Lane Detection Using Semantic Segmentation and Shape Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane detection methods are prone to errors, particularly in varying environments, and struggle to accurately identify lane lines with complex shapes or in the presence of similar objects like railings.
Innovation Solution
A method implemented on a computing device using a trained semantic segmentation network to determine the semantic category of each pixel in an image, followed by binarization and fitting operations to accurately detect lane lines of various shapes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If deep learning models are used for lane detection, then detection capability is improved, but error rate increases in complex scenarios
Solution Approach 1:
The patent segments the lane detection task into multiple processing stages: initial detection by deep learning model, extraction of detection results, identification of candidate regions, and verification through shape analysis. This multi-stage segmentation allows the system to leverage the strengths of deep learning while adding verification steps to eliminate false positives, thereby improving reliability without sacrificing detection capability.
Solution Approach 2:
The patent introduces an intermediary verification mechanism that acts as a mediator between the deep learning model and final detection results. This intermediary layer performs shape analysis and candidate region verification to filter out false detections from objects like railings and trees, resolving the contradiction between detection capability and reliability.
2Measurement precision
If lane detection methods are designed for specific scenarios, then detection accuracy is improved, but adaptability to varying environments deteriorates
Solution Approach 1:
The patent designs a universal lane detection method that can adapt to various environments and lane types. The shape-based verification mechanism and candidate region identification approach work effectively across different scenarios including straight lanes, curved lanes, and complex intersections, making the system multi-functional and environmentally adaptable while maintaining high detection accuracy.
Solution Approach 2:
The patent implements dynamic adaptation through the verification stage, where detection parameters and candidate region criteria are adjusted based on the specific scenario. The system dynamically identifies candidate regions and verifies them using shape analysis that adapts to different lane configurations, enabling the method to maintain high accuracy across varying environments.
3Productivity
If objects with similar appearance to lane lines are detected using surrounding context, then detection speed is improved, but false identification rate increases
Solution Approach 1:
The patent extracts and analyzes the shape characteristics of detected objects as a separate verification step. By taking out the shape analysis from the initial detection process and using it as an independent verification criterion, the system can quickly identify candidate regions while accurately distinguishing lane lines from similar objects like railings, maintaining detection speed while improving identification accuracy.
Solution Approach 2:
The patent uses shape templates and candidate region patterns as reference copies for verification. By comparing detected objects against known lane line shape patterns, the system can rapidly verify detections without relying solely on surrounding context, thereby maintaining detection speed while reducing false identification of objects with similar appearance.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods for lane detection are provided. The methods may include obtaining an image and determining, for each of a plurality of pixels in an image, a semantic category using a trained semantic segmentation network. The methods may further include determining one or more pixel sets based on the plurality of pixels according to a predetermined rule. Each of the one or more pixel sets includes one or more pixels of a same semantic category. The methods may further include, in response to a determination that there are one or more erroneous pixels in a pixel set, removing the one or more erroneous pixels from the pixel set to obtain a fitting line corresponding to a lane line. The methods may further include determining a position of the lane line based on the fitting line.