Panoramic Image Mosaics In The Ship Image System Based On Sift Algorithm

With the rapid development of the shipbuilding industry and marine economy,the importance of maritime traffic safety cannot be ignored.How to avoid collisions,plan routes and secure parking in accordance with the surrounding conditions has become a very important preventive measure before accidents.In the ship image system,the image mosaic technology is an effective way to generate high-resolution panoramic images,and it is also an important measure to assist the crew in driving safely and strengthen the informatization of ships.Taken image stitching based on SIFT feature extraction algorithm as the research object,and taken the peripheral scene of the ship as its application background,this paper deeply studies the key steps such as image registration,the elimination of mismatching points,the calculation of transformation matrix in image stitching,and proposes several optimization methods.The main contents are as follows:(1)The invariability of the traditional SIFT algorithm,such as rotation and scale change,was verified experimentally.Combined with the purpose of image registration,the insufficiency of SIFT in the image registration process was analyzed in detail.(2)The feature point extraction range of SIFT is the entire image to be stitched,but the search and matching of the feature points in the non-overlapped area between images will result in more time-consuming in image registration.In order to solve this problem,this paper proposes a SIFT optimization method for dividing overlapping areas.While using the hierarchical comparison method to obtain the rough matching results of images,the rough matching results of each layer are used to demarcate the scope of local extreme point extraction by SIFT,and the steps of the hierarchical comparison method and the SIFT algorithm are organically combined to avoid waste of time in pointless feature extraction.Through the comparison experiments between the optimization algorithm and the traditional SIFT algorithm,it is proved that the registration efficiency of this algorithm is at least twice that of the traditional algorithm.(3)Aiming at the problem of wrong matching points appear easily when SIFT is used in feature matching,a mismatching points elimination algorithm is proposed in this paper.The algorithm can effectively eliminate two types of mismatching point pairs.For a class of mismatching points that do not correspond to each other,The voting mechanism of Generalized Hough Transform(GHT)is introduced to find the rotation and scaling of the image and locate where the template image appears in the scene in order to completely reject unmatched points.For a type of mismatching point pairs caused by rotation and scaling,this paper rejects the wrong matching points by the difference between the correct matching pair and the wrong matching pair.Finally,by comparing with several common elimination algorithms,it is verified that the proposed algorithm has a higher correct matching rate and a lower false rejection rate.(4)The above-mentioned image registration optimization method and mismatching points elimination algorithm are used in the process of panorama stitching of ship images.Using VS2010 as the experimental platform,the image mosaic experiments were performed under the conditions of multiple images,similar objects,light changes,viewing angle changes,zoom changes,translation changes,rotation changes,and jitter.The experimental results show that the stitched images are smooth without ghosting,and there is no obvious stitching gap.Compared to the traditional method,the splicing speed has improved significantly.The research results of this paper are of great significance to the research work of advancing the ship image system.