Multiband Stereo Depth Estimation via 2D Disparity Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multiband stereo vision systems face challenges in accurate depth estimation due to increased calculation overhead and precision errors when fusing point cloud information from different bands, particularly in adverse weather conditions.
Innovation Solution
The method involves calibrating multiband binocular cameras, conducting semi-global matching with image compression and sparse propagation to obtain disparity diagrams, which are then fused to reduce calculation overhead and enhance accuracy, eliminating the need for 3D point cloud fusion and projection mapping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If point cloud information from multiple bands is fused to obtain depth information, then depth estimation accuracy is improved, but calculation overhead and resource consumption increase significantly
Solution Approach 1:
The patent changes the dimension of fusion from 3D point cloud level to 2D disparity map level. Instead of fusing depth information directly in three-dimensional space, the method performs fusion on two-dimensional disparity maps obtained from each band, then converts the fused disparity to depth. This dimensional reduction significantly decreases computational complexity while maintaining depth estimation accuracy.
Solution Approach 2:
The patent segments the depth estimation process into separate band-specific processing streams. Each band (infrared, visible light, etc.) independently performs stereo matching to generate its own disparity map, which are then fused. This segmentation allows parallel processing of different bands, reducing overall calculation overhead compared to unified point cloud fusion.
2Reliability
If point cloud registration and fusion algorithms are used to fuse multiband binocular point cloud, then complete depth information is obtained, but resource consumption increases making it difficult to apply to actual products
Solution Approach 1:
The patent simplifies the system by operating in the 2D disparity domain rather than 3D point cloud domain. This avoids complex point cloud registration, coordinate system transformations, and 3D spatial alignment operations. The fusion is performed on 2D disparity maps with simpler pixel-wise operations, dramatically reducing system complexity for real-time applications.
3Productivity
If images are compressed before matching and sparse propagation is used, then calculation overhead is reduced, but matching precision may be affected
Solution Approach 1:
The patent applies local quality by using sparse propagation that selectively processes only certain pixel locations (e.g., sampled pixels or pixels with significant cost changes) rather than all pixels uniformly. This localized processing reduces calculation overhead while maintaining matching precision at critical locations, achieving a balance between efficiency and accuracy.
Data Source
AI summary
The present invention belongs to the field of image processing and computer vision, and discloses an acceleration method of depth estimation for multiband stereo cameras. In the process of depth estimation, during binocular stereo matching in each band, through compression of matched images, on one hand, disparity equipotential errors caused by binocular image correction can be offset to make the matching more accurate, and on the other hand, calculation overhead is reduced. In addition, before cost aggregation, cost diagrams are transversely compressed and sparsely matched, thereby reducing the calculation overhead again. Disparity diagrams obtained under different modes are fused to obtain all-weather, more complete and more accurate depth information.