K-Means Clustering Algorithm for Image Segmentation and Classification Based On ANN for Underwater Applications

The objects in underwater is difficult to classify clearly. During acquisition the objects and organisms present in the underwater is suffered from a large amount of noise due to low contrast and scattering of light present in the environment. The noise in that image is filtered by a median filtering method. Then the filtered image is segmented by k-means clustering algorithm. Feature is extracted before the classification method. For classification Artificial Neural Network is used. The application of Artificial Neural Networks is found to have improved performance than other supervised algorithms. In this, the prototype of a system is for classifying underwater images into two broad categories such as natural shapes and unnatural shapes. Distinctive back propagation strategies and a variable number of concealed layers have been attempted with the model neural system framework for guaranteeing the robustness of the system.