Research And Application On Image Stitching

Robot vision navigation is an important research area of mobile robot autonomous navigation technology, ordinary equipment with limited shooting area is difficult to meet the needs of global location and navigation. The appearance of the stitching technique makes it possible to obtain wide field scene under the condition of not changing hardware.Image stitching technology mixes a series of images which are overlapped with each other into one image which has a wide viewing angle, low distortion and no sutures. We put forward the corresponding stitching algorithms according to different problems.In order to solve stitching algorithm signs while large differences in brightness or real-time problem, We present a fast image stitching algorithm based on ORB and OECF model. A image stitching system is development by using HALCON library and the MFC framework. Firstly, feature points are extracted by SIFT algorithm.Secondly, the binary descriptors of the feature points are obtained by ORB algorithm to achieve images matches. The project transformation model matrix between two matching images is calculated. Finally, to solve the color inconsistency problem of images stitching, the Opto-Electronic Conversion Function of the camera is estimated to conduct color adjustment. Experimental results show that the proposed algorithm is effective for image stitching,moreover the stitching quality and the stitching speed are much better than other image stitching algorithms.For fuzzy problem in linear scale-space edge, we propose a simple and effective algorithm AKAZE stitching. Firstly, FED algorithm is proposed to accelerate the rate of formation of nonlinear scale space, with the Hessian matrix extracting feature points, and then feature vectors are constructed using the M-LDB descriptor.Secondly,the matched pairs are extracted by computing the Hamming distance between two feature vectors.Thirdly,wrong matches are eliminated by RANSAC and the best transform parameters are estimated to accomplish the registration process.Finally, images are fused by the method of weighted fusion. A performance comparison test can be conducted aiming at KAZE, SIFT, SURF, ORB, BRISK.Experimental results show that the proposed algorithm has better robustness for the various transform, and the processing time is greatly reduced.Experiments show that the algorithm not only realize accurate image stitching,but also improve the real-time and robustness of system,the results also have no obvious signs of stitching. The algorithm can be applied to a variety of environments.Practical application results of this algorithm meet the requirements of road detection in mobile robot vision navigation. That is to say, this algorithm can be applied to a variety of scenarios. Large field of vision provide convenience for visual navigation and path planning.