WO2020107326A1

WO2020107326A1 - Lane line detection method, device and computer readale storage medium

Info

Publication number: WO2020107326A1
Application number: PCT/CN2018/118186
Authority: WO
Inventors: 崔健
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2018-11-29
Filing date: 2018-11-29
Publication date: 2020-06-04
Also published as: CN111433780A

Abstract

Disclosed are a lane line detection method, a device (500) and a computer readable storage medium. The method comprises the following steps: determining several candidate lane lines to be clustered in an image (S100); determining position parameters of the end points of respective candidate lane lines in the image (S200); clustering the candidate lane lines according to the relationship between the position parameters of the end points of respective candidate lane lines (S300); and determining the clustered candidate lane lines as lane lines detected from the image (S400).

Description

Lane line detection method, equipment, and computer-readable storage medium

Technical field

The present invention relates to the field of electronic technology, and in particular, to a lane line detection method, device, and computer-readable storage medium.

Background technique

In some scenarios such as automatic driving systems and ADAS (Advanced Driver Assistance Systems), the detection technology of lane lines has a very important significance. The accuracy of the detection results will directly affect the performance and reliability of the system.

In the related lane line detection method, the lane line in the image is detected by feature extraction, straight line or curve detection methods. However, in actual scenes, there will be dashed lane lines (including dashed lane lines caused by wear, and variable lane lane lines, etc.). A dashed lane line will contain more than two segmented lane lines. In the above detection method, a dotted lane line is detected as several different lane lines, and the detection result has low accuracy.

Summary of the invention

The invention provides a lane line detection method, device, and computer-readable storage medium, which can prevent a broken line lane line from being detected into several different lane lines, which is beneficial to improve the detection accuracy.

In a first aspect of an embodiment of the present invention, a lane line detection method is provided. The method includes:

Determine several candidate lane lines to be clustered in the image;

Determine the location parameters of the endpoints of each candidate lane line in the image;

Cluster the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line;

The candidate lane line after clustering is determined as the lane line detected from the image.

According to a second aspect of the embodiments of the present invention, an electronic device is provided, including: a memory and a processor;

The memory is used to store program codes;

The processor is used to call the program code, and when the program code is executed, it is used to perform the following operations:

Determine several candidate lane lines to be clustered in the image;

A third aspect of the embodiments of the present invention provides a computer-readable storage medium,

Computer instructions are stored on the computer-readable storage medium, and when the computer instructions are executed, the lane line detection method described in the foregoing embodiments is implemented.

Based on the above technical solutions, in the embodiments of the present invention,

In the embodiment of the present invention, the candidate lane lines are clustered based on the relationship between the endpoint position parameters of each candidate lane line, and the candidate lane lines that belong to the same dashed lane line can be aggregated into one category to prevent a dashed lane line Detecting as several different lane lines is helpful to improve the detection accuracy; and the position parameter on which the clustering is based is the endpoint of the candidate lane line, which will not cause the clustering of the two candidate lane lines when the middle part is closer Two candidate lane lines can prevent false clustering.

BRIEF DESCRIPTION

In order to more clearly explain the technical solutions in the embodiments of the present invention, the drawings required in the embodiments of the present invention will be briefly described below. Obviously, the drawings in the following description are only some implementations described in the present invention For example, for those of ordinary skill in the art, other drawings may also be obtained based on these drawings of the embodiments of the present invention.

1 is a schematic flowchart of a lane line detection method according to an embodiment of the invention;

2 is a schematic diagram of a candidate lane line determined in an image according to an embodiment of the present invention;

3 is a schematic flowchart of clustering candidate lane lines according to an embodiment of the present invention;

4 is a schematic diagram of a current candidate lane line and a first candidate lane line found according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of calculating a specified endpoint position parameter of a current candidate lane line and a target endpoint position parameter of a first candidate lane line found according to an embodiment of the present invention; FIG.

6 is a structural block diagram of an electronic device according to an embodiment of the invention.

detailed description

The technical solutions in the embodiments of the present invention will be described clearly and completely in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by a person of ordinary skill in the art without making creative efforts fall within the protection scope of the present invention. In addition, in the case of no conflict, the following embodiments and the features in the embodiments can be combined with each other.

The terminology used in the present invention is for the purpose of describing specific embodiments only, and does not limit the present invention. The singular forms "a", "said" and "the" used in the present invention and claims are also intended to include the majority forms unless the context clearly indicates other meanings. It should be understood that the term "and/or" as used herein refers to any or all possible combinations including one or more associated listed items.

Although the terms first, second, third, etc. may be used to describe various information in the present invention, the information should not be limited to these terms. These terms are used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as second information, and similarly, the second information may also be referred to as first information. Depending on the context, in addition, the word "if" can be interpreted as "when", or "when", or "in response to a determination".

The lane line detection method of the embodiment of the present invention will be described in detail below, but it should not be limited to this. In one embodiment, referring to FIG. 1, a lane line detection method may include the following steps:

S100: Determine several candidate lane lines to be clustered in the image;

S200: Determine position parameters of end points of each candidate lane line in the image;

S300: Cluster the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line;

S400: Determine the candidate lane line after clustering as the lane line detected from the image.

The execution subject of the lane line detection method of the embodiment of the present invention may be an electronic device, and more specifically may be a processor of the electronic device. The electronic device may be an imaging device that performs corresponding processing on the collected image; or, the electronic device may be a mobile device equipped with an imaging device, and the mobile device may acquire the image collected by the imaging device for corresponding processing. Ground robots, drones, vehicles, etc. Of course, the specific electronic device is not limited, and it only needs to have image processing capabilities.

Specifically, the lane line detection method of the embodiment of the present invention can be applied to a vehicle equipped with an automatic driving system and ADAS. During the driving process of the vehicle, the lane line in the collected image is detected to further control the driving, Planning etc.

In step S100, several candidate lane lines to be clustered in the image are determined.

The image may be a road image collected by an electronic device or obtained from an imaging device. Candidate lane lines can be initially detected from the image through related lane line detection methods, such as feature extraction, straight line, or curve detection methods.

However, in the related lane line detection method, a middle broken lane line is detected as several different lane lines, so among these candidate lane lines determined in step S100, there may be several candidate lane lines belonging to the same lane Lines, referring to FIG. 2, the candidate lane lines determined from the image include L1-L5, but L1-L3 among these candidate lane lines actually belong to the same lane line. Therefore, the candidate lane lines determined in step S100 need to be clustered to aggregate L1-L3 into the same category.

In step S200, the position parameters of the end points of each candidate lane line in the image are determined.

