A Fast Motion Deblurring Method in Visual SLAM

Motion blurring caused by camera shaking and scene movement often occurs, which is a key factor that reduces the accuracy of Simultaneous Localization and Mapping (SLAM). So a fast motion deblurring method in visual SLAM is proposed. Firstly, image quality is analyzed by sharpness evaluation module. Then motion deblurring network based on DeblurGANv2 is designed, which can deblur less sharpness images. Finally, the motion deblurring visual SLAM is acquired by combining above network with ORB-SLAM2. The experiments on the open dataset EuRoC show that the proposed algorithm has good deblurring effect and fast processing time. The total RMSE is 7.22cm. It can effectively eliminate the influence of blurred images on positioning accuracy.