High-Precision Traffic Light Recognition Using Map-Stored Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving technologies face challenges in accurately recognizing traffic lights, which is crucial for safe and efficient navigation, due to the complexity and variability of traffic light states and environments.
Innovation Solution
A high-precision map generation method and system that utilizes road test data to create a recognition model for traffic lights, including intelligent models and state machines, which are stored in a high-precision map to enable accurate identification of traffic light states and changes, using camera devices and a main control device to capture and analyze images in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional traffic light recognition methods are used, then the system can operate with simpler components, but the recognition accuracy and reliability are insufficient due to complexity and variability of traffic light states and environments
Solution Approach 1:
The patent applies preliminary action by pre-building a recognition model offline using road test data, marked data, and training processes. This pre-processing allows the system to have accurate recognition capabilities ready before actual deployment, resolving the contradiction by preparing complex computational work in advance rather than during real-time operation
Solution Approach 2:
The patent uses copying by creating a virtual recognition model that replicates traffic light recognition capabilities without requiring physical analysis of every traffic light instance. The model stores learned patterns and characteristics, allowing accurate recognition through data copying rather than complex real-time physical analysis
2Reliability
If real-time recognition of various traffic light states is achieved, then the system can make accurate driving decisions, but the processing time and computational resources increase
Solution Approach 1:
The recognition model is trained offline using marked data containing various traffic light states and environments. This preliminary training action transfers complex computational work to the offline phase, enabling fast and accurate real-time recognition during actual operation without time loss
Solution Approach 2:
The patent changes parameters by transforming raw video data into marked data with labeled states, colors, and characteristics during the training phase. This parameter transformation creates a structured recognition model that can quickly classify traffic light states during real-time operation, reducing processing time while maintaining reliability
Data Source
AI summary
A generation method of high-precision map for recognizing traffic lights is provided. The generation method comprised steps of: obtaining road test data comprising video data of traffic lights; marking the video data in order to obtain marked data of the traffic lights, the marked data comprising states of the traffic lights and traffic lights information; using the video data and the marked data to generate a recognition model of the traffic lights; and storing the recognition model and the traffic lights information in a high-precision map to generate a high-precision map for recognizing the traffic lights. Furthermore, a method and system for recognizing traffic lights using high-precision map are also provided. The recognition model is stored in the high-precision map, and cooperating with the high-precision map to effectively recognize the traffic lights.