Detection and recognition of Bangladeshi fishes using surf features and CNN classifier

Abstract

This thesis proposes the detection and recognition of Bangladeshi local fishes using image processing. In the proposed model, we have successfully detected fishes using grass-fire algorithm along with other methods like Gaussian Elimination, noise margin for image pre-processing at first and binary masking, flood-filling of binary image later for detection. After that the species of the detected fish is recognized using Speeded up Robust System (SURF) algorithm and for further analysis it was also done by CNN classifier to check whether the accuracy rate is better than the previous one. This is to be mentioned that we have done the whole research on various species of Fishes available on Bangladesh which has not been done before. We implemented our custom Dataset consisting of 400 sample images in the proposed method to measure out it credibility. Our aim is to detect and recognize fish more accurately with less error percentage.