Research On Image Stitching Technology

Panoramic image mosaic technology can get panoramic information in real time, which is widely used in tank sighting system, vehicle safety, geographic information acquisition and so on. This paper makes a further study in the distortion correction of large field of view and short focal length lens, and discusses the subject of panoramic image mosaic.In this paper, panoramic image stitching system, including the two parts, the image acquisition system and image processing system, is introduced first. Secondly, the part of the image processing system is specifically addressed and proposing a viable high-speed mosaic method for distorted image based on ORB feature:Before the entire process, the camera we want to use is calibrated using Zhang’s calibration method, and the interior camera parameters is determined. The gray centroid method is used to get the sub-pixel accuracy corner, which simplifies the calculation. Next, completing image mosaic:firstly, apply median filtering and histogram equalization to the acquired images. Then, correct the distorted images using the interior camera parameters just obtained. The cylindrical projection is applied to the corrected images and ORB algorithm is used to detect images’ feature points of projected images. Next, preliminary match feature points with brute-force method according to the principle of minimum Hamming distance. Apply USAC algorithm to the matched points to calculate the transformation matrix of projection between images. Finally, a weighted average fusion algorithm is used to synthesize the adjacent images. At the end of the paper, the feasibility of the above image mosic algorithm is proved by experiments.