A Hierarchical Approach To Seafloor Classification Using Neural Networks
暂无分享,去创建一个
Preliminary progress toward the automated segmen- tation and classification of undersea terrain is demonstrated with sidescan-sonar imagery from a mid-ocean spreading center. A multilayer, backpropagation neural network was applied to data from three distinct geoacoustic provinces: axial valley, ridge flank, and sediment pond. A suite of experiments compared the effectiveness of different feature vectors comprising :spectral esti- mates, gray-level run length, and gray-level differences with win- dow sizes and training patterns considered. Run-length features showed the best individual performance across different data sets, while hybrid features appear most promising.
[1] Azriel Rosenfeld,et al. A Comparative Study of Texture Measures for Terrain Classification , 1975, IEEE Transactions on Systems, Man, and Cybernetics.
[2] Belur V. Dasarathy,et al. Image characterizations based on joint gray level-run length distributions , 1991, Pattern Recognit. Lett..
[3] M. E. Jernigan,et al. Texture Analysis and Discrimination in Additive Noise , 1990, Comput. Vis. Graph. Image Process..