Lane Detection Error Correction Using Semantic Segmentation Instances
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane detection technologies face challenges with camera-based errors and high construction costs of semantic segmentation databases, leading to inaccurate lane detection results.
Innovation Solution
A lane detection apparatus and method that utilizes semantic segmentation data to generate instances, correct errors through region of interest analysis, and update deep learning models to improve accuracy and reduce construction time and costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If semantic segmentation database is constructed to implement lane detection software, then lane detection capability is provided, but construction time and cost increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-processing camera images to generate semantic segmentation data before the actual lane detection operation. The system extracts lane pixels and generates semantic segmentation maps in advance, which are then stored as training data. This allows the deep learning model to be trained efficiently without requiring time-consuming manual annotation of every lane detection scenario.
Solution Approach 2:
The patent uses copying by creating synthetic semantic segmentation data from camera images through automated pixel extraction and classification. Instead of manually annotating images or requiring complex database construction, the system automatically copies and transforms raw camera data into structured semantic segmentation labels, significantly reducing construction time and cost while maintaining data quality for model training.
2Reliability
If semantic segmentation database is constructed to implement lane detection software, then lane detection capability is provided, but construction cost increases significantly
Solution Approach 1:
The patent applies self-service by enabling the system to automatically generate its own training data from camera images without external intervention. The lane detection system extracts lane pixels, classifies them into semantic categories, and generates segmentation labels autonomously. This self-service approach eliminates the need for expensive manual annotation services or complex database construction processes, significantly reducing overall system development cost.
Solution Approach 2:
The patent uses copying by creating synthetic semantic segmentation data from camera images through automated pixel extraction and classification. Instead of manually annotating images or requiring complex database construction, the system automatically copies and transforms raw camera data into structured semantic segmentation labels, significantly reducing construction time and cost while maintaining data quality for model training.
3Productivity
If lane detection is performed based on semantic segmentation data, then lane detection software can operate, but detection errors occur
Solution Approach 1:
The patent applies feedback by implementing a verification mechanism that checks detected lane pixels against the generated semantic segmentation data. The system compares the detected lane information with the semantic labels and uses this feedback to correct detection errors. This feedback loop ensures that inaccurate detections are identified and corrected, maintaining high detection precision while preserving the automated operation capability.
4Speed
If camera-based lane detection is used, then real-time detection is achieved, but errors occur in difficult lighting conditions
Solution Approach 1:
The patent applies preliminary action by pre-processing camera images to generate semantic segmentation data before the actual lane detection operation. The system extracts lane pixels and generates semantic segmentation maps in advance, which are then stored as training data. This allows the deep learning model to be trained efficiently without requiring time-consuming manual annotation of every lane detection scenario.
Solution Approach 2:
The patent applies feedback by implementing a verification mechanism that checks detected lane pixels against the generated semantic segmentation data. The system compares the detected lane information with the semantic labels and uses this feedback to correct detection errors. This feedback loop ensures that inaccurate detections are identified and corrected, maintaining high detection precision while preserving the automated operation capability.
Data Source
AI summary
A lane detection apparatus includes storage for storing data and computer-readable instructions. The lane detection apparatus also includes a processor configured to execute the computer-readable instructions to generate a lane image based on semantic segmentation data including semantic information for a lane, detect the lane using a plurality of feature points extracted from the lane image, generate one or more instance for the lane, and correct a lane detection error based on the one or more instances.