A new method of pipeline detection in sonar imagery using self-organizing maps

The main purpose of this paper is to detect and follow the pipeline in sonar image. This work is performed in two steps. The first is to split a transformed line image of pipeline signal into regions of uniform texture using the gray level co-occurrence matrix method (GLCM) which is widely used in texture segmentation application. The second one addresses the unsupervised learning method based on the artificial neural networks (self-organizing map or SOM) used for determining the comparative model of pipeline from the image. To increase the performance of SOM, we propose a penalty function based on data histogram visualization for detecting the position of pipeline. After a brief review of both techniques (GLCM and SOM), we present our method and some results from several experiments on the real world data set.