摘要
提出了一种改进的OMP算法,在信号重构中用Jaccard系数替代传统的内积准则进行原子匹配。实验仿真证明,与内积准则和Dice系数相比,Jaccard系数在向量的相似性准则中具有最优性。将改进的OMP算法应用到侧扫声纳图像预处理上,可以在视觉上和信噪比上让低信噪比的声图获得一定改善,结合hough变换,并设计海底管线边缘轨迹的直线选择策略,可以提高海底管线的检测正确率。
An improved Orthogonal Matching Pursuit（ OMP） algorithm base on Jaccard coefficient has been proposed. In the procession of signal reconstruction,Jaccard coefficient has been used to match atom instead of inner product. It has been proved by the simulation that Jaccard coefficient is the better criterion than inner product and Dice coefficient in matching pursuit. When the improved OMP algorithm has been applied in the preprocessing of low SNR side- scan sonar image,the quality of it will be improved in visual and SNR terms. Using Hough transform and designing a line-choosing tactics,the improved OMP algorithm will improve the correct rate of detecting seabed pipeline on the side- scan sonar image.
作者
潘泽宇
许枫
刘佳
PAN Zeyu XU Feng LIU Jia(Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China University of Chinese Academy of Sciences, Beijing 100049, China)
出处
《网络新媒体技术》
2017年第2期58-64,共7页
Network New Media Technology
基金
国家自然科学基金项目(11404365)
