多条带侧扫声呐图像精拼接方法研究  

作　　者：高飞 王晓 杨敬华 张博宇 周海波 陈佳星 

机构地区：[1]青岛市勘察测绘研究院,山东青岛266032 [2]江苏海洋大学海洋技术与测绘学院,江苏连云港222005

出　　处：《科技创新与应用》2021年第5期1-4,共4页Technology Innovation and Application

基　　金：国家自然科学青年基金(编号:41806117)。

摘　　要：针对侧扫声呐图像地理编码拼接存在相邻条带共视目标错位,特征拼接存在效率低且位置漂移问题,给出一种基于几何变换的分段Speed-up Robust Features(SURF)特征匹配方法,提高匹配效率和精度;提出一种多条带图像分组拼接策略,实现联合SURF特征和地理编码的多条带图像拼接。实验验证了方法的有效性,5个条带图像SURF特征匹配耗时从135s缩减到30s,达到程序实时处理的要求;5个条带图像拼接后共视特征点对平均坐标偏差从传统SURF特征匹配的6.18m降低到1.07m,实现了位置统一;实现了大区域海底地貌“一张图”的精细获取。For the problem of Side Scan Sonar(SSS)image mosaic method based on geocoding image leads to the dislocation of the corresponding features in adjacent SSS strip,while the method of feature mosaic leads to the problem of low efficiency and migrating coordinates.A Speed-up Robust Features(SURF)matching method based on geometric transformation and block splicing image is given to improve the matching efficiency and accuracy.A grouping method for multi-strip SSS image is given for combining geocoding and feature mosaic of large areas SSS images.Experimental results show the effectiveness of the proposed method.The time-consuming for SURF feature matching of 5 strip images reduces from 135s to 30s,which meets the requirements of real-time processing.The average coordinate deviation of the corresponding feature points after 5 strip images mosaic is reduced from 6.18 m of traditional SURF to 1.07 m of improved SURF mosaic,and it is achieving a unified location.The fine sea bottom topography of a large area for one image is obtained finally.

关 键 词：侧扫声呐 SURF特征 几何变换 图像拼接 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]