Road width measurement method, system and equipment based on point cloud and image, and medium

Abstract

The invention discloses a road width measurement method, system and device based on point cloud and an image, and a medium. The method comprises the following steps: obtaining a road image and corresponding ground area point cloud data; performing road segmentation on the road image, and performing road mask fitting based on a road segmentation result to obtain a fitted road mask; projecting the fitted road mask into the point cloud data of the ground area, and performing filtering processing to obtain point cloud road surface data corresponding to the road mask; screening points on the left and right sides of a road in the point cloud road surface data, and constructing a left and right edge point set; and obtaining a fork road identification result, road width information and a width change curve corresponding to the advancing direction based on the left and right edge point sets. According to the invention, the problems of limitation of a single data source and limited visual field of a sensor are overcome, and more accurate, real-time and stable road width detection can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of road width detection, and in particular relates to a road width measurement method, system, equipment and medium based on point cloud and image. Background Art

[0002] An autonomous driving system refers to an autonomous system that can achieve safe driving without relying partially or completely on human drivers. It mainly includes core technical modules such as environmental perception, path planning, behavioral decision-making, and navigation control. At present, before performing local path planning, the autonomous driving system is usually based on the global route and needs to obtain the maximum drivable area size of the current road first. The extraction of the drivable area is a key technology of the autonomous driving system, which aims to use sensor perception technology to perceive the road environment around the driving vehicle, identify and segment the drivable area in the current driving scenario, and prevent lane deviation or illegal driving.

[0003] There are some limitations in the existing methods for detecting drivable areas of roads: although the map annotation-based method can provide a reference for global planning, due to the static nature of the annotation information, it cannot cope with the dynamic changes in the actual road environment, which can easily cause vehicle planning errors. The sensor perception-based method can obtain the road width in real time, but due to the accuracy of the sensor and the influence of external conditions such as weather, the measurement results may not be stable enough. In addition, the sensor's viewing angle and detection range are limited, especially in complex sections (such as tunnels and bridges), there may be perception blind spots. Although the combination of vehicle-road cooperative technology can make up for the perception blind spots, it relies heavily on external infrastructure, and there is a certain degree of uncertainty in the real-time and accuracy of information transmission. Although the fusion of multi-source data can improve the measurement accuracy by integrating multiple information, the data processing is complex and the cost is high. Summary of the invention

[0004] The purpose of the present invention is to provide a road width measurement method, system, device and medium based on point cloud and image to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a road width measurement method based on point cloud and image, comprising:

[0006] Obtain road images and corresponding ground area point cloud data;

[0007] Performing road segmentation on the road image, and performing road mask fitting based on the road segmentation result to obtain a fitted road mask;

[0008] The fitted road mask is projected onto the ground area point cloud data and filtered to obtain the point cloud road surface data corresponding to the road mask;

[0009] Selecting points located on the left and right sides of the road from the point cloud road surface data to construct left and right edge point sets;

[0010] Based on the left and right edge point sets, a fork road recognition result, road width information, and a width change curve corresponding to the traveling direction are obtained.

[0011] Optionally, the process of acquiring the ground area point cloud data specifically includes:

[0012] Obtain point cloud data corresponding to the road image;

[0013] Extract ground area point cloud data from point cloud data based on ground segmentation algorithm.

[0014] Optionally, performing road mask fitting based on the road segmentation result specifically includes:

[0015] Performing road semantic segmentation on the road image to obtain an original mask of the road area;

[0016] Performing morphological operations on the original mask to obtain a processed mask;

[0017] Perform edge detection on the processed mask to generate a road edge curve, fit the road edge curve based on a weighted curve fitting method, fill the mask based on the fitted road edge curve, and obtain a fitted road mask.

[0018] Optionally, projecting the fitted road mask onto the ground area point cloud data and performing filtering processing specifically includes:

[0019] Projecting the fitted road mask onto the ground area point cloud data based on the extrinsic parameter matrix of the road image to obtain a projection transformation result;

[0020] The point cloud data of the non-road area in the projection transformation result is filtered, and the point cloud data corresponding to the road area in the road mask is retained to obtain the point cloud road surface data.

