Multi-source feature lane edge line identification method and system based on scoring mechanism

Abstract

The invention relates to the technical field of lane edge detection, and provides a multi-source feature lane edge line identification method and system based on a scoring mechanism, and the method comprises the following steps: obtaining and preprocessing road point cloud data, and obtaining road region point cloud data; the color information and the laser intensity information are fused for scoring, and the road area point cloud is processed to obtain a lane line point cloud; for the lane line point cloud, extracting boundary points based on the central main axis and the normal feature jump score to obtain an edge line of each lane line; and performing smooth fitting on the edge line of the lane line to obtain a fitted lane edge line. According to the method, the color and laser intensity double-scoring function is constructed, the intensity threshold is adaptively determined through the OTSU algorithm, and the intensity threshold and RGB color difference are subjected to weighted fusion, so that the recognition accuracy and continuity of the lane marking points are effectively improved, and the problems of extraction fracture, misrecognition and the like in a traditional method are solved.

Description

Technical Field

[0001] The present invention relates to the technical field related to lane edge detection, and in particular to a multi-source feature lane edge line recognition method and system based on a scoring mechanism. Background Art

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

[0003] With the rapid development of transportation infrastructure, the number and scale of highway reconstruction and expansion projects continue to grow, and the requirements for road structural accuracy and safety continue to increase. As a key information carrier of road geometric features, lane edge lines not only define the spatial boundaries of lanes but also have important application value in areas such as digital modeling of road assets, intelligent driver assistance systems, and transportation infrastructure operation and maintenance. In particular, in the maintenance and reconstruction scenarios of multi-lane highways, the accurate extraction and geometric description of lane edge lines have become a critical support for high-precision map construction and road condition perception. With the development of mobile laser scanning (MLS) technology, point cloud data has become one of the core data sources for collecting highway geometric information due to its high density and three-dimensional spatial expression capabilities. However, how to efficiently and accurately extract multi-lane edge lines from complex highway point clouds remains a technical challenge in the industry.

[0004] Existing lane edge line extraction suffers from the following problems: On the one hand, traditional extraction methods, which use a laser reflection intensity threshold, struggle to cope with uneven reflection strength caused by varying road surface materials, lighting conditions, and reflective interference, often leading to broken or blurred edge line extraction, or even misjudgment. On the other hand, point cloud data inherently suffers from noise, uneven density, and small-scale perturbations, which severely impact the accurate positioning of edge points and curve continuity, leading to glitches and drift during line segment fitting. Furthermore, most current algorithms focus solely on the overall outline of lane lines or single-sided extraction, making it difficult to meet structural-level modeling requirements. Summary of the Invention

[0005] To solve the above problems, the present invention proposes a lane edge line recognition method and system based on a multi-source feature scoring mechanism. A dual scoring function of color and laser intensity is constructed. The intensity threshold is adaptively determined by the OTSU algorithm and weighted fused with the RGB color difference. This effectively improves the recognition accuracy and continuity of lane marking points, overcoming the problems of extraction breaks and misidentification in traditional methods.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A first aspect of the present invention provides a multi-source feature lane edge recognition method based on a scoring mechanism, comprising the following steps: Obtain road point cloud data and perform preprocessing to obtain road area point cloud data; The weighted score is calculated by fusing color information and laser intensity information, and the road area point cloud is processed to segment the lane line from the background to obtain the lane line point cloud. For lane point clouds, the center axis and normal feature jump score are used to score the normal feature jump based on the intensity and color changes in the normal direction. The boundary points are extracted to obtain the edge line of each lane line. A smooth fitting is performed on the edge line of the lane line to obtain the fitted lane edge line.

[0007] A second aspect of the present invention provides a lane edge recognition system based on a multi-source feature scoring mechanism, comprising: A road point cloud extraction module is configured to acquire road point cloud data and perform preprocessing to obtain road area point cloud data; The lane extraction module is configured to fuse color information and laser intensity information to calculate a weighted score, process the road area point cloud, and segment the lane lines from the background to obtain a lane point cloud; The multi-lane edge line extraction module is configured to extract the edge line of each lane line based on the central main axis and normal feature jump score. The normal feature jump score is performed based on the intensity and color changes in the normal direction with the central main axis as the reference. The multi-lane edge line curve fitting module is configured to perform smooth fitting on the edge lines of the lane lines to obtain fitted lane edge lines.

[0008] A third aspect of the present invention provides a multi-source feature lane edge line recognition system based on a scoring mechanism, comprising a point cloud data acquisition device and a processor; A point cloud data acquisition device, used to obtain road point cloud data; The processor is configured to execute the steps in the above-mentioned multi-source feature lane edge line recognition method based on the scoring mechanism; Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention fuses the color information in the point cloud data with the laser reflection intensity characteristics to extract lane lines. By weighted fusion of color score and intensity score, lane line points are screened out, realizing the joint analysis of point cloud data in the color domain and intensity domain, effectively overcoming the misjudgment problem caused by uneven color distribution and abnormal intensity.

