Indian Journal of Science and Technology

Abstract

Prawn Species Recognition is very essential as Prawn culture is expanding in particular because of high price and demand. Aquatic experts use manual methods for identifying various Penaeid Prawn Species. A Gabor wavelet classification method is used for the Prawn species recognition. Using Gabor wavelet the texture features mean and variance are extracted and classification is done based on Nearest Neighbour algorithm using Euclidean distance. The species recognition based on unique texture feature extraction is quick, simple and reliable and further will speed up the discovery of many Species yet to be named. The proposed approach identifies the species of the prawns without any expert. Three different species of Penaeid prawns each of 100 image samples were collected from various parts of Indian coasts and are experimented to prove that Gabor filter is comparatively better with good recognition rate. This technique is robust against rotations to some extent. The experimental results are reported to show that by using simple statistical features like mean, variance and Euclidean Distance as a measure, 93% of recognition rate has been achieved. As an improvement it would be remarkable to extract more statistical features from prawn images that does not increase the feature vector but affects the execution time.
Keywords: Feature Extraction, Gabor, Penaeid Prawn, Recognition, Species, Statistical Features