CN-122023457-A - Lane line tracking method, device and equipment

CN122023457ACN 122023457 ACN122023457 ACN 122023457ACN-122023457-A

Abstract

The application provides a lane line tracking method, a lane line tracking device and lane line tracking equipment. The method comprises the steps of detecting positions of lane lines in a target image to obtain lane line detection results, determining coordinates of corresponding point chains according to the lane line detection results, wherein the point chains are discrete point sets which correspond to the lane lines and are arranged according to a spatial sequence and are used for representing continuous geometric shapes of the lane lines, constructing a plurality of lane line models fitting the point chains based on the coordinates of the point chains and a preset highest polynomial order, respectively corresponding to different polynomial orders and not exceeding the preset highest polynomial order, respectively calculating fitting errors corresponding to the lane line models, taking the lane line model with the smallest fitting error as a final lane line model, and carrying out lane line tracking based on the final lane line model to obtain lane line tracks. By verifying the applicability of different models, the problem of high-order term overfitting caused by noise is avoided, the accuracy of track prediction is ensured, and the lane line fitting stability is improved.

Inventors

Zhao Jiangshuo
YANG JUNFENG
LU YUE

Assignees

北京经纬恒润科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260116

Claims (10)

1. A lane line tracking method, the method comprising: detecting the position of a lane line in a target image to obtain a lane line detection result, wherein the target image comprises a lane line image; determining coordinates of a corresponding point chain according to the lane line detection result, wherein the point chain is a discrete point set which corresponds to the lane line and is arranged according to a space sequence and is used for representing the continuous geometric shape of the lane line; Constructing a plurality of lane line models fitting the point chain based on the coordinates of the point chain and a preset highest polynomial order, wherein the lane line models respectively correspond to different polynomial orders and do not exceed the preset highest polynomial order; Calculating fitting errors corresponding to the lane line models respectively, and taking the lane line model with the minimum fitting error as a final lane line model; and carrying out lane line tracking based on the final lane line model to obtain a lane line track.
2. The lane line tracking method according to claim 1, wherein the calculating the fitting error corresponding to each lane line model, and taking the lane line model with the smallest fitting error as the final lane line model, comprises: based on a preset fitting algorithm, fitting errors corresponding to the lane line models are calculated respectively; And comparing fitting errors of all the lane line models, and taking the lane line model with the minimum fitting error as a final lane line model.
3. The lane line tracking method according to claim 1, wherein the performing lane line tracking based on the final lane line model to obtain a lane line track comprises: Based on a preset prediction algorithm, predicting prediction model parameters corresponding to historic lane line tracks of a plurality of lane lines in a target image respectively; matching tracking model parameters of each lane line in the final lane line model with a plurality of prediction model parameters; and according to the matching result, updating the corresponding historical lane line track based on the tracking model parameters of each lane line in the final lane line model, and determining the current lane line track.
4. The lane-line tracking method according to claim 3, wherein the matching of the tracking model parameters of each lane line in the final lane-line model with a plurality of the prediction model parameters includes: based on the primary term coefficient and the secondary term coefficient in the model parameters, performing primary matching on a plurality of tracking model parameters and a plurality of prediction model parameters; and carrying out secondary matching on the tracking model parameters and the prediction model parameters which are failed to be matched in the primary matching based on the coordinates of the point chain corresponding to the model parameters.
5. The lane-line tracking method according to claim 4, wherein the performing the secondary matching of the tracking model parameter and the prediction model parameter, which have failed the matching in the primary matching, based on the coordinates of the point chain to which the model parameter corresponds, comprises: Based on each tracking model parameter of the matching failure in the primary matching, respectively determining the coordinates of the point chain of each first lane line; Based on each prediction model parameter of the matching failure in the primary matching, respectively determining coordinates of a point chain of each second lane line; Calculating the matching distance between the first lane line and the second lane line based on the coordinates of the point chain of each first lane line and the coordinates of the point chain of each second lane line; and matching the plurality of first lane lines with the plurality of second lane lines based on the Hungary algorithm and the matching distance.
6. The lane-line tracking method according to claim 5, wherein after matching the plurality of first lane lines with the plurality of second lane lines based on the hungarian algorithm and the matching distance, further comprising: if at least one first lane line is not matched successfully, establishing a new lane line track based on the first lane line which is not matched successfully; And deleting the corresponding historical lane line track based on the second lane line which is not matched successfully when at least one second lane line is not matched successfully.
7. The lane-tracking method according to any one of claims 1 to 6, wherein the determining coordinates of the corresponding point chain based on the lane-detection result includes: Resolving the lane line detection result to obtain pixel coordinates to be filtered corresponding to the lane line in the image; And performing outlier filtering on the pixel coordinates to be filtered based on a preset filtering algorithm to obtain coordinates of a point chain corresponding to the lane line detection result.
8. The lane line tracking method according to claim 7, wherein the performing outlier filtering on the pixel coordinates to be filtered based on a preset filtering algorithm to obtain coordinates of a point chain corresponding to the lane line detection result includes: acquiring preset filtering parameters, wherein the preset filtering parameters comprise target times and an error threshold value; Randomly extracting part of pixel coordinates to be filtered from a plurality of pixel coordinates to be filtered as samples to fit, so as to obtain a filtering model; Calculating error values of a plurality of pixel coordinates to be filtered and the filtering model respectively; taking the pixel coordinates to be filtered corresponding to the error value lower than the error threshold value as interior point coordinates; Under the condition that the number of the interior point coordinates and the R square value of the filtering model meet preset conditions, taking the filtering model as an adaptation model, and returning to the step of randomly extracting part of pixel coordinates to be filtered from a plurality of pixel coordinates to be filtered as samples to fit, so as to obtain the filtering model until the iteration times reach the target times; And determining the coordinates of the inner points of the adaptation model as the coordinates of the point chain corresponding to the lane line detection result.
9. A lane line tracking apparatus, the apparatus comprising: The detection module is used for detecting the position of the lane line in the target image to obtain a lane line detection result, wherein the target image comprises a lane line image; The determining module is used for determining the coordinates of a corresponding point chain according to the lane line detection result, wherein the point chain is a discrete point set which corresponds to the lane line and is arranged according to the space sequence and is used for representing the continuous geometric shape of the lane line; The construction module is used for constructing a plurality of lane line models fitting the point chain based on the coordinates of the point chain and a preset highest polynomial order, wherein the lane line models respectively correspond to different polynomial orders and do not exceed the preset highest polynomial order; the calculation module is used for calculating fitting errors corresponding to the lane line models respectively, and taking the lane line model with the minimum fitting error as a final lane line model; And the tracking module is used for tracking the lane line based on the final lane line model to obtain a lane line track.
10. A lane tracking apparatus comprising a processor and a memory storing computer program instructions, the processor reading and executing the computer program instructions to implement the lane tracking method of any one of claims 1-8.