[0009] To address the issues of edge jitter and misjudgment caused by blurred boundaries and sparse point clouds during lane marking edge extraction, this paper proposes a dual-edge extraction method based on the central principal axis and normal feature jump scoring. This method uses the central principal axis curve as a benchmark and analyzes intensity and color changes in the normal direction to accurately locate the left and right boundary points, effectively improving the stability and geometric accuracy of boundary point extraction.

[0010] The advantages of the present invention and its additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.

[0012] Figure 1 is a flow chart of a lane edge line recognition method according to embodiment 1 of the present invention; Figure 2 1 is an overall architecture diagram of the lane edge line recognition method according to Example 1 of the present invention; Figure 3 is a flow chart of a method for lane line extraction according to embodiment 1 of the present invention; Figure 4 This is a schematic diagram of lane names according to Example 1 of the present invention; Figure 5 1 is a schematic diagram of an example road sign clustering result of Example 1 of the present invention; Figure 6 Schematic diagram of lane line point cloud extraction effect in Example 1 of the present invention; Figure 7 1 is a schematic diagram of the lane edge line extraction effect of Example 1 of the present invention; Figure 8 This is a lane edge line fitting effect diagram of Example 1 of the present invention; FIG9 is a curve after vectorization processing according to Example 1 of the present invention; Figure 10 2 is a schematic diagram of points extracted during the lane edge line boundary point recognition process according to Example 1 of the present invention; Figure 11 is a lane edge line boundary point recognition result diagram of Example 1 of the present invention; DETAILED DESCRIPTION The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0013] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0014] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments of the present invention and the features in the embodiments can be combined with each other. The embodiments will be described in detail below with reference to the accompanying drawings.

[0015] In response to the problems of difficulty and poor continuity in the extraction of lane lines and edge lines in road point clouds, the present invention proposes a high-precision lane edge line identification method based on a multi-source feature CRI-Score scoring mechanism. In terms of lane marking line extraction, the innovative CRI-Score fusion method is proposed, which comprehensively utilizes RGB color and laser intensity features. By constructing a scoring function and weighted fusion, it effectively improves the segmentation accuracy of lane lines and road surface backgrounds, and solves the problems of discontinuity and high misjudgment rate of traditional intensity methods. In terms of edge line extraction, an orthogonal normal is constructed in combination with the central main axis of the lane line, and sampling is performed in the normal direction and the color and intensity jump scores are calculated, so as to stably extract the left and right boundary points. This method adapts to complex situations such as lane line width changes and sparse point clouds, and improves the continuity and accuracy of edge extraction. The algorithm of the present invention has a clear structure, high precision, and strong adaptability, which meets the engineering application requirements of high-precision extraction and expression of lane edge lines in complex multi-lane scenes. It is explained below with specific embodiments; Example 1 In the technical solutions disclosed in one or more embodiments, Figures 1 to 11 As shown in FIG, a lane edge recognition method based on a multi-source feature scoring mechanism includes the following steps: Step 1: Obtain road point cloud data and preprocess it to obtain road area point cloud data; Step 2: Calculate a weighted score by fusing color information and laser intensity information, process the road area point cloud, segment the lane line from the background, and obtain the lane line point cloud; Step 3: For the lane point cloud, score the normal feature jump based on the central axis and normal feature jump. With the central axis as the reference, score the normal feature jump based on intensity and color changes in the normal direction, and extract the boundary points to obtain the edge line of each lane line. Step 4: Perform smooth fitting on the lane edge line to obtain the fitted lane edge line; The method in this embodiment utilizes the RGB color information and laser reflection intensity characteristics in point cloud data to propose a CRI-Score (Colour-Reflectivity-Intensity Score) fusion algorithm for lane line extraction. By weightedly fusing color scores and intensity scores, multi-source features are integrated to filter out lane line points. This allows for joint analysis of point cloud data in both the color and intensity domains, effectively overcoming misjudgments caused by uneven color distribution and abnormal intensity.

[0016] Furthermore, to address the issues of edge jitter and misjudgment caused by blurred boundaries and sparse point clouds during lane marking edge extraction, this example proposes a dual-edge extraction method based on the central principal axis and normal feature jump scoring. This method uses the central principal axis curve as a benchmark and analyzes intensity and color changes in the normal direction to accurately locate the left and right boundary points, effectively improving the stability and geometric accuracy of boundary point extraction.

[0017] RGB color is a method of expressing color by mixing three basic colors: red, green, and blue. In step 1, point cloud data can be collected through the MLS system. MLS (Mobile Laser Scanning) is a system that installs lidar equipment on a mobile carrier (such as a car or drone) to collect three-dimensional data.