Since the candidate lane line is determined from the image, the position of the candidate lane line in the image can be determined, and accordingly, the position parameter of the endpoint of each candidate lane line in the image can be determined.

The position parameter is a parameter that can characterize the position of the end point of the candidate lane line in the image. The position parameter may include a vector parameter and/or a scalar parameter, which is not particularly limited. For example, the position parameter may include the tangent vector and normal vector of the endpoint on the candidate lane line; or, the position parameter may include the coordinates of the endpoint in the coordinate system to which the image is applied, when calculating the coordinates of the endpoint in the coordinate system to which the image is applied The distance of the imaging device that collects images from the ground can be used as a scale.

It can be understood that, in the embodiment of the present invention, the candidate lane line may refer to a linear region with a certain width in the image. When calculating the endpoint position parameters, the skeleton line can be extracted from the candidate lane line, and the endpoint of the skeleton line can be determined as the endpoint of the candidate lane line; or, the candidate can be determined from the end of the candidate lane line in a preset manner The end point of the lane line (for example, the midpoint of the end is determined as the end point), the specific is not limited.

In step S300, the candidate lane lines are clustered according to the relationship between the endpoint position parameters of each candidate lane line.

The relationship between the endpoint position parameters of each candidate lane line may be obtained by vector operation and/or scalar operation, and may represent the position relationship between the endpoints of the candidate lane line. Taking two candidate lane lines as an example, it can be determined whether the two candidate lane lines need to be aggregated into one class according to the relationship between the endpoint position parameters of the two candidate lane lines. After judging the relationship between all the candidate lane lines, you can complete the clustering of all the candidate lane lines.

Candidate lane lines whose relationship between end point position parameters meet certain conditions can be aggregated into one category. After clustering the candidate lane lines according to the relationship between the endpoint position parameters, several candidate lane lines that are broken off on the dotted lane line can be clustered.

For example, in Figure 2, L1 and L2 are aggregated into a category, L2 and L3 are aggregated into a category, after clustering is completed, L4 belongs to a category, L5 belongs to a category, L1-L3 belongs to a category, that is, L1-L3 is a cluster A candidate lane line after the class.

In step S400, the clustered candidate lane line is determined as the lane line detected from the image.

After the clustering is completed, the candidate lane lines belonging to different categories are determined as the detected different lane lines, and the candidate lane lines belonging to the same category are determined as the detected one lane line. Continuing to refer to FIG. 2, L4 is a lane line detected from the image, L5 is a lane line detected from the image, and L1-L3 is a lane line detected from the image.

In the related clustering method, in the first step, each object (pixel or area in the image) is used as a category, and the minimum distance between the categories is calculated; in the second step, the two categories whose minimum distance is less than the threshold are aggregated It is a category; the third step is to calculate the minimum distance between the clustered categories, and return to the second step to calculate until all categories cannot be aggregated. In this clustering method, when the shortest distance between the first lane segment and the second lane segment in the dashed lane line is greater than the shortest distance between the first lane segment and the adjacent lane line, the first lane segment may be It is grouped with the adjacent lane line into one category. The first lane segment and the second lane segment in the dashed lane line cannot be aggregated into one category, resulting in incorrect clustering.

In the embodiment of the present invention, the candidate lane lines are clustered based on the relationship between the endpoint position parameters of each candidate lane line, and the candidate lane lines that belong to the same dashed lane line can be aggregated into a category to prevent a dashed lane The line detection is a few different lane lines, which is helpful to improve the detection accuracy; and, the location parameter on which the clustering is based is the endpoint of the candidate lane line, which will not result in clustering when the middle part of the two candidate lane lines is closer The two candidate lane lines can prevent erroneous clustering; the calculation is also less, and the power consumption and cost are lower.

In one embodiment, in step S200, the determining position parameters of the end points of each candidate lane line in the image includes:

For each candidate lane line in the image, a curve fitting process is performed on the candidate lane line using a preset first curve model, and the position parameter of the end point of the candidate lane line after fitting is calculated;

The position parameter of the endpoint includes the tangent vector and the normal vector of the endpoint on the corresponding candidate lane line after fitting.

Each candidate lane line to be clustered determined in step S100 is subjected to curve fitting processing. For curve fitting, for example, the least squares curve fitting method can be used, and of course, other methods can also be used, such as the method of approximating discrete data with analytical expressions.

The first curve model used in the curve fitting process is, for example, a polynomial curve model. Correspondingly, the fitted curve is a polynomial curve. Of course, the first curve model is not specifically limited to this, but can also be other curve models, such as logarithm. Function model, piecewise function model, etc.

After the curve fitting process, the tangent vector and normal vector of the end point of the corresponding lane line can be calculated according to the first curve model and the parameters obtained by fitting. For details, please refer to the calculation method of the curve tangent vector and normal vector in the mathematical operation. I will not repeat them here.

In the embodiment of the present invention, the candidate lane line is curve-fitted, and the tangent vector and the normal vector of the end point on the fitted candidate lane line are used as the endpoint position parameters of the candidate lane line. Therefore, the relationship between the endpoint position parameters Taking into account the direction of the candidate lane line, the embodiment of the present invention is suitable for the detection of the lane line of the curve, and the problem of cross lanes at the curve can be avoided. Of course, it can also be applied to the detection of straight lane lines.

In one embodiment, referring to FIG. 3, in step S300, the clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line includes the following steps:

S301: Traverse each candidate lane line in the specified order, and determine whether there is a Target candidate lane lines of the current candidate lane line cluster merge;

S302: If yes, determine that the current candidate lane line and the target candidate lane line belong to the same category, and use the preset second curve model to compare the current candidate lane line and the target candidate lane in the image The line performs a curve fitting process to obtain the fitted candidate lane line, calculates the end position parameter of the fitted candidate lane line, and returns to the step of traversing each candidate lane line in the specified order.

In step S301, each candidate lane line is traversed in the specified order, and each time a candidate lane line is traversed, subsequent steps are performed. The specific order is not limited. For example, the specified order may be determined according to the position of the candidate lane line in the image, or the specified order may be determined according to the length of the candidate lane line.

Preferably, the designated order is the order of length of each candidate lane line from long to short. That is, when traversing, the candidate lane lines with a longer length are traversed first, and all candidate lane lines may be sorted according to length before traversing.

Referring to FIG. 2, for example, traversing to L4 first, because there is no target candidate lane line that needs to be merged with L4 in the candidate lane line that has not been traversed, the process is not executed, and then traversing to L5, because the candidate lane line has not There is no target candidate lane line that needs to be merged with L5, and then traversed to L1, L2 is a target candidate lane line that needs to be merged with L1, so step S302 is executed.

