海上舰船目标可见光/红外图像匹配方法  

Visible and Infrared Images Matching Method for Maritime Ship Targets

作　　者：于乐凯 曹政 孙艳丽 刘宁波 王中训[1,2] YU Lekai;CAO Zheng;SUN Yanli;LIU Ningbo;WANG Zhongxun(School of Physics and Electronic Information,Yantai University;Shandong Data Open Innovation Application Laboratory of Smart grid Advanced Technology,Yantai University,Yantai Shandong 264005,China;Naval Aviation University,Yantai Shandong 264005,China)

机构地区：[1]烟台大学物理与电子信息学院 [2]烟台大学智慧电网先进技术山东省数据开放创新应用实验室,山东烟台264005 [3]海军航空大学,山东烟台264001

出　　处：《海军航空大学学报》2024年第6期755-764,772,共11页Journal of Naval Aviation University

基　　金：国家自然科学基金(62388102、62101583、61871392);泰山学者工程(tsqn202211246)。

摘　　要：为改善传统SURF算法匹配海上舰船图像存在计算时间长、配准精度低等不足,提出一种改进Canny-SURF-RANSAC算法。针对起伏海浪导致传统Canny算法产生大量非舰船目标边缘的问题,提出改进的Canny算法进行海面舰船边缘检测。通过引入形态学滤波中的开运算和闭运算的复合操作,以及使用Otsu阈值方法进行自适应阈值设置,优化了舰船的边缘连续性并消除了无关噪点。在舰船目标边缘信息的基础上,运用SURF算法提取特征点,进行特征匹配,在传统的RANSAC算法采用动态调整阈值即自适应RANSAC算法消除匹配结果中错误匹配点,提高匹配的准确率。实验证明,文章图像匹配技术匹配算法在同一传感器不同帧中、不同视角出现的同一舰船、可见光/红外传感器采集到的同一舰船目标的复杂场景下,比传统方法具有更快的匹配时间、更高的匹配正确率。完成匹配后,框定相同舰船目标连通域,得到舰船检测结果。To address the limitations of the traditional SURF algorithm,such as lengthy computation times and low registration 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 improved 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/infrared sensors,compared to traditional methods.After matching,the connected domains of the same ship targets are delineated to obtain ship detection results,thereby facilitating automatic annotation of target ships.

关 键 词：SURF算法 海上舰船目标图像 CANNY算法 自适应RANSAC算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]