation module and a comparative analysis module; wherein the path analysis module analyzes the original three-dimensional model of the road to obtain a cruising path; the real scene generation module receives the first data collected by the drone equipped with a laser radar when cruising along the cruising path and constructs a three-dimensional real scene model; the comparative analysis module compares and analyzes the three-dimensional real scene model and the original three-dimensional model to obtain an analysis result.

7. The drone-based road detection system according to claim 6, characterized in that: The original 3D model is constructed based on the design data or historical surveying data of the road.

8. The drone-based road detection system according to claim 6, characterized in that: The path analysis module includes: a segmentation and filling unit, a scanning unit and a correction unit; wherein the segmentation and filling unit performs area segmentation and removes the area that does not need to be detected, and performs area filling to make the area boundary a regular shape as the target to be scanned; the scanning unit scans the target to be scanned with a scanning frame determined by the scanning boundary of the laser radar to obtain a scanning path; the correction unit corrects the scanning path according to the distribution of the area to be scanned within the area scanned by the scanning frame to obtain a cruising path.

9. The drone-based road detection system according to claim 8, characterized in that: The comparative analysis module includes: a dismantling association unit and a comparative statistical unit; wherein the dismantling association unit dismantles the original three-dimensional model and the three-dimensional real scene model based on the area to be detected, and divides the original three-dimensional model and the three-dimensional real scene model after dismantling with a segmentation grid and associates the segmented units; the comparative statistical unit compares and performs statistics on the associated units.

10. The drone-based road detection system according to claim 8, characterized in that: The comparative analysis module also includes: a sampling association unit and a comparative statistical unit, wherein the sampling association unit performs point sampling on the original three-dimensional model and the three-dimensional real scene model respectively to obtain multiple associated sampling points; the comparative statistical unit compares and performs statistics on the associated sampling points.

Images