Underwater object Images Classification Based on Convolutional Neural Network

Abstract

In order to solve the problem of underwater object images classification under the condition of insufficient training data, a novel underwater object images classification method based on Convolutional Neural Network(CNN) is proposed. Firstly, an advanced method of Markov random field-Grabcut algorithm is adopted to segment images into two regions: shadow and sea-bottom. Then, considering the character of the dataset, a CNN is constructed referring to Alexnet structure, consisting of two parts with different functions: convolutional part and classification part. At last, the CNN is trained to classify three different shapes of underwater objects(cylinder, truncated cone and sphere) utilizing the transfer learning approach. The method is applied to synthetic aperture sonar(SAS) datasets for validation. Comparing with Support Vector Machine(SVM) and CNN which only use trial dataset, the proposed method can achieve a better accuracy.