Improved ORB-SLAM algorithm with deblurring image

Simultaneous Localization and Mapping (SLAM) technology is a key technique that utilizes scene information to achieve autonomous device localization and construct environmental maps in unknown environments. The ORB-SLAM algorithm based on image information has the advantages of small information processing and accurate positioning. However, the ORB-SLAM algorithm heavily relies on the clarity of the image, and moving cameras can easily cause image motion blur. The ORB-SLAM algorithm based on feature point extraction suffers from image blur during camera motion, resulting in inaccurate feature point extraction and inaccurate system tracking and positioning. To this end, an improved ORB-SLAM algorithm based on Wiener filtering for deblurring is proposed, which combines the attitude data provided by the inertial navigation module. By combining the previous frame attitude data of the SLAM system with the attitude data of the inertial navigation module, the motion blur degradation point spread function is further inferred. The fuzzy degradation point spread function is applied to the improved Wiener filtering algorithm, which effectively improves the image quality and solves the problem of inaccurate feature point extraction to a certain extent. Experimental comparisons were conducted on the TUM dataset, and the results showed that the algorithm effectively improved the tracking and positioning accuracy of the ORB-SLAM algorithm while ensuring real-time performance.