[0018] The preprocessing of point cloud data can be realized by filtering, denoising and removing interference points in sequence to obtain point cloud data of the road area; Optionally, the filtering denoising method adopts a statistical filtering-based method: a statistical filter is used to analyze the neighborhood distance distribution of each point, and points whose distance mean deviation exceeds a set threshold are eliminated, and outliers are automatically determined based on the set threshold; The filtering denoising method of the above embodiment can remove random, isolated, and unstructured noise points such as dust and raindrops, significantly improving the overall stability and continuity of point cloud data. Optionally, interference point removal can be performed using a region growing method, which includes the following steps: Step 11, initial seed point selection: select a plane point whose angle between the surface normal and the Z axis is less than the angle setting value and whose height is lower than the height setting value as a seed; In three-dimensional space, each point in a point cloud can be estimated to have a local fitting plane through its neighboring points (i.e., surrounding points). This local fitting plane is assumed to be a local model of the surface of the object to which the point belongs. The "surface normal" refers to the unit vector perpendicular to the local fitting plane, which characterizes the spatial tilt and direction of the surface at the point. It can reflect the spatial tilt and direction of the surface at the point. The Z axis is the vertical direction, i.e., the "upward" or "perpendicular to the ground" direction. In point clouds, the direction of the Z axis is usually used to indicate height or altitude. Set two thresholds: one for the angle between the surface normal and the vertical direction (Z axis), and the other for the point's height. When the angle between the surface normal and the Z axis is less than the set angle threshold, and the point's Z coordinate (height) is lower than the set height threshold, it is set as a seed point.

[0019] Step 12: Starting from the seed point, using normal similarity and distance between points as growth conditions, adjacent points that meet the growth conditions are added as growth points to the growth region for expansion. This process is repeated until further growth is impossible, and finally a complete road point cloud region with similar geometric characteristics and continuity is extracted. Normal similarity, that is, the angle between the surface normal of the plane where the seed point is located and the surface normal of the current point is less than the set threshold; The distance between points, that is, the spatial distance from the seed point is less than the set distance; In this embodiment, based on the region growing algorithm, combined with the flat geometric characteristics of the road surface and the consistency of the normal direction, interference such as vegetation and buildings is eliminated to achieve efficient segmentation of the road area point cloud.

[0020] To address the difficulties in lane line extraction, inaccurate results, and discontinuities caused by the uneven distribution of point cloud intensity features, a filtering extraction method that fuses RGB color and intensity is employed in step 2 of this embodiment. Lane lines typically have high-brightness color features (such as white or yellow) and strong reflective intensity. Therefore, scoring functions are constructed for both color and intensity to quantify each point's proximity to the ideal lane line feature. The color score is calculated based on the difference between the point's RGB color and the target color, while the intensity score is based on the OTSU algorithm, adaptively determining the threshold for reflective intensity and highlighting high-intensity areas. Finally, a weighted fusion of the two scores is used to generate a comprehensive score, and a threshold is then set to filter out high-confidence candidate lane line points, achieving the organic integration and complementary extraction of color and intensity information.

[0021] Among them, the OTSU algorithm (translated as: Otsu threshold segmentation algorithm) is an adaptive global threshold segmentation method based on image grayscale histogram analysis.

[0022] Step 2 is the lane line extraction process. The road area point cloud data extracted in step 1 is processed to identify lane lines. The method uses a scoring method that integrates color information and laser intensity information (also known as the CRI-Score scoring method) to process the road area point cloud, segmenting the lane lines from the background, and obtaining a lane line point cloud. The method includes the following steps: Step 21: Calculate color score: For the road area point cloud data, extract the RGB color of each point cloud point, calculate the color difference between the RGB color of each point cloud data and the lane target color, and obtain the color score; Among them, the lane target color is the standard lane line color, which is usually white or yellow; Each point in the point cloud The color is represented by , define the target lane line color, specifically including: white lane line , yellow lane lines ; Step 211: Calculate the color difference between the RGB color of each point and the lane target color using Euclidean distance; Traverse and extract the RGB color information of each point cloud point in the regional road point cloud data, and use the Euclidean distance to separate the color differences. Defined as: ; in, include and , The smaller the color difference value, the closer the color of the point is to the lane line color.

