摘要
为解决侧扫声纳条带图像拼接时共视目标的扭曲和错位问题,提出了基于地物不变性的特征点匹配方法。该方法对侧扫声纳图像重叠区域进行符合地物不变性的特征点提取、K-均值分区和薄板样条弹性变换等处理,实现匹配点对的精确配准。采用某海域实测侧扫声纳数据对算法进行了验证,结果表明,特征点匹配效率和准确率高于传统的RANSAC算法,配准精度达到0像素偏差,拼接后的图像能够准确反映海底地貌特征,对大面积、高质量的海底地形地貌图像获取有重要意义。
In order to solve the problem of distortion and misalignment of common view objects when stitching side scan sonar band images,a feature point matching method based on the invariance of ground objects was proposed.In this method,feature points extraction which are consistent with the invariance of ground objects,K-mean partition and thin plate spline elastic transformation are carried out to achieve accurate registration of matching point pairs in the overlapping regions of side scan sonar images.The algorithm is verified by using the measured side scan sonar data in a sea area,the results show that the matching efficiency and accuracy of feature points are higher than that of traditional RANSAC algorithm,and the registration accuracy can reach O pixel error,and the mosaic image can accurately reflect the features of the seafloor geomorphic features,which is of great significance for obtaining large area and high quality images of submarine topography and landform.
作者
李雪申
吴永亭
胡俊
豆虎林
李治远
LI Xueshen;WU Yongting;HU Jun;DOU Hulin;LI Zhiyuan(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China;First Institute of Oceanography,MNR,Qingdao 266061,China)
出处
《海洋测绘》
CSCD
北大核心
2023年第2期6-10,共5页
Hydrographic Surveying and Charting
基金
国家自然科学基金(42176186)。
关键词
侧扫声纳
地物不变性
图像匹配
K-均值分区
弹性变换
side scan sonar
ground object invariance
image matching
K-mean partition
elastic transformation