[0021] Optionally, the process of constructing the left and right edge point sets specifically includes:

[0022] Based on a preset judgment standard and the spatial position of the point cloud in a plane coordinate system, points located on the left and right sides of the road are screened out from the point cloud road surface data to obtain the left and right edge point sets.

[0023] Optionally, the process of obtaining the fork road identification result specifically includes:

[0024] Analyzing the branching morphology of the road area based on the road segmentation result to obtain a first fork road recognition result, wherein the first fork road recognition result includes a preliminary position and a preliminary direction of the fork road;

[0025] The degree of separation of the left and right edge point sets is analyzed to obtain a second fork in the road recognition result, and a final fork in the road recognition result is determined based on the first fork in the road recognition result and the second fork in the road recognition result.

[0026] Optionally, the process of acquiring the road width information specifically includes:

[0027] The center line of the road is calculated based on the left and right edge point sets, a tangent line of the center line at a fixed position is selected at the far end of the road, and a normal line of the tangent line is calculated, and based on the intersection points of the normal line and the edge fitting curves on both sides, the Euclidean distance between the intersection points is calculated to obtain the road width information;

[0028] Calculate the road width information of the road in real time and output the width change curve corresponding to the travel direction.

[0029] A road width measurement system based on point cloud and image, comprising:

[0030] Data acquisition module, used to obtain road images and corresponding ground area point cloud data;

[0031] An image segmentation and road mask fitting module, used to perform road segmentation on the road image, and perform road mask fitting based on the road segmentation result to obtain a fitted road mask;

[0032] The projection transformation module is used to project the fitted road mask into the ground area point cloud data and perform filtering processing to obtain the point cloud road surface data corresponding to the road mask;

[0033] The road width measurement module is used to select points located on the left and right sides of the road in the point cloud road surface data to construct left and right edge point sets; based on the left and right edge point sets, the fork road recognition result, road width information and the width change curve corresponding to the travel direction are obtained.

[0034] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a road width measurement method based on point clouds and images.

[0035] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for measuring road width based on point cloud and image is implemented.

[0036] The technical effects of the present invention are:

[0037] The present invention can overcome the shortcomings of the prior art, achieve more accurate, real-time and stable road width detection, adapt to dynamic road environments, and especially accurately identify road boundaries when the sensor field of view is limited, thereby ensuring safe driving and efficient path planning of the autonomous driving system.

[0038] The present invention combines lidar point cloud data with camera image data to perform multi-dimensional perception and analysis of the road environment, thus overcoming the limitations of a single data source, improving the accuracy and robustness of road width measurement, and reducing the risks brought by complex road sections. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0040] Figure 1 Schematic diagram of the steps of the method described in an embodiment of the present invention;

[0041] Figure 2 Schematic diagram of a road width calculation process based on a combination of an image and a point cloud in an embodiment of the present invention;

[0042] Figure 3 Schematic diagram of the Patchwork++ ground extraction process in an embodiment of the present invention;

[0043] Figure 4 Schematic diagram of the BiseNetV2 road segmentation process in an embodiment of the present invention;

[0044] Figure 5 Schematic diagram of the projection transformation process in an embodiment of the present invention;

[0045] Figure 6 Schematic diagram of comparison between the true value and calculated value of road width in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as limiting the present invention, but should be understood as a more detailed description of certain aspects, features, and embodiments of the present invention.

[0047] It should be understood that the terms described in the present invention are only for describing special embodiments and are not intended to limit the present invention. In addition, for the numerical range in the present invention, it should be understood that each intermediate value between the upper and lower limits of the scope is also specifically disclosed. Each smaller range between the intermediate value in any stated value or stated range and any other stated value or intermediate value in the described range is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded in the scope.