[0023] Step 212: Calculate a color score based on the ratio of the color difference to the maximum color difference in the RGB space; The color difference is converted into a color score. The score range is [0, 1]. The higher the score, the closer the color is to the lane line. The color score formula is: ; in: is the maximum possible distance in RGB space; Step 22: Calculate the intensity score: For the road area point cloud data, map the reflection intensity of the point cloud to a grayscale value, and use the OTSU algorithm to adaptively determine the intensity segmentation threshold; construct a normalized scoring function based on the relationship between the intensity value of each point and the intensity segmentation threshold. , we get the strength score of each point cloud’s proximity to the lane line feature. The process is as follows: Step 221: for the road area point cloud data, map the point cloud reflection intensity into grayscale values and perform partitioning based on a threshold method; Specifically, the reflection intensity of the point cloud is mapped to grayscale values, and partitioned based on the threshold method to divide the point cloud into high-intensity areas and low-intensity areas to distinguish potential lane lines from non-lane areas; For example, assuming that the grayscale value of a grayscale image ranges from 0 to 255, the image can be divided into two regions by setting a threshold: one region contains pixels with grayscale values less than or equal to the threshold, and the other contains pixels with grayscale values greater than the threshold; each region is regarded as a class; Step 222: Calculate the inter-class variance between each partition and extract the maximum inter-class variance; Calculate between-class variance , the formula is as follows: ; Where T is the candidate threshold; L is the total number of gray levels; Grayscale The probability distribution of Calculate the maximum inter-class variance. The goal of OTSU threshold segmentation is to find the inter-class variance. Maximum threshold , the maximum between-class variance calculation formula is as follows: ; Step 223: Calculate the intensity score of each point in the point cloud based on the reflection intensity and the maximum inter-class variance of each point. According to the threshold of OTSU output, a scoring function based on relative intensity difference is designed , and linearly normalize to the interval [0,1]: ; in, Indicates the strength of the current point, Indicates the maximum intensity of the point cloud in the road area; If the intensity of the point cloud data point Above threshold , then it enters the scoring area, the higher the intensity, the closer it is to 1; if it is below the threshold , directly set to 0 and not participate in the scoring; Step 23: Perform weighted fusion of the color score and intensity score of each point in the point cloud to obtain the fusion score. The calculation formula is as follows: ; in: and is the set weight, and ; Step 24: Screen candidate points: Based on the threshold method, select the points in the point cloud whose fusion score is greater than the set threshold as the marker line point cloud; Set a threshold , only retain the points whose fusion score is higher than this threshold, traverse all road point clouds, The points in the lane will be classified as marked line point clouds, thus achieving accurate extraction of lane markings.

[0024] Step 25: Marker line point cloud segmentation: Cluster the filtered marker line point cloud based on the Euclidean distance of the point cloud to separate the lane lines from the internal road markings to obtain the lane line point cloud; This step uses the Euclidean clustering method to group the point cloud data to extract the regional features of lane markings and other lane markings. This method clusters the point cloud based on the Euclidean distance, generates multiple cluster regions, and selects valid clusters by setting an appropriate cluster radius and minimum cluster point number to separate lane markings from the road interior. The middle area of the lane is the area between the two lane lines. like Figure 5 The figure shows a schematic diagram of road marking clustering. The markings on the road include lane lines and markings in the area between lane lines. Figure 5 The two blue lines are lane lines, and the green lines and green arrows are internal road markings. Through clustering in step 25, the internal road markings can be filtered out, and the point cloud data of the blue lane lines, i.e., the lane line point cloud, is retained. Figure 5 (a) is the road marking including lane markings and internal markings. Figure 5 (b) in the middle is the lane marking. Figure 5 (c) Road interior markings; In step 2 of this example, a filtering extraction method that fuses RGB color and intensity is used to accurately extract lane marking point clouds. A dual scoring function for color and laser intensity is constructed. The intensity threshold is adaptively determined using the OTSU algorithm and weightedly fused with RGB color differences. This effectively improves the recognition accuracy and continuity of lane markings, achieving high-precision segmentation of the lane marking point cloud from the road background. Compared to traditional lane marking extraction, this example, after determining the optimal segmentation threshold by maximizing the inter-class variance, weighted fusion of RGB color information is then performed, effectively addressing issues such as discontinuous and inaccurate extraction caused by uneven intensity distribution.

[0025] Specific extraction examples, such as Figure 6 As shown, the effect changes in the process of extracting lane lines are shown. Figure 6 (a) is the point cloud data after filtering and denoising in step 1, i.e., the road surface point cloud data; Figure 6 (b) is the road area point cloud after removing interference points based on the region growing method in step 1; Figure 6 (c) is the lane line point cloud obtained by executing step 2; In step 3, the multi-lane edge line extraction step is performed, based on the obtained lane lines, the lane line edges are identified; Figure 6 (c) Lane line point cloud, and Figure 5 The blue line in the figure is the lane line point cloud; Figure 5 There are two lane lines, each lane line has a certain width, so each lane line has two edge lines. Step 3 is to identify the edge lines of the lane lines; For lane point clouds, based on the central principal axis and normal feature jump score, the method uses the central principal axis as a reference, performs normal feature jump score based on intensity and color changes in the normal direction, and extracts boundary points to obtain the edge line of each lane line, including the following steps: Step 31: PCA central principal axis extraction: Use principal component analysis to determine the direction of the lane line and extract the central principal axis of the lane line; Step 311: Use principal component analysis (PCA) to estimate the direction of the lane point cloud set, and convert the first principal component direction vector As the overall direction of the lane line, the direction vector Fixed step size Perform projection slicing to obtain point cloud slices; Each lane marking point cloud appears as a strip-like structure with a width of approximately 0.15 meters. The morphological centerline needs to be extracted from this set of points as a reference for subsequent normal analysis and boundary extraction.

