Traffic Light Detection Using Obstruction State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing traffic light detection systems face difficulties in properly selecting a traffic light as a detection target when the view of the largest traffic light is obstructed, leading to temporary switching to the next-largest light even if the obstruction is temporary, which can result in improper selection.
Innovation Solution
A traffic light detection device and method that captures images, estimates traffic light locations using GPS and map information, sets search areas, detects traffic lights, and selects search areas based on continuous obstruction states, allowing for accurate detection by prioritizing less obstructed lights and switching only when obstructions are prolonged.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the detection target is switched to the next-largest traffic light when the largest traffic light is obstructed, then the detection can continue, but the detection accuracy deteriorates when the obstruction is temporary
Solution Approach 1:
The system performs preliminary actions by estimating traffic light locations using GPS and map information before actual detection, and by evaluating obstruction states in advance. This allows the system to predict which traffic lights will be obstructed and prepare appropriate detection strategies, avoiding unnecessary switching to secondary targets when obstructions are temporary.
Solution Approach 2:
The detection target selection is made dynamic through continuous evaluation of obstruction states. The system adapts the detection target based on real-time obstruction assessments, switching targets only when obstructions are determined to be continuous rather than temporary. This dynamic approach optimizes both detection continuity and accuracy by responding adaptively to changing conditions.
2Ease of operation
If the detection target is forced to switch to the next-largest traffic light, then the system maintains a detection target, but unnecessary switching occurs reducing detection reliability
Solution Approach 1:
The system implements feedback mechanisms by continuously evaluating obstruction states and using this information to determine whether to switch detection targets. The feedback loop assesses whether obstructions are temporary or continuous, and only triggers target switching when the obstruction persists. This feedback-based decision-making enhances detection reliability by avoiding unnecessary switches while maintaining operational simplicity.
3Measurement precision
If the system searches all potential traffic light areas, then detection coverage is maximized, but detection efficiency decreases
Solution Approach 1:
The system segments the search space by dividing potential traffic light areas into multiple candidate regions based on GPS coordinates and map information. Each segment is evaluated independently for obstruction states, allowing the system to focus computational resources on promising areas rather than uniformly searching all possible locations. This segmentation maintains comprehensive coverage while improving detection efficiency.
Solution Approach 2:
The system applies local quality assessment by evaluating obstruction states specifically in each candidate traffic light area rather than applying a uniform search strategy. Areas with no obstructions or temporary obstructions receive different treatment than areas with continuous obstructions, optimizing the detection process by concentrating efforts on high-probability detection zones while maintaining overall coverage.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A traffic light detection device capable of properly selecting a traffic light for a detection target depending on an obstruction state of the traffic light is provided. A traffic light detection device (10) includes: an image capture unit (31) that is mounted on a vehicle and captures an image of surroundings; a vehicle location detection unit (11) that detects a location of the vehicle; a map information unit (12) that stores map information; a traffic light location estimation unit (13) that estimates a location of a traffic light around the vehicle with reference to the location of the vehicle and the map information, and sets a traffic light search area in which the traffic light is estimated to be present; a traffic light detection unit (32) that detects the traffic light by searching the traffic light search area on the image; and an obstruction estimation unit (22) that estimates whether a continuous obstruction state where a view of the traffic light is continuously obstructed occurs in the traffic light search area. When the obstruction estimation unit (22) estimates that a continuous obstruction state occurs, the traffic light location estimation unit (13) selects the traffic light search area based on the continuous obstruction state.