Top-View Image Generation Using Ground Feature Point Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The degradation in display quality of top-view images generated by a vehicle's around view monitoring system due to camera tolerance issues caused by vehicle vibrations and camera adjustments is not effectively addressed by existing technologies.
Innovation Solution
A method and apparatus that automatically correct camera posture angles in real-time by detecting ground feature points and lanes from images captured by multiple cameras, using algorithms for rotation-conversion and synthesis to generate a stable top-view image, even when ground feature points are not detectable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image registration processing technique is used to generate top-view image, then top-view image can be generated from multiple camera images, but camera tolerance accumulation degrades registration accuracy and display quality
Solution Approach 1:
The system performs preliminary calibration to establish initial camera parameters and tolerance values before actual operation. This preliminary setup creates a baseline registration model that compensates for expected camera tolerances, allowing the system to maintain accurate image registration even as tolerances accumulate during vehicle operation.
Solution Approach 2:
The system continuously monitors image registration quality by analyzing feature point correspondence between adjacent frames. When registration deviation exceeds a threshold, the system automatically adjusts camera tolerance parameters to restore accurate alignment. This closed-loop feedback mechanism ensures sustained registration accuracy despite accumulating camera tolerances.
2Measurement precision
If manual camera adjustment is performed to correct tolerance, then registration accuracy can be improved, but the system requires manual intervention and cannot operate continuously
Solution Approach 1:
The system automatically detects registration deviations by analyzing feature points in consecutive image frames and self-corrects by adjusting camera tolerance parameters without manual intervention. The algorithm identifies when images need realignment and performs the correction autonomously, enabling continuous automatic operation while maintaining high registration accuracy.
3Measurement precision
If ground feature points are used for camera posture estimation, then accurate top-view image generation is achieved, but the method fails when ground feature points are not detectable
Solution Approach 1:
The system introduces lane detection as an intermediary method that bridges the gap when ground feature points are unavailable. By detecting lane markings in the image and using their geometric properties to estimate camera posture, the system maintains operational capability in diverse conditions such as snowy roads or areas with poor ground texture, thereby expanding adaptability while preserving estimation accuracy.
Data Source
AI summary
Provided is a top-view image generating method. The top-view image generating method estimates a posture angle of a camera providing at least one image, based on a ground feature point detected from the at least one image of front, rear, left, and right images and estimates a posture angle of a camera providing the other image from which the ground feature point is not detected. Subsequently, based on the estimated posture angles, the front, rear, left, and right images are rotation-converted, and then, a top-view image is generated by synthesizing images obtained through the rotation-conversion.