[0026] Optionally, principal component analysis (PCA) can be used to analyze the lane point cloud collection. Perform direction estimation and take the first principal component direction vector As the overall direction of the lane line. Then in the direction vector Fixed step size Projection slices are performed, each slice represents the transverse section of the lane line, and the transverse section and direction vector Vertically, the points contained in the kth slice are recorded as .

[0027] Step 312: Calculate the average horizontal coordinate in each projection slice as the center point to obtain an ordered center point set. ; The calculation formula of the horizontal average coordinate is: ; Step 313: Perform cubic B-spline curve fitting on the center point set to obtain the central axis o of the lane line, which is expressed as , is the normalized arc length, and the parameter s takes the value of [0,1].

[0028] Step 32: Determine the normal direction of the central axis o of the lane line and perform normal sampling to obtain candidate boundary points; Step 321: construct the normal direction of the central axis o of the lane line; The normal direction of the lane line is the orthogonal direction of the central main axis. For each point s in the equation, calculate its tangent direction and construct the corresponding unit normal vector. First, take the derivative about the central principal axis, then rotate the tangent vector 90 degrees counterclockwise to obtain the unit normal vector. The formula is as follows: ; ; in, is the unit vector in the tangent direction; is the unit normal vector in the normal direction, is the tangent direction angle.

[0029] Step 322: At each center point, sample multiple points along the normal direction to obtain left and right boundary candidate points; Given maximum extension distance , sampling interval , generate M sampling points in the left and right directions: The sampling formula for the left boundary sampling point is: ; The sampling formula for the right boundary sampling point is: ; in, ， ; The left sampling point and the right sampling point constitute the left and right boundary candidate points. Each candidate point Each corresponds to a position in space, and the RGB color information and intensity information of the point can be obtained through mapping, such as Figure 10 and Figure 11 As shown, the center point on the central axis o is represented by a green point, and the other points in the normal direction of the green point, including blue and red points, are candidate boundary points; Step 33: For the obtained candidate boundary points, perform feature jump scoring based on intensity and color changes, and select the point with the highest score as the boundary point; that is, compare each point one by one to see if there is a sudden change in color and intensity. The point where a sudden change occurs is the dividing point between the road surface and the edge line, such as Figure 10 In the figure, the red points are the boundary points; At the edge of the lane line, the road surface is gray and the lane line is white or yellow because of the obvious difference between the road surface and the lane line. The RGB color and intensity information of the point cloud points at the edge are prone to sudden changes. In addition, edge points may be mis-extracted, broken, or noisy, making them prone to misjudgment based on distance alone. In this embodiment, a boundary scoring function is innovatively constructed Perform auxiliary screening and calculate the left and right sampling points respectively, then the boundary scoring function is: ; in, Represents the color vector, the first candidate point in the boundary Color vector of sampling points; Indicates the first candidate point in the boundary The reflection intensity of each sampling point; and are the weights of the reflection intensity and color change scores, is the Euclidean distance between two RGB color vectors; +1, j and j-1 are adjacent points sampled in the normal direction of the central main axis o; Specifically, such as Figure 10 and Figure 11 As shown in the figure, M points are extracted from the normal line in the left and right directions of the central axis, and the numbers are increased in sequence from the central axis, such as j-1, j, j+1. If the outermost point is j, then j+1 does not exist and its data is 0. In this case, the scoring function value is the largest, that is, a mutation occurs. For illustration purposes, Figure 10 In the example, j+1 is drawn outside the lane edge line; Set sampling points Signed distance relative to the normal , where the distance to the left point is negative and the distance to the right point is positive. In order to avoid interference with judgment due to color mutation or abnormal reflection intensity in the area near the normal line, the scoring function is selected from the first K sampling points farthest from the normal line on each side. The largest one is taken as the boundary point, as follows: Left boundary point extraction, the formula is: ; ; The right boundary point extraction formula is: ; ; in, represents the jth boundary candidate point; is the offset distance of the sampling point relative to the main axis of the lane center, On the left side of the normal, On the right side of the normal; is the boundary scoring function; Represents the index set of the first K points on the left that are farthest from the normal line; Represents the index set of the first K points on the left and right sides that are farthest from the normal; Indicates selecting the first K points from the points that meet the conditions by sorting by distance; 、 Indicates the sampling point number finally selected as the left and right boundary points.