[0048] It will be apparent to those skilled in the art that various modifications and variations may be made to the specific embodiments of the present invention description without departing from the scope or spirit of the present invention. Other embodiments derived from the present invention description will be apparent to those skilled in the art. The present application description and examples are exemplary only.

[0049] The words “include,” “including,” “have,” “contain,” etc. used in this article are open-ended terms, meaning including but not limited to.

[0050] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0051] like Figure 1 - Figure 6 As shown, in this embodiment, a road width measurement method based on point cloud and image is provided, including: obtaining a road image and corresponding ground area point cloud data; performing road segmentation on the road image, and performing road mask fitting based on the road segmentation result to obtain a fitted road mask; projecting the fitted road mask onto the ground area point cloud data and performing filtering processing to obtain point cloud road surface data corresponding to the road mask; screening points located on the left and right sides of the road in the point cloud road surface data to construct left and right edge point sets; and obtaining a fork road recognition result, road width information, and a width change curve corresponding to the traveling direction based on the left and right edge point sets.

[0052] The present embodiment provides a road width measurement method and system based on point cloud and image, which can overcome the shortcomings of the prior art, achieve more accurate, real-time and stable road width detection, adapt to dynamic road environments, and especially accurately identify road boundaries when the sensor field of view is limited, thereby ensuring safe driving and efficient path planning of the autonomous driving system.

[0053] This embodiment aims to combine lidar point cloud data with camera image data to conduct multi-dimensional perception and analysis of the road environment, overcome the limitations of a single data source, improve the accuracy and robustness of road width measurement, and reduce the risks brought by complex road sections.

[0054] The technical solution of this embodiment is as follows:

[0055] S1. Point cloud data acquisition and ground extraction. Use LiDAR to collect point cloud data of road scenes. Point cloud data is transmitted through ROS communication. Ground segmentation is performed on point cloud data to extract point cloud data of the area of ​​interest (i.e., ground area) and filter out non-ground point clouds. Point cloud data of ground area Indicates that P contains N points, each of which contains (x, y, z) three-dimensional spatial coordinates, where the three-dimensional spatial coordinate system takes the center of the laser radar as the origin, the front of the laser radar is the positive x direction, the left is the positive y direction, and the top is the positive z direction;

[0056] S2, image data acquisition and road segmentation, use industrial cameras to collect image I synchronized with point cloud data, image I contains m*n pixels, each pixel is represented by coordinates (u, v). In order to accurately segment the road area, the semantic segmentation model is used to segment the road area:

[0057] S21, semantic segmentation mask generation: perform road semantic segmentation on the image and generate a road surface mask M. The semantic segmentation mask M(u,v)∈{0,1} is used to distinguish between road and non-road areas;

[0058]

[0059] S22, mask post-processing: In order to eliminate noise and improve the accuracy of the road area, this embodiment performs morphological operations (such as dilation and erosion) on the generated mask M (u, v), removes small noise areas, smoothes the mask edge, and obtains a processed mask M' (u, v);

[0060] S23, edge fitting: perform edge detection on the processed mask M'(u,v) to generate the road edge curve C edge , and fit the road edge curve, and fill the mask to obtain the fitted mask M" (u, v);

[0061] C edge ={(u,v)|M'(u,v)=1}

[0062] S3. Perform road mask fitting based on the image segmentation results. Since segmentation and edge detection may produce noise and irregular boundary curves, these edge curves are smoothed and curve fitted:

[0063] S31, edge smoothing: Smoothing the fitted road edge curve to reduce the impact of noise on the edge position and obtain a smooth curve C smooth ;

[0064] S32, curve refinement: refine the edge curve to ensure the smoothness and continuity of the curve;

[0065] S33, edge weighted fitting: Use the weighted curve fitting method to further fit the edge curves on both sides of the road to ensure that the fitted curve can reflect the actual boundary of the road. On the actual road, some parts of the edge may be more critical, so the weighted curve fitting method can be used to assign different weights to the edges of different areas, thereby better reflecting the actual shape of the road. Weighted curve fitting assigns weights ω to the points in the fitting process. i , the final fitted edge curve equation f left (u), f right (u) represents the curve model of the left and right edges, v left =f left (u;ω i ),v right =f right (u;ω i ), where ω i is the weight of each point;

