Lane Line Intersection Detection Using Category Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane line intersection detection methods in intelligent driving systems are limited by the accuracy of lane line detection models, and the detection of intersections with a small number of pixels is challenging, especially when pixels are too far away or too close, leading to inaccurate geometric logic judgments.
Innovation Solution
A method and apparatus using a lane line intersection prediction model to determine coordinates and categories of lane line intersections based on a category feature map, trained with iterative learning using sample heatmaps, independent of lane line detection model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If lane line detection is performed using a lane line detection model with geometric logic judgment, then the detection process can be implemented, but detection accuracy is limited by the accuracy of the lane line detection model
Solution Approach 1:
The patent segments the detection task into two independent parts: lane line detection and intersection detection. The lane line detection model detects lane lines, and then a separate intersection detection module uses geometric logic to determine intersections. This segmentation allows each module to be optimized independently, improving overall detection accuracy without being constrained by the lane line detection model's limitations.
Solution Approach 2:
The patent introduces an intermediary step by detecting lane line segments first, then using these segments as input for intersection detection. The lane line segments serve as intermediaries that bridge the gap between raw image data and intersection detection, allowing geometric logic judgment to operate on more reliable segmented data rather than directly on pixel data.
2Productivity
If geometric logic judgment is used to determine lane line intersections, then intersection detection can be performed, but it becomes difficult when there are relatively small numbers of pixels at the intersection
Solution Approach 1:
The patent transitions from 2D pixel-based geometric judgment to a multi-dimensional approach by detecting lane line segments in multiple scales and orientations. The system detects lane lines at different resolutions and angles, then synthesizes this multi-dimensional information to determine intersections, making the detection robust regardless of the number of pixels at the intersection location.
Solution Approach 2:
The patent performs preliminary lane line detection and segmentation before intersection detection. By pre-detecting and segmenting lane lines into multiple segments with associated geometric properties, the system prepares rich feature data in advance, enabling accurate intersection detection even when the intersection itself has few pixels, as the geometric logic can operate on the pre-segmented lane line structures.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A method for detecting a lane line intersection, a device, and a storage medium are disclosed, and relate to the field of intelligent driving technologies. The method includes: determining a to-be-detected image acquired by an ego vehicle during driving; processing the to-be-detected image based on a lane line intersection prediction model to obtain a lane line intersection category feature map; and determining coordinates and a category of a lane line intersection in the to-be-detected image based on the lane line intersection category feature map.