Coastal Indian Prawn Species Recognition System Through Molecular Identification and Neural Networks

Abstract

Species delineation is one of the strenuous tasks to be done. In Crustaceans species, which are lookalike in appearance, is very intricate and need much effort for the right species identification. Prawns & Shrimps are being consumed widely and have more health benefits. According to Holthius (1980), prawns include 33 genera with about 2,500 species globally in which only a few hundred have their commercial importance. In such case it is very important to differentiate the trade from important prawns to noncommercial ones. For parting such species, a special technique is being applied named DNA barcoding, which is a fresh, fast and accurate technique for the species delineation. DNA barcoding concept uses nucleotide sequences of precise mitochondrial marker (Cytochrome Oxidase I gene) to generate the DNA barcode. The nucleotide sequences are compared in parallel to the sequences in the BOLD (Barcode of Life Data system) reference library for appropriate recognition of the species. This is further followed by phylogenetic analysis to know their genetic distance. This study addresses the difficulties in species discrimination using morphology-based taxonomy through molecular identification. These barcodes are preserved in a large-scale system. In order to identify the species, it is essential to measure and recognize the features in detail. Feature extraction of barcode is an important problem to identify the prawn species through barcodes. On the computational part, the focus is on the feature extraction of prawn species using image processing. Image pre-processing is the main step where filtering, interpolation and Sobel edge detection are followed. From the pre-processed image, the segmentation process is employed. The results of the analyses are the time, accuracy and evaluation measures. The provisional results are evaluated by the segmentation adaptability for pre-processing and width magnitude for feature extraction under each family thereby, facilitating automatic identification of prawn species.