[0030] The default range of the window range K is 3, which can be adjusted to adapt to different point densities; the final selection 、 As the left and right boundary points.

[0031] Step 34: Cluster and sort the boundary points on both sides of all lane lines to obtain the inner and outer edge point clouds of the two lane lines, i.e., the lane edge lines.

[0032] like Figure 7 As shown in the figure, the point cloud of the two lane lines is obtained through step 2, which is in the form of points; Figure 11 The left and right boundary points in the image are the lane edges that form a lane line. The middle yellow line is used as the dividing line to separate the lane into right and left lanes. The right edge of the right lane and the left edge of the left lane are called the outer edge lines. The left edge of the right lane and the right edge of the left lane are called the inner edge lines.

[0033] In the above implementation of this embodiment, a dual boundary extraction method based on the central principal axis and normal feature jump score is used. This method uses the fitting of the central principal axis curve as a benchmark, analyzes the intensity and color mutations in the normal direction, accurately locates the left and right boundary points, and effectively improves the stability and geometric accuracy of boundary point extraction. In order to solve the problem of lane marking line boundary recognition, the central principal axis of the lane line point cloud is first extracted by direction slicing and average position calculation, and its derivative is used to construct the orthogonal normal direction. Boundary candidate points are sampled at fixed intervals in the normal direction, and a boundary scoring function is constructed by combining intensity and color gradients to extract the boundary points with the most obvious mutations in each normal direction. Finally, by fitting the left and right boundary point sets, a continuous and smooth lane line bilateral contour is obtained, which can adapt to various practical situations such as lane line bending, subtle width changes, and missing boundaries.

[0034] In step 3, a normal direction is constructed based on the central principal axis. Boundary sampling and feature analysis along this normal direction enable high-precision identification and contour fitting of the left and right lane edge points. Unlike traditional contour extraction methods, this method leverages the structural characteristics of lane lines and employs a combined strategy of "central principal axis + normal extension + feature transition scoring." This strategy enables stable and continuous boundary extraction even in situations with limited width, unclear boundaries, or uneven point cloud distribution.

[0035] In step 4, curve fitting of multiple lane edges is performed: curve fitting is performed on the lane edge lines to obtain fitted lane edge lines. Specifically, the lane edge lines extracted in step 3 are used in combination with the improved cubic B-spline curve fitting algorithm and the least squares curve approximation method to perform smooth fitting on the lane edges to generate a continuous curve, namely the fitted lane edge lines, including the following steps: Step 41, data point parameterization: parameterize the point cloud data points of the lane edge line using a centripetal parameterization method to obtain a parameter value for each point cloud data point so that the point cloud is reasonably distributed in the parameter space; The centripetal parameterization method is used to parameterize the point cloud data points of the lane edge line. The process is as follows: Step 411: Calculate the chord length between adjacent data points based on the coordinate values of the point cloud: ; ; Step 412: Use the accumulated chord length as the parameter value of the centripetal parameterization, and the calculation formula is: ; in, It is data points; is the total number of points; Indicates the The chord length between a data point and the previous data point; represents the cumulative chord length between all data points; Represents a data point The corresponding parameter value; is the adjustment parameter, which is the set value.

[0036] Step 42: Construct a node vector, determine the starting node and the ending node, and use the point cloud data points between the starting node and the ending node as the intermediate points. With the starting node as the reference, calculate the position of each intermediate point through the parameterized value of the data point. Step 43: Normalize the obtained parameter values so that all parameter values are distributed in the interval between; Step 44: Select control points, establish a relationship matrix between the control points, determine the initial positions of the control points by solving the corresponding linear equations, and generate a preliminary B-spline curve. The linear equations are as follows: ; Where, Represents the mth lane edge line point cloud data point, such as Figure 10 The red dot in represents the nth control point, which is a data point on the lane edge line selected according to the distance; Represents the B-spline basis function value.

[0037] Step 46: For the obtained B-spline curve, the least squares method is used to further optimize the positions of the control points so that the final generated curve is close to the true edge line, and the fitted lane edge line is obtained, thereby achieving an accurate expression of the edge line morphological characteristics.

[0038] The B-spline-least squares fitting model is: ; The least squares optimization objective function is: ; Where, It represents the fitted value of parameter 𝑡 under the influence of the 𝑖th sample point and its surrounding control points; The weight coefficient corresponding to the kth control point is determined by least squares; is the fitting residual; is the kth control point; (t) is the n-order B-spline basis function; ; This step improves the point cloud data of each lane edge obtained from lane line extraction in Step 3 by using a cubic B-spline curve and a least-squares fusion algorithm. This achieves high-precision smooth fitting and vectorized output of multiple lane edge lines. Unlike traditional curve fitting methods, this step incorporates a centripetal parameterization method to optimize control point weights and smoothing constraints, effectively addressing drift and glitches on curved sections. Figure 8 (a) shows the lane edge line effect after fitting, where the blue curve represents the lane edge line obtained by fitting. Figure 7 The points are connected into a curve; Figure 8 (b) shows the fit between the lane edge line and the road marking line after fitting, where the red horizontal line segment between the two lane edge lines represents the lane line point cloud data obtained in step 2.