In step S301, each candidate lane line is traversed in the specified order, and when the last candidate lane line is traversed, the clustering ends. Referring to Figure 2, for example, when traversing to L3, it means that for any current candidate lane line traversed during this traversal, there is no candidate lane line that has not been traversed. The target candidate lane line of clustering merge indicates that the clustering has been completed and the clustering can be ended.

When traversing each candidate lane line in the specified order, you can first determine whether the number of candidate lane lines is greater than 1, and if so, based on the relationship between the endpoint position parameters of the traversed current candidate lane line and the untraversed candidate lane line Determine whether there is a target candidate lane line to be merged with the current candidate lane line cluster among the untraversed candidate lane lines; if not, the clustering ends.

In step S302, when there is a target candidate lane line to be merged with the current candidate lane line among the untraversed candidate lane lines, the current candidate lane line and the target candidate lane line are determined to belong to the same category, And perform a curve fitting process on the current candidate lane line and the target candidate lane line in the image using a preset second curve model to obtain a fitted candidate lane line, and calculate the fitted candidate lane line End position parameter.

Referring to FIG. 2, for example, L2 is a target candidate lane line that needs to be merged with L1, so L1 and L2 are determined to belong to the same category (the category can be a new category, a category of L1, or a category of L2, Specific is not limited, as long as it is different from other categories), and the second curve model is used to perform curve fitting processing on L1 and L2.

The curve fitting process here can also adopt least squares curve fitting and other methods, which will not be repeated here. The second curve model may be, for example, a polynomial curve model, a logarithmic function model, a piecewise function model, or the like.

In an embodiment, while determining that the current candidate lane line and the target candidate lane line belong to the same category, curve fitting processing is also performed on the current candidate lane line and the target candidate lane line, that is, edge aggregation is used The method of class edge fitting can adjust the position parameters of the candidate lane line and its endpoints during the aggregation process, which is helpful to improve the accuracy of clustering.

As an optional method, the curve fitting process may not be performed during the clustering process, and after the clustering is completed, the curve fitting process may be performed on the candidate lane lines belonging to the same category.

In step S302, returning to the step of traversing each candidate lane line in the specified order, that is, restarting traversing each candidate lane line in the specified order. In the new traversal, the specified order is still followed. Of course, the length order and number of candidate lane lines will be different due to clustering and curve fitting, that is, the length of the candidate lane line after fitting will become longer, The number of all candidate lane lines will be reduced.

In one embodiment, in step S301, it is determined whether there is a need in the untraversed candidate lane line according to the relationship between the endpoint position parameters of the traversed current candidate lane line and the untraversed candidate lane line The target candidate lane line merged with the current candidate lane line cluster includes the following steps:

S3011: Determine whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line;

S3012: If yes, for each first candidate lane line, calculate the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, and determine whether the relationship satisfies the set relationship, and if so, Then, the first candidate lane line is the target candidate lane line.

Each candidate lane line has two endpoints, but performing the corresponding processing according to one endpoint on the candidate lane line can also achieve the purpose of determining the target candidate lane line, and can also reduce the amount of calculation.

In step S3011, find the first candidate lane line that satisfies the specified condition from the candidate lane lines that have not been traversed according to the specified endpoint position parameter. Performing step S3012 for the first candidate lane line can reduce the amount of calculation required to perform step S3012 . Of course, instead of performing the judgment process, all the candidate lane lines that have not been traversed may be determined as the first candidate lane line.

In step S3012, the relationship between the target endpoint position parameter of each first candidate lane line and the specified endpoint position parameter is calculated. If the calculated relationship satisfies the set relationship, it is the target candidate lane line.

The designated end point of the current candidate lane line may be any one of the two end points of the current candidate lane line. When the designated end point is determined, the target endpoint position parameter of the first candidate lane line is also determined correspondingly.

Optionally, the designated endpoint is an endpoint with a smaller coordinate value in the designated direction of the current candidate lane line, and the target endpoint is an endpoint with a larger coordinate value in the designated direction of the target candidate lane line; or,

The designated endpoint is an endpoint with a larger coordinate value in a designated direction of the current candidate lane line, and the target endpoint is an endpoint with a smaller coordinate value in a designated direction of the target candidate lane line.

In one embodiment, the specified direction is a vertical direction or a horizontal direction applied in the coordinate system of the image. Of course, if the image has undergone transformation processing (such as inverse perspective transformation processing) before step S300, the designated direction is the vertical direction or the horizontal direction in the coordinate system applied to the transformed image.

2 and 4, for example, L1 is the current candidate lane line, L2 is the determined first candidate lane line, and P1 is the designated end point of L1 (the vertical axis of the current candidate lane line applied to the coordinate system of the image The end point with the larger coordinate value in the vertical direction), correspondingly, P2 is the target end point of L2 (the end point with the lower coordinate value in the vertical direction in the coordinate system of the target candidate lane line applied to the image), calculate The relationship between P1 and P2. When the relationship satisfies the set relationship, P2 is the target candidate lane line.

In one embodiment, the position parameter of the endpoint includes the tangent vector and normal vector of the endpoint on the corresponding candidate lane line;

In step S3012, the calculation of the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line includes the following steps:

S30121: Calculate the first tangential distance obtained by projecting the first vector onto the tangent vector of the specified endpoint, the first normal distance obtained by projecting the first vector onto the normal vector of the specified endpoint, and the second vector projected at A second tangential distance obtained on the tangent vector of the target endpoint, and a second normal distance obtained by projecting a second vector on the normal vector of the target endpoint; the first vector is the specified endpoint to the target endpoint , The second vector is the vector from the target endpoint to the specified endpoint;

S30122: Determine the larger of the first tangential distance and the second tangential distance as the target tangential distance, and determine the larger of the first normal distance and the second normal distance as the target Normal distance

S30123: Determine the target tangential distance and the target normal distance as the relationship between the specified endpoint position parameter and the target endpoint position parameter.

Referring to FIGS. 4 and 5, taking the calculation of the second tangential distance obtained by projecting the second vector onto the tangent vector of the target endpoint as an example

Is the second vector (that is, the vector from the target endpoint to the specified endpoint), n is the tangent vector of the target endpoint, and P2Q1 is

The second tangential distance projected on n, where P1Q1 is perpendicular to P2Q1. The other distances are similar and will not be repeated here.

In one embodiment, in step S3012, the judging whether the relationship satisfies the set relationship includes:

S30124: When the target tangential distance is less than the set tangential threshold, and the target normal distance is less than the set normal threshold, it is determined that the relationship satisfies the set relationship.

