Asynchronous Stereo Vision Disparity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual perception processing systems for moving robots face challenges in accurately calculating three-dimensional position information due to errors caused by differences in frame time points when using images from multiple cameras, which increases processing burden and costs.
Innovation Solution
An image processing apparatus that acquires images from multiple cameras at different frame time points, measures disparity using a disparity measuring device, corrects disparities using motion information, and calculates three-dimensional position information through virtual synchronization, allowing for accurate positioning even with asynchronous image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If images are taken alternately from multiple cameras to reduce processing burden, then processing capacity requirement is reduced, but errors occur due to frame time point differences
Solution Approach 1:
The patent applies preliminary action by performing disparity correction using motion information before final three-dimensional position calculation. The motion information calculating device pre-calculates motion vectors from temporal image sequences, and the disparity correcting device uses these pre-calculated motion vectors to correct disparities arising from frame time differences, thereby resolving the accuracy issue before the final positioning step
Solution Approach 2:
The patent introduces motion information as an intermediary element between the raw images and the final three-dimensional position calculation. The motion information calculating device extracts motion vectors from temporal sequences, and these motion vectors serve as a mediator to correct the disparities in the disparity measuring device, enabling accurate positioning despite asynchronous frame capture
2Measurement precision
If processor performance is enhanced to improve processing accuracy, then measurement precision is improved, but product costs increase
Solution Approach 1:
The patent changes the processing parameters by introducing temporal dimension utilization - processing images from multiple time points (t-Δt, t, t+αt) rather than simultaneous images. This parameter change enables the use of motion information to correct disparities, achieving high measurement precision through algorithmic improvement rather than hardware enhancement, thus avoiding increased product costs
3Productivity
If simple stereo image process is performed with asynchronous images, then processing speed is improved, but errors occur due to frame time point differences
Solution Approach 1:
The patent maintains continuous useful action by processing asynchronous images from multiple cameras without requiring frame synchronization. The system continuously captures images at different time points (t-Δt, t, t+αt) and processes them through motion information calculation and disparity correction, maintaining both high processing speed and accuracy through uninterrupted asynchronous processing
Data Source
AI summary
A first image obtained by taking a moving object at a frame time point t by a first imaging device, a second image obtained by taking the moving object at a frame time point t+Δt by a second imaging device having a view point different from that of the first imaging device, and a third image obtained by taking the moving object at a frame time point t−Δt by the second imaging device are acquired. A first corresponding point on the first image corresponding to an arbitrary point on the second image is searched for, and the disparity of the arbitrary point is measured. Motion information of the arbitrary point is calculated on the basis of the third image, and by use of the motion information as a correction amount, the disparity is corrected. Three-dimensional position information of the arbitrary point is calculated on the basis of the corrected disparity.