AVM Stop Line Detection Range Extension via Image Remapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stop line detection systems in autonomous vehicles are unreliable due to dirt, illumination variations, and unclear markers, leading to inaccurate edge detection and a limited detection distance, posing safety risks.
Innovation Solution
A system and method using a front view camera and a controller that extracts the upper portion of a front view image and remaps it onto an extended bird's eye view image, employing a line filter algorithm and projection algorithm to enhance stop line detection range and robustness, allowing for simultaneous detection of horizontal and vertical lines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If known edge detectors (Sobel and Canny) are used for stop line detection, then the detection process can be performed, but the detection accuracy deteriorates due to dirt, illumination variation, and noise
Solution Approach 1:
The patent transforms the 2D image detection problem into a 3D spatial problem by using depth information from stereo cameras. The stop line detection is performed in both the image plane and the depth map plane, allowing the system to distinguish actual stop lines from false detections caused by dirt or illumination variations by checking consistency across dimensions.
Solution Approach 2:
The patent introduces an intermediate depth map as a mediator between the image data and the final detection result. The depth information acts as an intermediary that helps verify whether detected edges correspond to actual stop lines or are artifacts of environmental conditions, thereby improving reliability without sacrificing precision.
2Reliability
If the detection range is extended to increase vehicle safety, then the detection distance is improved, but the processing complexity increases
Solution Approach 1:
The patent divides the extended detection range into multiple processing zones: a near-field zone processed with high precision using both cameras, and a far-field zone processed with optimized algorithms. This segmentation allows the system to maintain high safety standards across the entire range while managing processing complexity through zone-specific strategies.
Solution Approach 2:
The patent applies full stereo processing only to critical detection zones where stop lines are most relevant for immediate safety, while using simplified processing for peripheral or distant areas. This partial application of the full processing pipeline maintains vehicle safety for critical detections while reducing overall processing complexity.
3Reliability
If the detection algorithm is made more robust to handle challenging conditions, then the detection reliability is improved, but the processing time increases
Solution Approach 1:
The patent implements a multi-pass detection algorithm where the first pass performs quick filtering to eliminate obvious false detections, and subsequent passes perform more thorough analysis only on candidate regions. This periodic processing approach maintains high reliability for critical detections while reducing average processing time by avoiding exhaustive analysis of all image regions.
Solution Approach 2:
The patent performs preliminary processing of the stereo image pair to pre-compute depth maps and detect potential stop line candidates before the main detection algorithm runs. This preliminary action prepares the data in advance, allowing the main algorithm to focus only on verifying and refining candidates rather than processing the entire image from scratch, thus reducing processing time while maintaining reliability.
Data Source
AI summary
A system and method on a vehicle for providing extended detection range of a stop line on a roadway. The system and method includes a front view camera on the vehicle that provides images of the roadway in front of the vehicle and a controller that receives images from the front view camera. The controller is programmed to extract an upper portion of a front view image that is remapped onto an extended bird's eye view image such that the detection range of the stop line is detected.