The target tangential distance is the larger of the first tangential distance and the second tangential distance. The target normal distance is the larger of the first normal distance and the second normal distance. In the above setting relationship, the smaller one must also satisfy the above setting relationship.

The specific values of the set tangential threshold and the set normal threshold are not limited, and can be determined according to the specific lane line conditions.

Under normal circumstances, only one target candidate lane line will be determined among all the first candidate lane lines, that is, there will be only one first candidate lane line whose relationship between the target endpoint position parameter and the specified endpoint position parameter satisfies the set relationship. Therefore, after calculating the relationship between the target endpoint position parameter of each first candidate lane line and the specified endpoint position parameter, you can first sort the target tangential distance or target normal distance of all the first candidate lane lines and select The relationship between the target tangential distance or the target normal distance is the smallest and compared with the set relationship. Of course, the relationship corresponding to each first candidate lane line may also be compared with the set relationship.

In one embodiment, in step S3011, determining whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line includes the following steps:

S30111: Determine the boundary for determining the search range according to the specified endpoint position parameter;

S30112: Determine a search range required to search for the first candidate lane line in the image according to the boundary;

S30113: Search for a candidate lane line within the search range in the image, and if found, determine the found candidate lane line as the first candidate lane line.

First determine the boundary according to the specified endpoint position parameters, and the search range determined according to the boundary is more suitable to avoid missing the first candidate lane line, and it will not contain too many first candidate lane lines. The candidate lane line is determined as the first candidate lane line. Compared with using the candidate lane lines in the entire image as the first candidate lane line, the number of first candidate lane lines can be reduced, and accordingly the subsequent required The amount of calculation does not affect the determination of the target candidate lane line.

In one embodiment, the boundary is the normal of the specified endpoint on the current candidate lane line;

In step S30112, determining the search range required to search for the first candidate lane line in the image according to the boundary includes the following steps:

Determining a first area and a second area on both sides of the boundary in the image, wherein the current candidate lane line is located in the first area;

The second area is determined as the search range.

Continuing to refer to FIG. 2, taking L1 as the current subsequent lane line as an example, and P1 as the designated end point of L1, the normal of L1 at P1 is determined as the boundary, and the image is divided into the second area where L1 is not located and the L1 is In the first area, determine the second area as the search range, and search for the candidate lane line in the search range. You can judge whether it is in the search range according to the endpoints at both ends of the candidate lane line. Then, L2 and L3 are found in the search range, and both L2 and L3 are determined as the first candidate lane line.

In one embodiment, after clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line in step S300, the method further includes the following steps:

S310: Perform a curve fitting process on each candidate lane line after clustering using a preset third curve model; wherein, the highest number of terms of the third curve model is greater than the highest number of terms of the second curve model;

In step S400, the determining the clustered candidate lane line as the lane line detected from the image includes:

Each candidate lane line after performing curve fitting processing using the third curve model is determined as the lane line detected from the image.

The curve fitting process is performed on the candidate lane lines belonging to the same category. The curve fitting process here can also adopt the method of least square curve fitting and the like, which will not be repeated here. The third curve model may be, for example, a polynomial curve model, a logarithmic function model, a piecewise function model, or the like.

Specifically, the second curve model may be: x=a*y^2+b*y+c; the third curve model may be: x=a*y^3+b*y^2+c*y +d.

The highest order number of the second curve model is lower than the highest order number of the third curve model, to prevent a very curved curve from being generated during the clustering iteration process, resulting in incorrect clustering results.

In one embodiment, in step S100, the determining a plurality of candidate lane lines to be clustered in the image may include the following steps:

S101: Detect, according to a preset lane line detection method, several candidate lane lines to be classified from the image;

S102: Determine a corresponding category for each candidate lane line to be classified;

S103: Determine each candidate lane line after the category is determined as the several candidate lane lines to be clustered.

In step S101, the preset lane line detection method may be, for example, edge detection, feature extraction and other lane line detection methods, and the candidate lane lines in the image are detected as candidate lane lines to be classified.

In step S102, a corresponding category is determined for each candidate lane line to be classified, for example, different candidate lane lines may be marked with different colors, and each color represents a category. In subsequent clustering, if two categories are If they are aggregated into the same category, the colors of the two categories will be unified, for example, one of the two colors will be modified to the other of the two colors.

It can be understood that the color marking is only a way to determine the category of the candidate lane line, which is not limited to this, and it may not need to be marked in the image, for example, the category mark (including characters, numbers, etc.) can be set for the candidate lane line ), store the candidate lane line and the category identifier in the memory, so that the candidate lane line can be determined to determine the category identifier corresponding to the candidate lane line, and the corresponding category can also be determined for each candidate lane line to be classified.

In S103, each candidate lane line after determining the category is determined as the several candidate lane lines to be clustered, and then the subsequent steps S200-S400 are performed.

In one embodiment, after determining the corresponding category for each candidate lane line to be classified in step S102, the method further includes the following steps:

S112: Perform skeleton extraction processing on each candidate lane line to refine each candidate lane line;

In step S103, the determination of each candidate lane line after the determination of the category as the several candidate lane lines to be clustered includes:

The refined candidate lane lines are determined as the candidate lane lines to be clustered.

Since the candidate lane line detected in step S101 usually has a certain width, and contains too many pixels, the number of pixels to be processed in the subsequent fitting and other steps is too large, and the calculation amount is huge. Therefore, in step S112, pass each candidate Carrying out skeleton extraction processing on lane lines can greatly reduce the amount of processing such as subsequent fitting.

The manner of performing skeleton extraction processing on each candidate lane line may be, for example, to retain only a single pixel point in the middle of the width direction of the candidate lane line to obtain the refined candidate lane line. Of course, the specific is not limited to this, for example, the edge in the length direction of the candidate lane line may also be extracted. After the skeleton extraction process is performed, certain image processing can also be performed on the candidate lane lines, such as smoothing processing and the like.

In one embodiment, after performing skeleton extraction processing on each candidate lane line in step S112, the method further includes the following steps:

S122: Perform an inverse perspective transformation process on each of the thinned candidate lane lines, so that the candidate lane lines are in a target perspective in the image, and the target perspective is to look down on the candidate lane lines when collecting the candidate lane lines Perspective

In step S103, the determination of each refined lane line as the candidate lane lines to be clustered includes:

Each candidate lane line after inverse perspective transformation processing is determined as the several candidate lane lines to be clustered.

