Traffic Signal Detection Using Map Confidence and Sensor Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle systems face challenges in accurately detecting and interpreting traffic signals, particularly in situations with visual obstructions or false positives, which can lead to incorrect navigation decisions.
Innovation Solution
The method involves scanning a target area using sensors to detect traffic signals, determining their location and state, and assessing confidence in the detection by comparing it to known locations, using a combination of imaging and 3D scanning technologies, and adjusting vehicle control based on the signal state and confidence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template-matching algorithm is applied to detect traffic lights, then detection speed is improved, but false positives increase due to visual obstructions or similar objects
Solution Approach 1:
The system uses a histogram filter that incorporates feedback from multiple sources including template matching results, geometric constraints, and temporal consistency to adjust detection confidence. The filter multiplies template matching results with histogram data and applies blurring to account for uncertainty, creating a feedback loop that refines detection accuracy while maintaining speed.
Solution Approach 2:
The patent introduces an intermediary histogram filter that mediates between template matching results and final detection decisions. This filter acts as a buffer that smooths out false positives by comparing detection results against historical data and geometric constraints, reducing the direct impact of template matching errors while preserving detection speed.
2Reliability
If prior map of traffic light locations is used, then false positives are reduced, but system fails when traffic lights are relocated or missing from map
Solution Approach 1:
The system dynamically adjusts the region of interest and detection parameters based on vehicle location and map data. When approaching an intersection, the system activates detection in specific regions based on map information, but maintains the capability to detect lights outside expected locations. The histogram filter is continuously updated to adapt to new traffic light positions while maintaining reliability for known locations.
Solution Approach 2:
The system performs preliminary detection in regions of interest defined by map data before final confirmation. By pre-defining search areas based on known traffic light locations, the system reduces false positives from unrelated objects. Simultaneously, the system maintains vigilance for lights outside these regions by using broader scanning parameters, enabling detection of relocated lights while preserving false positive reduction benefits.
3Reliability
If conservative action is taken when traffic light detection is uncertain, then safety is improved, but navigation efficiency decreases due to unnecessary braking
Solution Approach 1:
The system changes detection parameters dynamically based on confidence levels. When confidence is high (strong template match with consistent histogram data), the system uses standard detection parameters allowing smooth navigation. When confidence is low (ambiguous results or obstructions detected), the system switches to conservative parameters including reduced speed and increased scanning frequency, balancing safety with navigation efficiency through parameter adaptation.
Data Source
Figure 1
Figure 2A~2B
Figure 2C
AI summary
A camera viewing the area in front of a vehicle equipped with an autopilot is used to register images from the environment to detect traffic lights, determine their location and their state and determine whether there is a known location within a threshold distance to the determined location of the detected traffic light of a known traffic light. Upon determining that there is no known traffic light at proximity of the detection traffic light location, obtaining additional information about the detected traffic light.