Traffic Light State Estimation for Partially Captured Arrow Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cameras struggle to accurately recognize the state of arrow lights in traffic lights due to distance and angle of view, leading to incomplete capture of the traffic light image and improper recognition.
Innovation Solution
A light state recognition device that includes a determination unit to determine the light states of first and second traffic lights, an estimation unit to estimate the current light state based on past and current light states, and units to track and hold light state information, allowing for accurate recognition even when parts of the traffic light are not captured.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a camera is used to capture traffic light images, then the device can detect traffic light states, but the camera may fail to capture the entire traffic light due to distance and angle of view limitations
Solution Approach 1:
The system merges information from multiple traffic lights (first traffic light and second traffic light) to compensate for partial occlusion. When one traffic light is not fully captured, the system combines its detected state with the state of another traffic light at the same intersection to determine the current light state, thereby maintaining recognition accuracy despite limited camera coverage.
2Area of stationary object
If the camera focuses on capturing the entire traffic light, then the coverage area is maximized, but the recognition precision of arrow light states deteriorates
Solution Approach 1:
The system uses the second traffic light as an intermediary when the first traffic light is not fully captured. By referencing the synchronized operation of traffic lights at the same intersection, the system can infer the state of the occluded traffic light from the visible one, maintaining arrow light recognition accuracy without requiring the camera to capture every detail of each individual traffic light.
3Speed
If the system processes traffic light images at every processing cycle, then the response speed is maximized, but the computational complexity increases when dealing with partially captured images
Solution Approach 1:
The system performs partial processing by determining traffic light states only for the extent visible in each captured image, rather than requiring complete capture of all traffic lights. When part of a traffic light is occluded, the system processes the visible portion and supplements it with state information from other traffic lights, reducing computational complexity while maintaining response speed through incremental updates at each processing cycle.
Data Source
AI summary
A light state recognition device includes: a determination unit that determines respective light states of first and second traffic lights at predetermined processing cycles, based on an image in which the first traffic light with an arrow light that is on a closer side when viewed from a host vehicle and the second traffic light with an arrow light that is on a further side when viewed from the host vehicle are captured; and an estimation unit that estimates, in a current processing cycle, when at least part of the first traffic light is not captured in the image, a current light state, based on the light state of the first traffic light that is determined by the determination unit in a past processing cycle, and the light state of the second traffic light that is determined by the determination unit in the current processing cycle.