After performing the inverse perspective transformation processing on each of the refined candidate lane lines, the candidate lane lines are all under the target angle of view. Under the target angle of view, the size ratio of the candidate lane lines conforms to the true size ratio, and no distortion of the candidate lane lines will occur The phenomenon can improve the robustness.

The method of inverse perspective transformation is not limited. For example, the coordinate conversion relationship between the candidate lane line from the current perspective to the target perspective can be established in advance and stored in the electronic device. During calculation, the electronic device directly calls the coordinate conversion relationship, that is It may be determined that inverse perspective transformation processing is performed on each of the thinned candidate lane lines.

The coordinate conversion relationship from the current angle of view to the target angle of view can be established according to the focal length of the imaging device, the optical axis position parameters, and the angle and height parameters of the imaging device relative to the ground, etc., without limitation.

In one embodiment, before determining the corresponding category for each candidate lane line to be classified in step S102, the method further includes:

S111: Perform dilation processing for each candidate lane line to be classified, so that the invalid pixel value in the candidate lane line is modified to a valid pixel value;

S121: Perform corrosion processing on the candidate lane line after the expansion process, so that the candidate lane line after the corrosion has the same size as the corresponding candidate lane line before expansion.

There may be some voids in the candidate lane line detected in step S101 (voids refer to pixels or pixel blocks in the candidate lane line that have different pixel values from neighboring pixel points. The neighborhood may be 4 neighborhoods or 8 neighbors. Field, etc.), the pixel value at the hole is an invalid pixel value and needs to be modified to a valid pixel value. However, the amount of processing required to detect a hole is very large. Therefore, in order to reduce the amount of processing, in this embodiment, the determination of whether there is a hole is not performed, and each candidate lane line to be classified is directly subjected to inflation processing.

The expansion processing method may be, for example: for each pixel in each candidate lane line, the pixel is used as the center to expand the pixel by the first radius in the four directions of up, down, left, and right (that is, the pixels of the pixels within the first radius) The values are all modified to the pixel values of the expanded pixels). Overall, the candidate lane line becomes larger in length and width by two first radii.

In order to restore the candidate lane line to the original length and width, in step S121, a corrosion process is performed on the candidate lane line after the expansion process. The corrosion processing method may be, for example: for each inflated candidate lane line, the edge of the candidate lane line is eroded inwardly to the pixels of the first radius (that is, the pixel values of the pixels within the first radius range are all modified to Pixel values of pixels adjacent to the edge of the candidate lane line outside the candidate lane line), the eroded candidate lane line has the same size as the corresponding candidate lane line before inflation.

After the expansion and corrosion treatment, the holes in the candidate lane line can be filled with the pixel values of the neighboring pixels.

In one embodiment, after clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line in step S300, the method may further include the following steps:

S320: Perform curve fitting processing on each candidate lane line after clustering by using a preset curve model;

In step S400, the determining the clustered candidate lane line as the lane line detected from the image includes the following steps:

Each candidate lane line after performing curve fitting processing using the preset curve model is determined as the lane line detected from the image.

In the clustering process of step S300, the curve fitting process may not be performed, but after the clustering ends, the curve fitting process is performed on the candidate lane lines belonging to the same category.

The curve fitting process in step S320 may also use least squares curve fitting and other methods, which will not be repeated here. The preset curve model may be, for example, a polynomial curve model, a logarithmic function model, a piecewise function model, or the like.

Specifically, the preset curve model may be, for example: x=a*y^3+b*y^2+c*y+d.

In one embodiment, in step S400, after the clustered candidate lane line is determined as the lane line detected from the image, the method further includes the following steps:

S501: Calculate the specified characteristic value of each detected lane line;

S502: Determine whether the specified characteristic value is within a set value range;

S503: If not, delete the lane line from all detected lane lines.

Preferably, the specified characteristic value includes at least one of the following parameters:

The curvature of the lane line;

The slope of the lane line;

The width of the lane line.

It can be understood that the specified feature value is not limited to the above three, but may be other parameters as long as it can characterize the shape characteristics of the lane line. The setting value range can also be determined according to a priori knowledge, which is not limited to specific, and the setting value range corresponding to different characteristic values can also be different.

According to the specified characteristic values such as the curvature, slope, and width of the lane line, deleting the detected lane line that does not conform to the actual situation is helpful to improve the robustness.

Based on the same concept as the above lane line detection method, referring to FIG. 6, an electronic device 500 includes a memory 501 and a processor 502 (such as one or more processors). The specific type of the electronic device is not limited, and the electronic device may be an imaging device but not limited to an imaging device. The electronic device may also be a device electrically connected to the imaging device, for example, and may acquire the image collected by the imaging device, and then execute the corresponding method.

In one embodiment, the memory is used to store program code;

Determine several candidate lane lines to be clustered in the image;

Preferably, the processor is specifically used when determining the position parameter of the endpoint of each candidate lane line in the image:

Preferably, the processor is specifically used for clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line:

Traverse each candidate lane line in the specified order, and determine whether there is a current candidate in the untraversed candidate lane line according to the relationship between the endpoint location parameters of the traversed current candidate lane line and the untraversed candidate lane line The target candidate lane line of cluster merge of lane lines;

If yes, determine that the current candidate lane line and the target candidate lane line belong to the same category, and perform the current candidate lane line and the target candidate lane line in the image using a preset second curve model The curve fitting process obtains the fitted candidate lane line, calculates the end position parameter of the fitted candidate lane line, and returns to the step of traversing each candidate lane line in the specified order.

Preferably, the processor determines whether there is a current candidate lane line among the untraversed candidate lane lines according to the relationship between the endpoint position parameters of the traversed current candidate lane lines and the untraversed candidate lane lines The target candidate lane line of cluster merge is specifically used for:

Determine whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line;

If so, for each first candidate lane line, calculate the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, and determine whether the relationship satisfies the set relationship. The first candidate lane line is the target candidate lane line.

Preferably, the position parameter of the endpoint includes a tangent vector and a normal vector of the endpoint on the corresponding candidate lane line;

When the processor calculates the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, it is specifically used to:

Calculate the first tangential distance obtained by projecting the first vector onto the tangent vector at the specified endpoint, the first normal distance obtained by projecting the first vector onto the normal vector at the specified endpoint, and the second vector projected at the A second tangential distance obtained on the tangent vector of the target endpoint, and a second normal distance obtained by projecting the second vector on the normal vector of the target endpoint; the first vector is the vector from the specified endpoint to the target endpoint , The second vector is a vector from the target endpoint to a specified endpoint;

The larger of the first and second tangential distances is determined as the target tangential distance, and the larger of the first and second normal distances is determined as the target normal direction distance;

The target tangential distance and the target normal distance are determined as the relationship between the specified endpoint position parameter and the target endpoint position parameter.

