一种基于多级关键点匹配和重构网络的移动机器人图像拼接系统和方法
This image stitching method, which uses multi-level keypoint matching and reconstruction networks, solves the ghosting problem in traditional image stitching methods, enabling high-quality image stitching for mobile robots in complex environments. It is applicable to images from any viewpoint and at different depths, thus improving the accuracy of autonomous navigation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI UNIV
- Filing Date
- 2023-01-31
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional image stitching methods cannot effectively handle the ghosting phenomenon caused by changes in depth and posture of mobile robots in complex environments, and it is difficult to achieve image stitching from arbitrary perspectives, which affects the accuracy of autonomous navigation.
An image stitching method based on multi-level keypoint matching and reconstruction network is adopted. The homography offset information is obtained through multi-stage keypoint matching. Combined with dynamic fusion and scale transformation, the multi-level reconstruction network is used to optimize the stitched image, ensuring the semantic structure and texture details of the stitched image.
It effectively overcomes the ghosting defect of traditional algorithms in image stitching at different depths, can be applied to image stitching from any viewpoint, improves the quality and accuracy of stitched images, and is suitable for autonomous navigation of mobile robots.
Smart Images
Figure CN118608737B_ABST