[0039] To verify the accuracy of the curve fitting, 20 fitting points were randomly selected from the test section and their coordinates were compared with the coordinates of the corresponding original lane points. As shown in Table 1, the results show an average error of less than 1 cm, a high degree of fit between the fitting curve and the original point cloud, and no glitches. This fully demonstrates the high accuracy and good stability of this method for lane extraction.

[0040] Table 1 Comparison table of the fitted coordinates and original coordinates of the test section;

[0041] The fitted lane edge line outputted in step 4 is discrete data consisting of points, i.e., in point cloud data format. Furthermore, a vectorization step is included to convert the discrete point data into a vector form with a clear topological and geometric structure, and output it as a curve. Vectorization means that the curve can be exported to DXF format instead of point cloud format. After vectorization, it can be used across platforms and can be opened with existing drawing software; Optionally, vectorization uses a KD-Tree (K-Dimensional Tree) spatial index to topologically sort the fitted lane edge lines, i.e., the point cloud curves. Accurate point cloud adjacency relationships are established based on neighboring point searches to obtain vectorized lane lines, ensuring topological connectivity of the vector geometry.

[0042] The lane edge detection and fitting method of this embodiment has demonstrated excellent performance in practical highway applications. It not only significantly improves the accuracy and efficiency of multi-lane edge extraction, but also ensures a smooth transition between new and old road sections in terms of width, slope, and overall geometry, effectively reducing errors during construction.

[0043] In order to verify the effectiveness of the lane edge line recognition method in this embodiment, a simulation experiment was carried out. For the lane edge line point cloud extracted in step 3, the existing RANSAC (RANdom SAmple Consensus) straight line fitting algorithm was used for fitting comparison analysis. Figure 9 (a) shows the overall effect of RANSAC straight line fitting, and Figure 9 (b) is a local enlarged view. It can be seen from the figure that RANSAC straight line fitting has obvious drift and "burr" phenomenon in curved road sections, and the fitting accuracy in curved sections is poor. Comparing Figure 9 with Figure 8 Comparing the lane edge line fitting effects of the method of this embodiment with those of the method of this embodiment, it can be seen that the curvature of the edge line obtained by the method proposed in this embodiment changes more gently, without obvious drift, and has better smoothness and fitting accuracy.

[0044] Example 2 Based on Example 1, this embodiment provides a multi-source feature lane edge recognition system based on a scoring mechanism, including: A road point cloud extraction module is configured to acquire road point cloud data and perform preprocessing to obtain road area point cloud data; The lane extraction module is configured to fuse color information and laser intensity information to calculate a weighted score, process the road area point cloud, and segment the lane lines from the background to obtain a lane point cloud; The multi-lane edge line extraction module is configured to extract the edge line of each lane line based on the central main axis and normal feature jump score. The normal feature jump score is performed based on the intensity and color changes in the normal direction with the central main axis as the reference. The multi-lane edge line curve fitting module is configured to perform smooth fitting on the edge lines of the lane lines to obtain fitted lane edge lines.

[0045] Furthermore, it also includes a vectorization module, which uses KD-Tree spatial index to topologically sort the fitted lane edge lines, establishes accurate point cloud adjacency relationships based on neighboring point search, and obtains vectorized lane lines.

[0046] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation processes are the same, which will not be repeated here.

[0047] Example 3 Based on Example 1, this embodiment provides a multi-source feature lane edge line recognition system based on a scoring mechanism, including a point cloud data acquisition device and a processor; A point cloud data acquisition device, used to obtain road point cloud data; The processor is configured to execute the steps in the multi-source feature lane edge line recognition method based on the scoring mechanism described in Example 1.

[0048] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0049] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A multi-source feature lane edge recognition method based on a scoring mechanism, characterized by: The steps include: Obtain road point cloud data and perform preprocessing to obtain road area point cloud data; The weighted score is calculated by fusing color information and laser intensity information, and the road area point cloud is processed to segment the lane line from the background to obtain the lane line point cloud. For lane point clouds, the center axis and normal feature jump score are used to score the normal feature jump based on the intensity and color changes in the normal direction. The boundary points are extracted to obtain the edge line of each lane line. A smooth fitting is performed on the edge line of the lane line to obtain the fitted lane edge line.

2. The multi-source feature lane edge recognition method based on a scoring mechanism according to claim 1, characterized in that: The preprocessing of point cloud data includes filtering denoising and interference point removal in sequence: the filtering denoising method adopts a statistical filtering-based method, and the interference point removal adopts a region growing-based method.

