Lane Line Prediction Model for Damaged Road Markings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane detection systems struggle with recognizing lane lines that are damaged, shaded, or located far away, leading to inaccurate fitting and reduced efficiency in systems like lane departure warning, lane keeping support, and autonomous driving.
Innovation Solution
A method and system that transform lane line images into characteristic values, generate prediction lane lines using these values, and form a lane line prediction model to dynamically complete lane lines, improving recognition precision and efficiency, especially for far lanes and damaged or shaded lines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a processor fits characteristic points of all lanes in the image using conventional algorithms, then the system can recognize lane lines, but the recognition accuracy deteriorates when lane lines are damaged, shaded, or located far away
Solution Approach 1:
The patent applies preliminary action by first identifying characteristic points only in the lower region of the image (closer to the vehicle) where lane lines are clearer, then using these points to predict the positions of lane lines in the upper region (farther away). This avoids the problem of trying to directly recognize characteristic points in difficult conditions, instead preparing accurate reference data from easier conditions first.
Solution Approach 2:
The patent introduces prediction lane lines as an intermediary element. These prediction lane lines are generated based on characteristic points from the lower region and serve as a bridge to determine characteristic points in the upper region. This intermediary approach allows the system to overcome direct recognition difficulties by using an intermediate prediction step.
2Reliability
If the system processes all characteristic points in the entire image, then comprehensive lane recognition is achieved, but the recognition time increases
Solution Approach 1:
The patent segments the image processing task into two distinct regions: the lower region (near the vehicle) and the upper region (far from the vehicle). Characteristic points are extracted only from the lower region, while the upper region uses prediction based on these points. This segmentation reduces the number of points requiring direct processing while maintaining comprehensive lane recognition.
Solution Approach 2:
The patent applies partial action by extracting characteristic points only from the lower region of the image rather than processing the entire image. This partial processing is sufficient because the prediction algorithm can extend these points to cover the upper region, reducing overall processing time while maintaining recognition completeness.
3Adaptability or versatility
If the system uses conventional lane recognition methods, then it can handle clear lane lines, but it fails when lane lines are damaged or shaded
Solution Approach 1:
The patent changes the parameter of characteristic point extraction location from the entire image to specifically the lower region. This parameter change allows the system to operate in conditions where lane lines are clearly visible (lower region) and use this information to infer conditions where lane lines may be damaged or shaded (upper region), thereby improving adaptability to various road conditions.
Data Source
AI summary
A system for detecting a lane line of a road and a method thereof is disclosed. An image-retrieving device retrieves an image in front of a vehicle to generate a picture, and transmits the picture to a processing device. The processing device transforms one or two lane lines on the picture into a plurality of characteristic values, processes the characteristic values to generate one or two prediction lane lines, and uses them to dynamically predict and complete the lane lines, thereby forming a lane line prediction model. The processing device uses the lane line prediction model to determine whether the vehicle deviates, and transmits the picture and the lane line prediction model to a display device to display. When the vehicle deviates, the display device warns. The present invention can improve to recognize an unclear lane line, a single lane line and a far lane line.