Based On The The Wan Scene Of The Video Image Sequence Automatically Stitching Technology, And Target Recognition

This paper studies the technology of the wide-area scene mosaics and the target extraction based on the video image sequences.Automatic wide-area scene mosaic based on the video image sequence and the video image sequence object extraction technology is researched in the paper.Firstly, the development of image mosaic is summarized. The process of image mosaic is overviewed, including theoretical basis, image acquisition, geometric correction, image registration, image fusion.Secondly, Harris corner detection algorithm is highlighted. The normalized correlation matching algorithm and singular value decomposition-based corner matching algorithm are compared with experiment. An improved method of weighted smoothing fusion algorithm is presented to eliminate the gap between images. Compared with the original and improved method above, the results shows that the improved method can effectively eliminate the gap under there is a rotation between the two images. The result also shows that the improved method is valid and robust.Thirdly, we analyze cumulative error of multiple images. The method of combination of neighbor small image stitch and mosaic is presented with the global strategy for fast image mosaic.Lastly, we research on the detection of move targets. First, we use the median filtering to eliminate noise of the video image. And then, we detect the move targets with the method of combination of two difference and background difference. Finally, we use shape center to locate the target and acquire the move status of targets.Using VC++6.0 software to realize the image mosaic and target extraction, the summary of the paper and the prospects for future work are given in the final part.