Classification Of Coral Reefs in Marine Environments Using Deep Encoder-Decoder Mechanism

Abstract

The aim of this paper is to investigate the efficiency of the various techniques that have been proposed to be followed when processing underwater optical image datasets. The study makes use of a variety of different texture datasets to assess the efficacy of the techniques that are proposed. The results of the study show that the speed of the technique that was recommended is noticeably higher than that of other methods. In addition, the evidence that will be presented in this paper will show that this technique can accomplish a higher rate of accurate classification than any of the other methods that were used in the past. The findings of this study will be the subject of proposals for future work: using key points instead of all points for feature extraction, using feature selection techniques for removing redundant features, proposing some structural descriptors that can be combined with statistical descriptores for better description of the textures of coral reefs, and finally applying some new preprocessing operations to extract more discriminative features.