摘要
为克服传统侧扫声呐图像拼接方法在低纹理或重复纹理场景下受限的问题,利用VFISNet(View-Free Image Stitching Network)无监督拼接网络进行侧扫声呐重叠区域的拼接,并与传统的地理信息拼接方法和特征检测拼接方法中的SURF(Speeded Up Robust Features)、ORB(Oriented FAST and Rotated BRIEF)算法进行对比分析,分别比较它们拼接后图像的AG、SF、STD、MI、SSIM指数,并通过与真实侧扫声呐图像进行对比验证三种方法的优劣性。实验结果表明,基于VFISNet无监督拼接网络的图像拼接方法相较于传统的拼接算法在一些指标上有不错的提升,且消除了重叠区域的伪影,对于侧扫声呐图像的拼接工作有一定的参考价值。
To overcome the limitation of traditional side scan sonar image stitching methods in low texture or repetitive texture scenes,the VFISNet unsupervised stitching network is used to stitch the overlapping areas of side scan sonar,and compared with the SURF and ORB algorithms in traditional geographic information stitching methods and feature detection stitching methods.The AG,SF,STD,MI,and SSIM indices of the stitched images are compared,and the advantages and disadvantages of the three methods are verified by real side scan sonar images.The experiments show that the unsupervised stitching network based on VFISNet is well improved in some indicators compared to traditional stitching algorithms,and eliminates the artifacts in overlapping areas.Ithas certain reference value forthe stitching work ofside scan sonarimages.
作者
邹历
袁飞翔
宋晓伟
曾毅
ZOU Li;YUAN Feixiang;SONG Xiaowei;ZENG Yi(Nanchang Surveying and Mapping Research Institute Co.,Ltd.,Nanchang 330038,China;School of Surveying and Geoinformation Engineering,East China University of Technology,Nanchang 330013,China;Shanghai Huace Navigation Technology Co.Ltd.,Shanghai 201703,China)
出处
《江西测绘》
2024年第2期39-42,共4页
JIANGXI CEHUI
关键词
侧扫声呐
图像拼接
无监督拼接网络
Side Scan Sonar
Image Mosaic
Unsupervised Stitching Network
