Betta Fish Image Identification using Feature Extraction GLCM and K-Nearest Neighbour Classification

Freshwater fish known as bettas have their natural habitats in a number of Southeast Asian nations, including Thailand, Malaysia, Brunei Darussalam, Singapore, Vietnam, and Indonesia. In addition to having a distinctive appearance and personality, this fish is aggressive while defending its area. The Betta Fish is distinctive in its size, pattern, and body color. It is known that there are 73 species of Betta Fish. The main reasons for carrying out this study were due to the limitations in recognizing Betta Fish with the human eye and to build an initial model for the Betta Fish pattern recognition application. The fish data set consists of 300 with limited species: 60 Halfmoon Fancy, 60 Hell boy, 60 Red Koi Galaxy, 60 Solid Blue, and 60 Yellow Koi Galaxy. This study uses three schemes between training data and testing data: scheme 1 consist of training data 75% and testing data 25%, scheme 2 training data 80% and testing data 20%, scheme 3 training data 85% and testing data 15%. In the pre-processing stage, the scaling, segmentation and gray scale methods are carried out. Characteristics of fish features were obtained through the gray level co-occurence matrix (GLCM) method with angular direction and the K-NN method was used for classification with values of k = 1, 2, 3, 4, 5, 6, 7. The results of this study show in Scheme 1 the highest accuracy is 92% with k = 5 and angular direction is 135 and error rate = 8%, scheme 2 has the highest accuracy 95% with k = 7 and angular direction is 135 and error rate = 5%, scheme 3 has the highest accuracy 100% with k = 2 and the angular direction is 180 and error rate is 0%.