Underwater Image Classification using Machine Learning Technique

For the past few years, underwater exploration has increased exponentially. Currently available instruments for data collections (Side Scan Sonar, Multi Beam echo sounder, sub bottom profiler and Remotely Operated Vehicle) in underwater research and observation not only provide the data on objects and species, but also provide data about the sea surface. In this regard, selecting suitable features is a huge task. Due to limited datasets in Underwater, it is difficult to classify the objects/features from underwater images. In order to overcome this, machine learning based Bag of Features model is adopted in this paper. The dataset is obtained from shallow water using ROV. Since the underwater optical images have low light intensity, making the classification of features a difficult task; SURF (Speeded-Up Robust Features) and SVM (Support Vector Machines) algorithms are implemented in Bag of Features model to attain maximum accuracy. The performance evaluation of training and testing datasets gives better performances.