Traffic Light Recognition Using Semantic Aggregation and Color Space Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned vehicles face challenges in accurately recognizing traffic light states, especially when lights are damaged, abnormally displayed, covered, or in a blink state, leading to incorrect state detection.
Innovation Solution
A method and apparatus for recognizing traffic lights that involve obtaining semantic information from images captured by a camera, determining uniform states of traffic lights based on preset confidences, and generating indication information for navigation, which includes correcting for historical states and identifying abnormal conditions like blinking or shielding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple color recognition is used for traffic light detection, then the device complexity is low, but the measurement precision deteriorates when lights are damaged, covered, or abnormally displayed
Solution Approach 1:
The detection system is segmented into multiple functional modules: a detection module that captures traffic light images, a recognition module that performs color space conversion and state judgment, and a determination module that aggregates results from multiple traffic lights. This segmentation allows each module to specialize in specific tasks, improving overall accuracy without requiring a completely complex system redesign.
Solution Approach 2:
The system performs more color space conversions than strictly necessary (converting to multiple color spaces like HSV, YCrCb, and L*a*b*) and evaluates multiple potential states for each traffic light. This excessive action ensures that even if some detection attempts fail due to damaged or covered lights, the system has multiple opportunities to correctly identify the traffic light state.
2Speed
If multiple traffic lights are detected independently, then the detection speed is high, but the reliability deteriorates when some lights are damaged or abnormally displayed
Solution Approach 1:
The system merges the detection results of multiple traffic lights through the determination module. Instead of treating each traffic light detection as a completely independent decision, the system aggregates results and determines a unified state. This combining approach ensures that if some traffic lights are damaged or abnormal, the overall system can still reach a reliable conclusion based on the collective information from all detected lights.
Solution Approach 2:
The system implements feedback mechanisms where the recognition module continuously evaluates traffic light states and adjusts its determination based on aggregated results. The determination module uses feedback from multiple detection attempts and color space analyses to refine the final state judgment, improving reliability while maintaining detection speed through efficient information integration.
3Measurement precision
If color space conversion is performed for each detected traffic light, then the measurement precision is improved, but the use of energy increases due to multiple conversions and state evaluations
Solution Approach 1:
The system performs preliminary color space conversions to multiple color spaces simultaneously rather than sequentially attempting different conversions. By preparing all necessary color space representations in advance, the system avoids redundant conversion operations and reduces overall energy consumption while maintaining high measurement precision for traffic light state detection.
Data Source
AI summary
Embodiments of the present disclosure provide a method and an apparatus for recognizing a traffic light, a device, a storage medium, and a vehicle. The method may include: obtaining semantic information of one or more traffic lights provided at an intersection, and recognizing states of the one or more traffic lights in an image captured by a camera; determining traffic lights representing same semantics from the one or more traffic lights based on the semantic information, and determining a uniform state corresponding to the traffic lights representing the same semantics according to states of the traffic lights representing the same semantics and preset confidences of respective traffic light states; and determining indication information of the traffic lights representing the same semantics according to the uniform state.