Image Based Individual Identification of Sumatra Barb (Puntigrus Tetrazona)

Abstract

The paper deal with the individual fish identification of the same species based on digital image of the fish. The proof of concept of image based individual identification is introduced on the small group fish. The method is completely noninvasive and can overcome the disadvantages of standard invasive identification such as tagging. The experiments proved the hypothesis that the visible patterns on Sumatra Barb (Puntigrus tetrazona) body can be used for individual identification. In the first step, the database of 43 fish (was created by the taking of the images of fish in different pose. Images were taken in an aquarium with a water. After data collection, data was processed by the image processing methods to determine the features. The simple nearest neighbor classification was used to test individual identification. The accuracy of classification was 100%. The method proved the hypothesis that the visible pattern on Sumatra Barb can be used for fully automated individual fish identification. It can be substituted current practice of fish identification based on tagging and marking. The long-term stability of the pattern and the classification power for large fish group should be studied in the future.