Description

Lane line tracking method, device and equipment Technical Field The present application relates to the field of data processing technologies, and in particular, to a lane tracking method, apparatus, and device. Background Lane keeping assistance (Lane-KEEPING ASSISTANCE, LKA) and Lane departure warning (Lane Departure Warning, LDW) are vital safety functions in modern autonomous and advanced driving assistance systems. The functions depend on real-time, accurate and stable perception of the current driving lane line of the vehicle, and the position, the direction, the curvature and other information of the lane line relative to the vehicle are required to be continuously acquired, so that decision basis is provided for transverse control of the vehicle. However, the actual road environment is complex and changeable, and great challenges are brought to subsequent lane line model fitting and stable tracking. The existing scheme generally adopts a polynomial curve to fit a lane line model, but an unsuitable polynomial can introduce distortion and instability, so that curve overfitting is easy to cause, track divergence is easy to cause when detecting discontinuous, serious shielding or complex road conditions, the continuity of lane line perception is difficult to maintain, and the stability of lane line fitting is not high. Disclosure of Invention The embodiment of the application provides a lane line tracking method, a lane line tracking device and lane line tracking equipment, which can solve the technical problem of how to improve the lane line fitting stability. In a first aspect, the present application provides a lane line tracking method, including: detecting the position of a lane line in a target image to obtain a lane line detection result, wherein the target image comprises a lane line image; determining coordinates of a corresponding point chain according to the lane line detection result, wherein the point chain is a discrete point set which corresponds to the lane line and is arranged according to a space sequence and is used for representing the continuous geometric shape of the lane line; Constructing a plurality of lane line models fitting the point chain based on the coordinates of the point chain and a preset highest polynomial order, wherein the lane line models respectively correspond to different polynomial orders and do not exceed the preset highest polynomial order; Calculating fitting errors corresponding to the lane line models respectively, and taking the lane line model with the minimum fitting error as a final lane line model; and carrying out lane line tracking based on the final lane line model to obtain a lane line track. In some possible implementations, the calculating the fitting error corresponding to each lane line model, taking the lane line model with the smallest fitting error as the final lane line model includes: based on a preset fitting algorithm, fitting errors corresponding to the lane line models are calculated respectively; And comparing fitting errors of all the lane line models, and taking the lane line model with the minimum fitting error as a final lane line model. In some possible implementations, the performing lane line tracking based on the final lane line model to obtain a lane line track includes: Based on a preset prediction algorithm, predicting prediction model parameters corresponding to historic lane line tracks of a plurality of lane lines in a target image respectively; matching tracking model parameters of each lane line in the final lane line model with a plurality of prediction model parameters; and according to the matching result, updating the corresponding historical lane line track based on the tracking model parameters of each lane line in the final lane line model, and determining the current lane line track. In some possible implementations, the matching the tracking model parameters of each lane line in the final lane line model with the plurality of prediction model parameters includes: based on the primary term coefficient and the secondary term coefficient in the model parameters, performing primary matching on a plurality of tracking model parameters and a plurality of prediction model parameters; and carrying out secondary matching on the tracking model parameters and the prediction model parameters which are failed to be matched in the primary matching based on the coordinates of the point chain corresponding to the model parameters. In some possible implementations, the performing secondary matching on the tracking model parameter and the prediction model parameter that are failed to match in the primary matching based on the coordinates of the point chain corresponding to the model parameter includes: Based on each tracking model parameter of the matching failure in the primary matching, respectively determining the coordinates of the point chain of each first lane line; Based on each prediction model parameter of the matching