[0066] S4. Project the mask onto the point cloud to filter the point cloud and obtain the point cloud road data. Project the processed mask M" (u, v) onto the ground point cloud data P, filter the point cloud through the projection mask, and obtain the road point cloud P. road . Further segmentation of the ground point cloud data P is achieved to distinguish between road and non-road parts:

[0067] S41, using the laser radar external parameter matrix T to project the pixel points of the image to the point cloud coordinate system. Point P in the laser radar coordinate system i =(x i ,y i ,z i ) corresponds to the pixel point M i ”(u i ,v i ) is obtained through projection transformation:

[0068]

[0069] Among them, K is the intrinsic parameter matrix of the camera, and T is the extrinsic parameter matrix from the laser radar to the camera;

[0070] S42, filter the point cloud data P to obtain the point cloud corresponding to the mask M" (u, v), and obtain the road point cloud P road = {P i ∈P|M i ”(u i ,v i )=1}, where i=1,2,...,N;

[0071] S5. Road width measurement, using segmented point cloud road surface data P road To measure road width:

[0072] S51, point cloud recognition on both sides of the road: set a judgment standard, and filter out the points on the left and right sides of the road according to the spatial position of the point cloud in the plane coordinate system (such as the XOY plane). i =(x i ,y i ,z i )∈P road , based on its horizontal coordinate y i The size of the value determines whether it is on the left or right side of the road centerline; if y i >y center (Assume that y center is the lateral position of the road centerline), then point P i Classified as the left edge point set P left ; if y i <y center , then point P i Classified as the left edge point set P right . We can get the road point cloud P road Identify the edge point set P on both sides of the road point cloud left and P right , corresponding to the point cloud data of the left and right edges of the road respectively;

[0073] S52, centerline calculation: by identifying the edge point clouds P on both sides of the road left and P right , calculate the center point of the road

[0074] S53, edge and center line fitting: through the left and right edge points and the center point of the road, the left and right edge curves of the road are fitted by a third-order curve C left , C right and the center curve C center .

[0075] S54, road width measurement: select the centerline tangent T(x) at a fixed position at the far end of the road, and calculate the normal line N(x) of the tangent. The intersection of the normal line N(x) and the fitting curves of the two side edges is P' left and P' right , calculate the Euclidean distance between the intersection points W(x) = || P' left (x)-P' right (x)|| is the road width.

[0076] S6, fork road processing:

[0077] S61, identifying fork in the road based on image segmentation results: using image road semantic segmentation results M" (u, v) and edge fitting curve f left (u), fright (u), the branching morphology of the road area can be analyzed. The edge information in the image can be used to detect sudden changes in the road morphology to determine whether there is a fork in the road. The location and direction of the fork in the road can also be preliminarily determined by detecting features such as the sudden increase in road width and the drastic changes in edge curves;

[0078] S62. Further identification of forks in the road using point cloud data: In point cloud data, forks in the road are usually represented by separate, smaller point cloud clusters, which can be identified by clustering methods. left and P right The degree of separation is used to judge the fork in the road. If the edge point set suddenly forks in a certain area, it can be inferred that there is a fork in the road.

[0079] S7. Dynamic update and output of road width: Calculate the road width W(x) in real time and output the data, including the width change curve and intersection information.

[0080] S71. Width output: output road width data and its variation curve along the travel direction.

[0081] S72, fork in the road information: output the information of the fork in the road according to the fork in the road identification situation.

