Underground video stitching algorithm based on improved speeded up robust features

For the problem of low real—time performance of speeded up robust features(SURF) algorithm used in underground video stitching, the SURF algorithm was improved by decreasing dimensions of feature points and extracting feature points only in region of interest. On this basis, an underground video stitching algorithm was proposed.Firstly, feature points of video images are extracted by using the improved SURF algorithm. Then the number of feature points is dynamically tracked. If the number of feature points in the non—first frame image exceeds the threshold, some operations will be performed again including feature point matching and purifying and calculation and storage of projective transformation matrix. Otherwise, the projective transformation matrix from the previous frame is used. Finally, images are fused by gradual weighted average fusion method to generate a stitched video. The experimental results show that the underground video stitching algorithm based on the improved SURF algorithm has high real—time performance and good stitching effect.