Preferably, the processor is specifically used to determine whether the relationship satisfies the set relationship:

When the target tangential distance is less than the set tangential threshold and the target normal distance is less than the set normal threshold, it is determined that the relationship satisfies the set relationship.

Preferably, the processor is specifically used to determine whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line:

Determine the boundary for determining the search range according to the specified endpoint position parameter;

Determining a search range required to search for the first candidate lane line in the image according to the boundary;

Searching for a candidate lane line within the search range in the image, and if found, determining the found candidate lane line as the first candidate lane line.

Preferably, the boundary is a normal line of the specified endpoint on the current candidate lane line;

The determining the search range required to search for the first candidate lane line in the image according to the boundary is specifically used to:

The second area is determined as the search range.

Preferably,

The designated endpoint is an endpoint with a smaller coordinate value in the designated direction of the current candidate lane line, and the target endpoint is an endpoint with a larger coordinate value in the designated direction of the target candidate lane line; or,

Preferably,

The specified direction is a vertical direction or a horizontal direction applied in the coordinate system of the image.

Preferably, after the processor clusters the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line, it is further used to:

Using a preset third curve model to perform curve fitting processing on each candidate lane line after clustering; wherein, the highest number of terms of the third curve model is greater than the highest number of terms of the second curve model;

When the processor determines the clustered candidate lane line as the lane line detected from the image, it is specifically used for:

Preferably,

The designated order is an order in which the length of each candidate lane line is long to short.

Preferably, when the processor determines several candidate lane lines to be clustered in the image, it is specifically used to:

Detecting, according to a preset lane line detection method, several candidate lane lines to be classified from the image;

Determine the corresponding category for each candidate lane line to be classified;

Each candidate lane line after determining the category is determined as the several candidate lane lines to be clustered.

Preferably, after the processor determines the corresponding category for each candidate lane line to be classified, it is further used to:

Perform skeleton extraction processing on each candidate lane line to refine each candidate lane line;

When the processor determines each candidate lane line after determining the category as the several candidate lane lines to be clustered, it is specifically used for:

Preferably, after the processor performs skeleton extraction processing on each candidate lane line, it is further used to:

Perform an inverse perspective transformation process on each of the refined lane lines so that the candidate lane lines are in the target angle of view in the image, and the target angle of view is the perspective of looking down the candidate lane lines when collecting the candidate lane lines ;

When the processor determines the refined candidate lane lines as the candidate lane lines to be clustered, it is specifically used for:

Preferably, before the processor determines the corresponding category for each candidate lane line to be classified, it is further used to:

Performing dilation processing for each candidate lane line to be classified, so that the invalid pixel value in the candidate lane line is modified to a valid pixel value;

Corrosion processing is performed on the candidate lane line after the expansion process so that the candidate lane line after the corrosion has the same size as the corresponding candidate lane line before expansion.

Preferably, after the processor clusters the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line, it is also used to:

Perform curve fitting processing on each candidate lane line after clustering by using a preset curve model;

Preferably, after the processor determines the clustered candidate lane line as the lane line detected from the image, it is further used to:

Calculating the specified characteristic value of each detected lane line;

Determine whether the specified characteristic value is within the set value range;

If not, the lane line is deleted from all detected lane lines.

The curvature of the lane line;

The slope of the lane line;

The width of the lane line.

Based on the same inventive concept as the above method, the present invention also provides a computer-readable storage medium that stores computer instructions, and when the computer instructions are executed, the lane described in the foregoing embodiment is implemented Line detection method.

The system, device, module or unit explained in the above embodiments may be realized by a computer chip or entity, or by a product with a certain function. A typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email sending and receiving device, and a game control Desk, tablet computer, wearable device, or any combination of these devices.

For the convenience of description, when describing the above device, the functions are divided into various units and described separately. Of course, when implementing the present invention, the functions of each unit may be implemented in one or more software and/or hardware.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Moreover, embodiments of the present invention may take the form of computer program products implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

The present invention is described with reference to flowcharts and/or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and/or block in the flowchart and/or block diagram and a combination of the flow and/or block in the flowchart and/or block diagram may be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processing machine, or other programmable data processing device to produce a machine that enables the generation of instructions executed by the processor of the computer or other programmable data processing device An apparatus for realizing the functions specified in one block or multiple blocks of one flow or multiple flows of a flowchart and/or one block or multiple blocks of a block diagram.

Moreover, these computer program instructions may also be stored in a computer readable memory that can guide the computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including instruction means, The instruction device implements the functions specified in one block or multiple blocks of one flow or multiple blocks of the flowchart and/or block diagram.

These computer program instructions can also be loaded into a computer or other programmable data processing device so that a series of operating steps are performed on the computer or other programmable device to generate computer-implemented processing, thereby executing instructions on the computer or other programmable device Steps are provided for implementing the functions specified in one flow or multiple flows of the flowchart and/or one block or multiple blocks of the block diagram.

The above are only the embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

A lane line detection method, characterized in that the method includes:

Determine several candidate lane lines to be clustered in the image;

Determine the location parameters of the endpoints of each candidate lane line in the image;

Cluster the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line;

The candidate lane line after clustering is determined as the lane line detected from the image.
The lane line detection method according to claim 1, wherein the determining position parameters of the endpoints of each candidate lane line in the image includes:

For each candidate lane line in the image, a curve fitting process is performed on the candidate lane line using a preset first curve model, and the position parameter of the end point of the candidate lane line after fitting is calculated;

The position parameter of the endpoint includes the tangent vector and the normal vector of the endpoint on the corresponding candidate lane line after fitting.
The lane line detection method according to claim 1, wherein the clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line includes:

Traverse each candidate lane line in the specified order, and determine whether there is a current candidate in the untraversed candidate lane line according to the relationship between the endpoint location parameters of the traversed current candidate lane line and the untraversed candidate lane line The target candidate lane line of cluster merge of lane lines;

If yes, determine that the current candidate lane line and the target candidate lane line belong to the same category, and perform the current candidate lane line and the target candidate lane line in the image using a preset second curve model The curve fitting process obtains the fitted candidate lane line, calculates the end position parameter of the fitted candidate lane line, and returns to the step of traversing each candidate lane line in the specified order.
The lane line detection method according to claim 3, wherein the untraversed candidate is determined according to the relationship between the endpoint position parameters of the traversed current candidate lane line and the untraversed candidate lane line Whether there is a target candidate lane line in the lane line to be merged with the current candidate lane line cluster, including:

Determine whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line;