[0082] Figure 2 It is a schematic diagram of the road width calculation based on the combination of image and point cloud described in this embodiment. In Part A, 1 is a diagram of the original image data collected by the camera, 2 is a diagram of the road mask effect based on semantic segmentation, and 3 is a diagram of the edge fitting and weighted smoothing effect based on the image mask. In Part B, 4 is a diagram of the original point cloud data collected by the radar, and 5 is a diagram of the point cloud data after the region of interest is extracted. In Part C, 6 is a visualization effect diagram of the point cloud filtered and projected onto the image, and 7 is a diagram of the road width measurement using software. In Part D, 8 is a diagram of the experimental calculation effect of this embodiment.

[0083] Compared with the existing technology, this embodiment can dynamically and in real time detect the road width: by fusing the laser radar point cloud and camera image data, this method can perceive the road environment in real time and dynamically obtain the passable area of ​​the road. Compared with the traditional map-based annotation method, this embodiment can adapt to changes in the actual road environment, such as temporary obstacles, construction areas or weather changes, to ensure that the vehicle can adjust the path planning in time.

[0084] This embodiment has high accuracy and robustness: the laser radar point cloud data has extremely high spatial resolution and can accurately identify road boundaries and obstacle locations. The camera image can supplement rich texture information, such as road markings, traffic signs, etc. In actual tests, the system's road width measurement error in various environments is less than 2%.

[0085] This embodiment can adapt to complex road conditions: This embodiment can handle various road conditions, including school lanes, highways, rural roads, etc. Especially in narrow or complex sections, through the joint perception of sensors and data fusion, the system can accurately calculate the maximum passable width of the vehicle to avoid collision or crossing the line risk.

[0086] This embodiment can reduce sensor dependence: although the perception capability of a single sensor is limited in certain scenarios, for example, the accuracy of lidar is reduced in strong light or rainy and foggy weather, and the camera performs poorly at night or in low visibility conditions, this embodiment combines data from multiple sensors to make up for the shortcomings of a single sensor, improves the robustness and stability of the system, and reduces the impact of the environment on the perception results.

[0087] This embodiment has economic and social benefits: through accurate road width measurement, the autonomous driving system can better perform path planning and obstacle avoidance, improve traffic safety, and reduce accident rates. Especially in complex road sections or areas without accurate map data, the application of this embodiment will significantly improve the decision-making ability of autonomous driving vehicles, reduce dependence on high-precision maps, and reduce development and maintenance costs. In addition, the widespread application of autonomous driving technology will help reduce traffic congestion and improve fuel efficiency, thereby bringing significant social and economic benefits.

[0088] Experiments have shown that the road width measurement error of this embodiment in various environments is stably controlled within 10 centimeters, and it has high measurement accuracy and adaptability.

[0089] Specific algorithm implementation examples of this embodiment include:

[0090] This embodiment uses point cloud to perform preliminary ground extraction, and uses image to obtain road mask; then fits the road mask; then projects the point cloud to the image to extract the point cloud road surface; uses edge point tangent method to calculate road width; finally, processes forks based on point cloud road surface extraction results and calculated road width. This embodiment uses RoboSense's M1-plus laser radar as the point cloud acquisition device and Hikvision industrial camera as the image acquisition device to calculate the road width on campus. According to the process Figure 1 As shown, the following steps need to be taken in this embodiment:

[0091] S1. Use LiDAR to collect point cloud data of road scenes, and use Patchwork++ algorithm to estimate the ground, extract point cloud data of the area of ​​interest (i.e., the ground area), and filter out non-ground point clouds.

[0092] S2. Use industrial cameras to collect road scene image data synchronized with point cloud data, perform road semantic segmentation on the collected image data through BiseNet V2, and preliminarily obtain road masks based on semantic segmentation.

[0093] Then, in order to eliminate noise and improve the accuracy of the road area, this embodiment performs morphological operations (such as dilation and erosion) on the generated mask M(u,v), where the dilation and erosion convolution kernel kernel=(9,9) and the number of operations is set to 1. Small noise areas are removed and the mask edges are smoothed to obtain the processed mask M'(u,v)

[0094] Finally, edge detection is performed on the processed mask M'(u,v) to generate the road edge curve C edge , and fit the road edge curve, and fill the mask to obtain the fitted mask M" (u, v);

[0095] C edge ={(u,v)|M'(u,v)=1}

[0096] S4. The mask is projected onto the point cloud for point cloud filtering to obtain point cloud road surface data.

