摘要
利用SURF(Speeded-up Robust Features)算法对多波束和侧扫声呐图像配准时,因为图像分辨率差异大而导致配准困难,通过对低分辨率的图像进行升采样,使图像配准达到了较好的效果;另外,对SURF算法中粗匹配的距离测度函数进行改进,提高了SURF算法的配准速度;然后利用RANSAC算法实现了多波束与侧扫声纳图像的精准配准;最后对配准后的图像进行小波变换融合,利用信息熵和平均梯度对图像融合效果进行了评价,并通过实例数据验证了该算法的有效性。
The problem in the image registration with high-low resolutions is solved by preprocessing the multi-beam echo sounder(MBES) image and side scan sonar(SSS) image. Through replacing the similarity measurement in Speeded-Up Robust Features algorithm, the algorithm speed of SURF is improved. Then the image is registered precisely by Random Sample Consensus algorithm. Finally, MBES image and SSS image which are registered are fused according to the wavelet transform method. By comparing and analyzing information entropy and average gradient of fused images and original images,the averages of fused images are verified. These methods used in this paper are proved by experiments.
出处
《海洋通报》
CAS
CSCD
北大核心
2016年第1期38-45,共8页
Marine Science Bulletin
基金
中央高校基本科研业务费项目(GY02-2011T05)
山东省自然科学基金(ZR2009FM005)
国家国际科技合作专项(2014DFA21710)
国土资源部海洋油气资源与环境地质重点实验室项目(MRE201107)
国家自然科学基金(49706038)
关键词
多波束图像
侧扫声纳图像
高低分辨率匹配
融合
SURF算法
RANSAC算法
曼哈顿距离
multi-beam echo sounder image
side scan sonar image
high-low resolution image registration
fusion
Speeded-up Robust Features algorithm
Random Sample Consensus algorithm
Manhattan distance
作者简介
侯雪(1989-),女,硕士研究生,主要从事海洋测绘与多波束应用研究,电子邮箱:houxujiayou@126.com。