Traffic Light Recognition Using Dynamic ROI Across Video Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems face challenges in accurately recognizing tiny traffic lights due to their small size in images captured by cameras, which affects the system's ability to make timely and accurate driving decisions.
Innovation Solution
The method involves an object recognition model that estimates a region of interest (RoI) based on prior information about traffic light objects, adjusts the RoI using detection results from previous frames, and recognizes traffic lights by processing an enlarged target image within the determined RoI.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If the camera captures the entire scene to ensure traffic lights are within the field of view, then the coverage area is improved, but the resolution of distant traffic lights deteriorates
Solution Approach 1:
The patent divides the image processing into two stages: first capturing the entire scene to identify potential traffic light regions, then selectively enlarging only those specific regions for detailed recognition. This segmentation allows the system to maintain both broad coverage and high resolution for critical objects.
Solution Approach 2:
The patent transitions from processing the entire image at original resolution to extracting and enlarging specific regions of interest. By focusing computational resources on selected areas rather than the whole image, the system achieves higher effective resolution for traffic lights without sacrificing overall scene coverage.
2Reliability
If the object recognition model processes the entire input image to ensure no traffic lights are missed, then the detection completeness is improved, but the processing time increases
Solution Approach 1:
The patent extracts only the relevant regions containing traffic lights from the full image for detailed processing. By using the initial full-image analysis to identify traffic light locations, the system can then focus subsequent processing only on those specific regions, significantly reducing overall processing time while maintaining detection completeness.
Solution Approach 2:
The patent performs a preliminary analysis of the entire image to identify potential traffic light regions before conducting detailed recognition. This preliminary step allows the system to prepare region-specific processing in advance, optimizing the balance between comprehensive detection and processing efficiency.
3Productivity
If the object recognition model processes only specific regions to reduce computation, then the processing speed is improved, but the risk of missing distant traffic lights increases
Solution Approach 1:
The patent uses detection results from previous frames as feedback to guide region selection in the current frame. This temporal feedback mechanism ensures that regions containing traffic lights are consistently identified and processed, maintaining high detection accuracy while enabling focused processing that improves speed.
Solution Approach 2:
The patent dynamically adjusts which regions are selected for detailed processing based on real-time detection needs and historical information. Rather than using a fixed processing strategy, the system adapts its region selection to the specific driving context, ensuring optimal balance between speed and accuracy.
Data Source
AI summary
Disclosed is a method of detecting a traffic light with an object recognition model configured to recognize traffic lights. The method includes: obtaining an input image from a camera included in a vehicle, the input image among frames, including previous frames, captured by the camera; estimating, based on prior information about traffic light objects, a first region of interest (RoI) for the input image; determining a second RoI based on the first RoI and based on detection results of the previous frames, wherein the detection of results correspond to recognition results of recognizing traffic lights in the previous frames by the object recognition model; and recognizing, by the object recognition model, a traffic light in the input image, wherein the recognizing is based on the input image and the second RoI.