[0097] First, the processed mask M”(u,v) is projected onto the ground point cloud data P, and the point cloud is filtered by the projection mask to obtain the road point cloud P road . Further segmentation of the ground point cloud data P is achieved to distinguish between road and non-road parts:

[0098] Then, the image pixels are projected into the point cloud coordinate system using the laser radar’s external parameter matrix T. Point P in the laser radar coordinate system i =(x i ,y i ,z i ) corresponds to the pixel point M i ”(u i ,v i ) is obtained through projection transformation:

[0099]

[0100] Among them, K is the intrinsic parameter matrix of the camera, and T is the extrinsic parameter matrix from the laser radar to the camera;

[0101] Finally, the point cloud data P is filtered to obtain the point cloud corresponding to the mask M" (u, v), and the road point cloud P is obtained. road = {P i ∈P|M i ”(u i ,v i )=1}, where i=1,2,...,N;

[0102] S5. Use the segmented point cloud road surface data P road to calculate the road width.

[0103] First, divide the road point cloud P obtained by segmentation in S4 road in the x-axis direction with a division step size of 0.1 m and a division range of 3 < x < 15 to obtain point cloud data Px with an interval of 0.1 m for each. road .

[0104] Then set a judgment criterion, and based on the spatial position of the point cloud in the XOY plane coordinate system, select the points on the left and right sides of the road. For each point P i =(x i , y i , z i ) ∈ P road , judge whether it is on the left or right side of the road center line based on the magnitude of its lateral coordinate y i value; if y i > y center (assuming y center is the lateral position of the road center line), then the point P i is classified into the left edge point set P left ; if y i < y center , then the point P i is classified into the left edge point set P right . Thus, the edge point sets P road and P left on both sides of the road point cloud can be identified from the road surface point cloud P right , corresponding to the point cloud data on the left and right edges of the road respectively;

[0105] Then, through the identified edge point clouds P left and P right on both sides of the road, calculate the center point of the road and perform third-order polynomial curve fitting on the left and right edges and the center point of the road to obtain the left and right edge curves C left , C right , C center .

[0106]

[0107] In the formula, n takes the value of the order of the fitting polynomial.

[0108] Then, at a fixed x-axis position x = 7 m at the far end of the road, calculate the tangent line T(x) of the center line and calculate the normal line N(x) of this tangent line, which is expressed as y - y0 = m perp (x - x0), where m perpis the normal slope. The intersection of the normal line N(x) and the edge fitting curves on both sides is P' left and P' right , calculate the Euclidean distance between the intersection points W(x) = || P' left (x)-P' right (x)|| is the road width.

[0109] This embodiment calculates the road width by segmenting the road point cloud, uses patchwork++ to extract the ground point cloud, then uses BiseNetV2 to perform image road mask segmentation, and performs road point cloud segmentation based on projection transformation and calculates the road width through the road point cloud, which can effectively solve the problem of road width calculation throughout the day.

[0110] A road width measurement system based on point cloud and image, comprising:

[0111] Data acquisition module, used to obtain road images and corresponding ground area point cloud data;

[0112] An image segmentation and road mask fitting module, used to perform road segmentation on the road image, and perform road mask fitting based on the road segmentation result to obtain a fitted road mask;

[0113] The projection transformation module is used to project the fitted road mask into the ground area point cloud data and perform filtering processing to obtain the point cloud road surface data corresponding to the road mask;

[0114] The road width measurement module is used to select points located on the left and right sides of the road in the point cloud road surface data to construct left and right edge point sets; based on the left and right edge point sets, the fork road recognition result, road width information and the width change curve corresponding to the travel direction are obtained.

[0115] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a road width measurement method based on point clouds and images.