If so, for each first candidate lane line, calculate the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, and determine whether the relationship satisfies the set relationship. The first candidate lane line is the target candidate lane line.
The lane line detection method according to claim 4, wherein the position parameter of the endpoint includes a tangent vector and a normal vector of the endpoint on the corresponding candidate lane line;

The calculating the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line includes:

Calculate the first tangential distance obtained by projecting the first vector onto the tangent vector at the specified endpoint, the first normal distance obtained by projecting the first vector onto the normal vector at the specified endpoint, and the second vector projected at the A second tangential distance obtained on the tangent vector of the target endpoint, and a second normal distance obtained by projecting the second vector on the normal vector of the target endpoint; the first vector is the vector from the specified endpoint to the target endpoint , The second vector is a vector from the target endpoint to a specified endpoint;

The larger of the first and second tangential distances is determined as the target tangential distance, and the larger of the first and second normal distances is determined as the target normal direction distance;

The target tangential distance and the target normal distance are determined as the relationship between the specified endpoint position parameter and the target endpoint position parameter.
The lane line detection method according to claim 5, wherein the judging whether the relationship satisfies the set relationship includes:

When the target tangential distance is less than the set tangential threshold and the target normal distance is less than the set normal threshold, it is determined that the relationship satisfies the set relationship.
The lane line detection method according to claim 4, characterized in that it is determined whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line ,include:

Determine the boundary for determining the search range according to the specified endpoint position parameter;

Determining a search range required to search for the first candidate lane line in the image according to the boundary;

Searching for a candidate lane line within the search range in the image, and if found, determining the found candidate lane line as the first candidate lane line.
The lane line detection method according to claim 7, wherein the boundary is a normal line of the specified endpoint on the current candidate lane line;

The search range required to search for the first candidate lane line in the image according to the boundary includes:

Determining a first area and a second area on both sides of the boundary in the image, wherein the current candidate lane line is located in the first area;

The second area is determined as the search range.
The lane line detection method according to claim 4, wherein:

The designated endpoint is an endpoint with a smaller coordinate value in the designated direction of the current candidate lane line, and the target endpoint is an endpoint with a larger coordinate value in the designated direction of the target candidate lane line; or,

The designated endpoint is an endpoint with a larger coordinate value in a designated direction of the current candidate lane line, and the target endpoint is an endpoint with a smaller coordinate value in a designated direction of the target candidate lane line.
The lane line detection method according to claim 9, wherein:

The specified direction is a vertical direction or a horizontal direction applied in the coordinate system of the image.
The lane line detection method according to claim 3, wherein after clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line, the method further comprises:

Using a preset third curve model to perform curve fitting processing on each candidate lane line after clustering; wherein, the highest number of terms of the third curve model is greater than the highest number of terms of the second curve model;

The determination of the candidate lane line after clustering as the lane line detected from the image includes:

Each candidate lane line after performing curve fitting processing using the third curve model is determined as the lane line detected from the image.
The lane line detection method according to claim 3, wherein:

The designated order is an order in which the length of each candidate lane line is long to short.
The lane line detection method according to any one of claims 1-12, wherein the determining a plurality of candidate lane lines to be clustered in the image includes:

Detecting, according to a preset lane line detection method, several candidate lane lines to be classified from the image;

Determine the corresponding category for each candidate lane line to be classified;

Each candidate lane line after determining the category is determined as the several candidate lane lines to be clustered.
The lane line detection method according to claim 13, wherein after the corresponding category is determined for each candidate lane line to be classified, the method further comprises:

Perform skeleton extraction processing on each candidate lane line to refine each candidate lane line;

The determination of each candidate lane line after determining the category as the several candidate lane lines to be clustered includes:

The refined candidate lane lines are determined as the candidate lane lines to be clustered.
The lane line detection method according to claim 14, wherein after performing skeleton extraction processing on each candidate lane line, the method further comprises:

Perform an inverse perspective transformation process on each of the refined lane lines so that the candidate lane lines are in the target angle of view in the image, and the target angle of view is the perspective of looking down the candidate lane lines when collecting the candidate lane lines ;

The determination of each candidate lane line after refinement as the several candidate lane lines to be clustered includes:

Each candidate lane line after inverse perspective transformation processing is determined as the several candidate lane lines to be clustered.
The lane line detection method according to claim 13, wherein before the corresponding category is determined for each candidate lane line to be classified, the method further comprises:

Performing dilation processing for each candidate lane line to be classified, so that the invalid pixel value in the candidate lane line is modified to a valid pixel value;

Corrosion processing is performed on the candidate lane line after the expansion process so that the candidate lane line after the corrosion has the same size as the corresponding candidate lane line before expansion.
The lane line detection method according to claim 1, wherein after clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line, further comprising:

Perform curve fitting processing on each candidate lane line after clustering by using a preset curve model;

The determination of the candidate lane line after clustering as the lane line detected from the image includes:

Each candidate lane line after performing curve fitting processing using the preset curve model is determined as the lane line detected from the image.
The lane line detection method according to claim 1, wherein after the clustered candidate lane line is determined as the lane line detected from the image, the method further comprises:

Calculating the specified characteristic value of each detected lane line;

Determine whether the specified characteristic value is within the set value range;

If not, the lane line is deleted from all detected lane lines.
The lane line detection method according to claim 18, wherein the specified characteristic value includes at least one of the following parameters:

The curvature of the lane line;

The slope of the lane line;

The width of the lane line.
An electronic device, characterized in that it includes: a memory and a processor;

The memory is used to store program codes;

The processor is used to call the program code, and when the program code is executed, it is used to perform the following operations:

Determine several candidate lane lines to be clustered in the image;

Determine the location parameters of the endpoints of each candidate lane line in the image;

Cluster the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line;

The candidate lane line after clustering is determined as the lane line detected from the image.
The device according to claim 20, wherein the processor is specifically used when determining the position parameter of the endpoint of each candidate lane line in the image:

For each candidate lane line in the image, a curve fitting process is performed on the candidate lane line using a preset first curve model, and the position parameter of the end point of the candidate lane line after fitting is calculated;

The position parameter of the endpoint includes the tangent vector and the normal vector of the endpoint on the corresponding candidate lane line after fitting.
The device according to claim 20, wherein the processor is specifically used for clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line:

Traverse each candidate lane line in the specified order, and determine whether there is a current candidate in the untraversed candidate lane line according to the relationship between the endpoint location parameters of the traversed current candidate lane line and the untraversed candidate lane line The target candidate lane line of cluster merge of lane lines;

