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 small number of pixels at intersections in images complicates accurate geometric logic judgment, leading to reduced detection accuracy.
Innovation Solution
A lane line intersection prediction model is used to process images and generate a category feature map, allowing for the determination of intersection coordinates and categories based on the model's accuracy, independent of the lane line detection model's precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If lane line detection model is used for intersection detection, then detection process is simple, but detection accuracy is limited by lane line detection model accuracy
Solution Approach 1:
The patent segments the detection task into two independent parts: lane line detection and intersection detection. The lane line detection model processes images to extract lane line information, while a separate intersection detection model processes the same images to detect intersections. This segmentation allows each model to be optimized for its specific function, improving overall detection accuracy without complicating the overall process.
Solution Approach 2:
The patent introduces a feature map as an intermediary between lane line detection and intersection detection. The feature map contains enriched features extracted from the image that serve as input to the intersection detection model. This intermediary enables the intersection detection model to access detailed visual information without being constrained by the lane line detection model's accuracy limitations.
2Device complexity
If geometric logic judgment is used for intersection detection, then detection method is simple, but detection accuracy deteriorates when few pixels exist at intersection
Solution Approach 1:
The patent replaces geometric logic judgment (a mechanical/mathematical approach) with a deep learning-based detection model. The neural network model processes image data through multiple layers to automatically learn intersection patterns, eliminating the need for complex geometric calculations and threshold-based judgments that fail when pixel counts are low.
Solution Approach 2:
The patent changes the detection approach from pixel-count-based geometric judgment to feature-based neural network classification. By transforming the input from raw pixel data to enriched feature maps and using neural network parameters (weights and biases) to determine intersections, the system becomes insensitive to the number of pixels at intersection locations.
3Measurement precision
If lane line detection model accuracy is improved, then intersection detection accuracy improves, but computational cost increases
Solution Approach 1:
The patent segments the computational workload into two separate models: a lane line detection model and an intersection detection model. Each model processes the same image independently, allowing them to be optimized for their specific tasks. This segmentation prevents the need to reprocess all image data multiple times through a single complex model, reducing overall computational cost.
Solution Approach 2:
The patent performs preliminary feature extraction and representation learning in the lane line detection model, which then serves as input to the intersection detection model. By pre-processing the image data to extract meaningful features before intersection detection, the system reduces the computational burden on the intersection detection model compared to processing raw images from scratch.
Data Source
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.