3. The lane edge recognition method based on a multi-source feature scoring mechanism according to claim 1, characterized in that: The method for calculating a weighted score by fusing color information and laser intensity information, processing a road area point cloud, and segmenting lane lines from the background to obtain a lane line point cloud includes the following steps: For the road area point cloud data, the RGB color value of each point cloud point is extracted, and the color difference between the RGB color value of each point and the lane target color is calculated to obtain the color score; For road area point cloud data, the reflection intensity of the point cloud is mapped to grayscale values, and the OTSU algorithm is used to adaptively determine the intensity segmentation threshold. A normalized scoring function is constructed based on the relationship between the intensity value of each point and the intensity segmentation threshold to obtain an intensity score for each point cloud's proximity to the lane line feature. Perform weighted fusion of the color score and intensity score of each point in the point cloud to obtain a fusion score; Based on the threshold method, the points of the point cloud with a fusion score greater than the set threshold are screened out as the marker line point cloud; The filtered marking line point cloud is clustered based on the Euclidean distance of the point cloud, and the lane lines and the point clouds marked between the lanes are separated to obtain the lane line point cloud.

4. The multi-source feature lane edge recognition method based on a scoring mechanism according to claim 3, characterized in that: The method for determining the strength score of each point cloud's proximity to the lane feature includes the following steps: For the point cloud data of the road area, the reflection intensity of the point cloud is mapped to grayscale value and partitioned based on the threshold method; Calculate the inter-class variance between each partition and extract the maximum inter-class variance; Calculate the intensity score of each point in the point cloud based on the reflection intensity and the maximum inter-class variance of each point.

5. The lane edge recognition method based on a multi-source feature scoring mechanism according to claim 1, characterized in that: The lane edge line extraction method includes the following steps: The principal component analysis method is used to determine the direction of the lane line and extract the central axis of the lane line; Determine the normal direction of the central axis of the lane line and perform normal sampling to obtain candidate boundary points; For the obtained boundary candidate points, feature jump scores are performed based on intensity and color changes, and the points with the highest scores are selected as boundary points to obtain the extracted boundary points; All the boundary points on both sides of the lane lines are clustered and sorted to obtain the inner and outer edge point clouds of the lane lines, i.e., the lane edge lines.

6. The lane edge recognition method based on a multi-source feature scoring mechanism according to claim 1, characterized in that: Performing smooth fitting on the lane edge line to generate the fitted lane edge line includes the following steps: The centripetal parameterization method is used to parameterize the point cloud data points of the lane edge line to obtain the parameter value of each point cloud data point; Normalize the obtained parameter values; Construct a node vector, determine the starting node and the ending node, and use the point cloud data points between the starting node and the ending node as the intermediate points. With the starting node as the reference, calculate the position of each intermediate point through the parameterized value of the data point; Select control points, establish a relationship matrix between the control points, determine the initial positions of the control points by solving the linear equation of the B-spline curve, and generate a preliminary B-spline curve; For the obtained B-spline curve, the least squares method is used to further optimize the position of the control points so that the final generated curve is close to the real edge line, and the fitted lane edge line is obtained.

7. The lane edge recognition method based on a multi-source feature scoring mechanism according to claim 1, characterized in that: It also includes a vectorization processing step, using the KD-Tree spatial index to topologically sort the fitted lane edge lines, establishing accurate point cloud adjacency relationships based on neighboring point search, and obtaining the vectorized lane lines.

8. A multi-source feature lane edge recognition system based on a scoring mechanism, characterized by: include: A road point cloud extraction module is configured to acquire road point cloud data and perform preprocessing to obtain road area point cloud data; The lane extraction module is configured to fuse color information and laser intensity information to calculate a weighted score, process the road area point cloud, and segment the lane lines from the background to obtain a lane point cloud; The multi-lane edge line extraction module is configured to extract the edge line of each lane line based on the central main axis and normal feature jump score. The normal feature jump score is performed based on the intensity and color changes in the normal direction with the central main axis as the reference. The multi-lane edge line curve fitting module is configured to perform smooth fitting on the edge lines of the lane lines to obtain fitted lane edge lines.

9. The multi-source feature lane edge recognition system based on a scoring mechanism according to claim 8, characterized in that: It also includes a vectorization module that uses KD-Tree spatial indexing to topologically sort the fitted lane edge lines, establishes accurate point cloud adjacency relationships based on neighboring point search, and obtains vectorized lane lines.

10. A multi-source feature lane edge recognition system based on a scoring mechanism, characterized by: including a point cloud data acquisition device and a processor; A point cloud data acquisition device, used to obtain road point cloud data; The processor is configured to execute the steps in the multi-source feature lane edge line recognition method based on a scoring mechanism as described in any one of claims 1 to 7.

Images