If yes, determine that the current candidate lane line and the target candidate lane line belong to the same category, and perform the current candidate lane line and the target candidate lane line in the image using a preset second curve model The curve fitting process obtains the fitted candidate lane line, calculates the end position parameter of the fitted candidate lane line, and returns to the step of traversing each candidate lane line in the specified order.
The device according to claim 22, wherein the processor determines the untraversed candidate lane according to the relationship between the endpoint position parameters of the traversed current candidate lane line and the untraversed candidate lane line Whether there is a target candidate lane line to be merged with the current candidate lane line cluster is specifically used for:

Determine whether there is at least one first candidate lane line that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line;

If so, for each first candidate lane line, calculate the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, and determine whether the relationship satisfies the set relationship. The first candidate lane line is the target candidate lane line.
The device according to claim 23, wherein the position parameter of the endpoint includes a tangent vector and a normal vector of the endpoint on the corresponding candidate lane line;

When the processor calculates the relationship between the specified endpoint position parameter and the target endpoint position parameter of the first candidate lane line, it is specifically used to:

Calculate the first tangential distance obtained by projecting the first vector onto the tangent vector at the specified endpoint, the first normal distance obtained by projecting the first vector onto the normal vector at the specified endpoint, and the second vector projected at the A second tangential distance obtained on the tangent vector of the target endpoint, and a second normal distance obtained by projecting the second vector on the normal vector of the target endpoint; the first vector is the vector from the specified endpoint to the target endpoint , The second vector is a vector from the target endpoint to a specified endpoint;

The larger of the first and second tangential distances is determined as the target tangential distance, and the larger of the first and second normal distances is determined as the target normal direction distance;

The target tangential distance and the target normal distance are determined as the relationship between the specified endpoint position parameter and the target endpoint position parameter.
The device according to claim 24, wherein the processor is specifically used to determine whether the relationship satisfies the set relationship:

When the target tangential distance is less than the set tangential threshold and the target normal distance is less than the set normal threshold, it is determined that the relationship satisfies the set relationship.
The device according to claim 23, wherein the processor determines whether there is at least one first candidate lane that satisfies the specified condition among the candidate lane lines that have not been traversed according to the specified endpoint position parameter of the current candidate lane line The line time is specifically used for:

Determine the boundary for determining the search range according to the specified endpoint position parameter;

Determining a search range required to search for the first candidate lane line in the image according to the boundary;

Searching for a candidate lane line within the search range in the image, and if found, determining the found candidate lane line as the first candidate lane line.
The apparatus according to claim 26, wherein the boundary is a normal line of the specified endpoint on the current candidate lane line;

The determining the search range required to search for the first candidate lane line in the image according to the boundary is specifically used to:

Determining a first area and a second area on both sides of the boundary in the image, wherein the current candidate lane line is located in the first area;

The second area is determined as the search range.
The device of claim 23, wherein

The designated endpoint is an endpoint with a smaller coordinate value in the designated direction of the current candidate lane line, and the target endpoint is an endpoint with a larger coordinate value in the designated direction of the target candidate lane line; or,

The designated endpoint is an endpoint with a larger coordinate value in a designated direction of the current candidate lane line, and the target endpoint is an endpoint with a smaller coordinate value in a designated direction of the target candidate lane line.
The device according to claim 28, characterized in that

The specified direction is a vertical direction or a horizontal direction applied in the coordinate system of the image.
The device according to claim 22, wherein the processor further clusters the candidate lane lines after clustering the candidate lane lines according to the relationship between the endpoint position parameters of each candidate lane line:

Using a preset third curve model to perform curve fitting processing on each candidate lane line after clustering; wherein, the highest number of terms of the third curve model is greater than the highest number of terms of the second curve model;

When the processor determines the clustered candidate lane line as the lane line detected from the image, it is specifically used for:

Each candidate lane line after performing curve fitting processing using the third curve model is determined as the lane line detected from the image.
The device of claim 22, wherein

The designated order is an order in which the length of each candidate lane line is long to short.
The device according to any one of claims 20 to 31, wherein the processor is specifically used when determining a plurality of candidate lane lines to be clustered in the image:

Detecting, according to a preset lane line detection method, several candidate lane lines to be classified from the image;

Determine the corresponding category for each candidate lane line to be classified;

Each candidate lane line after determining the category is determined as the several candidate lane lines to be clustered.
The device of claim 32, wherein after the processor determines a corresponding category for each candidate lane line to be classified, it is further used to:

Perform skeleton extraction processing on each candidate lane line to refine each candidate lane line;

When the processor determines each candidate lane line after determining the category as the several candidate lane lines to be clustered, it is specifically used for:

The refined candidate lane lines are determined as the candidate lane lines to be clustered.
The device according to claim 33, wherein after the processor performs skeleton extraction processing on each candidate lane line, it is further used to:

Perform an inverse perspective transformation process on each of the refined lane lines so that the candidate lane lines are in the target angle of view in the image, and the target angle of view is the perspective of looking down the candidate lane lines when collecting the candidate lane lines ;

When the processor determines the refined candidate lane lines as the candidate lane lines to be clustered, it is specifically used for:

Each candidate lane line after inverse perspective transformation processing is determined as the several candidate lane lines to be clustered.
The device according to claim 32, wherein before the processor determines a corresponding category for each candidate lane line to be classified, it is further used to:

Performing dilation processing for each candidate lane line to be classified, so that the invalid pixel value in the candidate lane line is modified to a valid pixel value;

Corrosion processing is performed on the candidate lane line after the expansion process so that the candidate lane line after the corrosion has the same size as the corresponding candidate lane line before expansion.
The apparatus according to claim 20, wherein the processor is further used to: after clustering the candidate lane lines according to the relationship between the endpoint position parameters of the respective lane lane lines

Perform curve fitting processing on each candidate lane line after clustering by using a preset curve model;

When the processor determines the clustered candidate lane line as the lane line detected from the image, it is specifically used to:

Each candidate lane line after performing curve fitting processing using the preset curve model is determined as the lane line detected from the image.
The device of claim 20, wherein the processor determines the clustered candidate lane line as the lane line detected from the image, and is further used to:

Calculating the specified characteristic value of each detected lane line;

Determine whether the specified characteristic value is within the set value range;

If not, the lane line is deleted from all detected lane lines.
The device according to claim 37, wherein the specified characteristic value includes at least one of the following parameters:

The curvature of the lane line;

The slope of the lane line;

The width of the lane line.
A computer-readable storage medium, characterized in that

Computer instructions are stored on the computer-readable storage medium, and when the computer instructions are executed, the lane line detection method according to any one of claims 1-19 is implemented.