[0116] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for measuring road width based on point cloud and image is implemented.

[0117] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A road width measurement method based on point cloud and image, characterized in that: include: Obtain road images and corresponding ground area point cloud data; Performing road segmentation on the road image, and performing road mask fitting based on the road segmentation result to obtain a fitted road mask; The fitted road mask is projected onto the ground area point cloud data and filtered to obtain the point cloud road surface data corresponding to the road mask; Selecting points located on the left and right sides of the road from the point cloud road surface data to construct left and right edge point sets; Based on the left and right edge point sets, a fork road recognition result, road width information, and a width change curve corresponding to the traveling direction are obtained.

2. The road width measurement method based on point cloud and image according to claim 1, characterized in that: The process of acquiring the ground area point cloud data specifically includes: Obtain point cloud data corresponding to the road image; Extract ground area point cloud data from point cloud data based on ground segmentation algorithm.

3. The road width measurement method based on point cloud and image according to claim 1, characterized in that: The road mask fitting based on the road segmentation result specifically includes: Performing road semantic segmentation on the road image to obtain an original mask of the road area; Performing morphological operations on the original mask to obtain a processed mask; Perform edge detection on the processed mask to generate a road edge curve, fit the road edge curve based on a weighted curve fitting method, fill the mask based on the fitted road edge curve, and obtain a fitted road mask.

4. The method for measuring road width based on point cloud and image according to claim 1, characterized in that: The projecting of the fitted road mask onto the ground area point cloud data and filtering the data specifically includes: Projecting the fitted road mask onto the ground area point cloud data based on the extrinsic parameter matrix of the road image to obtain a projection transformation result; The point cloud data of the non-road area in the projection transformation result is filtered, and the point cloud data corresponding to the road area in the road mask is retained to obtain the point cloud road surface data.

5. The road width measurement method based on point cloud and image according to claim 1, characterized in that: The process of constructing the left and right edge point sets specifically includes: Based on a preset judgment standard and the spatial position of the point cloud in a plane coordinate system, points located on the left and right sides of the road are screened out from the point cloud road surface data to obtain the left and right edge point sets.

6. The method for measuring road width based on point cloud and image according to claim 1, characterized in that: The process of obtaining the fork road identification result specifically includes: Analyzing the branching morphology of the road area based on the road segmentation result to obtain a first fork road recognition result, wherein the first fork road recognition result includes a preliminary position and a preliminary direction of the fork road; The degree of separation of the left and right edge point sets is analyzed to obtain a second fork in the road recognition result, and a final fork in the road recognition result is determined based on the first fork in the road recognition result and the second fork in the road recognition result.

7. The method for measuring road width based on point cloud and image according to claim 1, characterized in that: The process of obtaining the road width information specifically includes: The center line of the road is calculated based on the left and right edge point sets, a tangent line of the center line at a fixed position is selected at the far end of the road, and a normal line of the tangent line is calculated, and based on the intersection points of the normal line and the edge fitting curves on both sides, the Euclidean distance between the intersection points is calculated to obtain the road width information; Calculate the road width information of the road in real time and output the width change curve corresponding to the travel direction.

8. A road width measurement system based on point cloud and image, characterized in that: include: Data acquisition module, used to obtain road images and corresponding ground area point cloud data; An image segmentation and road mask fitting module, used to perform road segmentation on the road image, and perform road mask fitting based on the road segmentation result to obtain a fitted road mask; The projection transformation module is used to project the fitted road mask into the ground area point cloud data and perform filtering processing to obtain the point cloud road surface data corresponding to the road mask; The road width measurement module is used to select points located on the left and right sides of the road in the point cloud road surface data to construct left and right edge point sets; based on the left and right edge point sets, the fork road recognition result, road width information and the width change curve corresponding to the travel direction are obtained.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute a road width measurement method based on point cloud and image according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: It stores a computer program, and when the computer program is executed by a processor, a road width measurement method based on point cloud and image as described in any one of claims 1 to 7 is implemented.

Images