Visible and Infrared Images Matching Method for Maritime Ship Targets
To address the limitations of the traditional SURF algorithm,such as lengthy computation times and low regis-tration accuracy in matching maritime ship images,an enhanced Canny-SURF-RANSAC algorithm is proposed.To tackle the issue of traditional Canny algorithms generating numerous non-ship target edges due to undulating waves,an im-proved Canny algorithm is introduced for detecting edges of maritime ships.The continuity of ship edges is optimized,and irrelevant noise is eliminated by incorporating morphological filtering operations,such as opening and closing,and by using the Otsu threshold method for adaptive threshold setting.Based on the edge information of ship targets,the SURF algorithm is utilized to extract feature points and perform feature matching.The traditional RANSAC algorithm is modified to dynamically adjust the threshold,i.e.,an adaptive RANSAC algorithm is used to eliminate incorrect matches in the results,thereby enhancing the accuracy of matching.Experiments demonstrate that this image matching technique achieves faster matching times and higher accuracy rates under complex scenarios involving the same ship in different frames of the same sensor,the same ship from different perspectives,and the same ship target captured by visible light/in-frared sensors,compared to traditional methods.After matching,the connected domains of the same ship targets are delin-eated to obtain ship detection results,thereby facilitating automatic annotation of target ships.
万方数据
维普
