Identifikasi Citra Karang Menggunakan Jaringan Syaraf Tiruan: Kasus Family Pocilloporidae
Abstract
Nowadays, the trend method for coral identification is by using the visual technique since it does not have to take the coral sample which can damage the coral growth. This technique needs visual skill and the ability to classify the patterns which are formed by the color, the texture and the shape of coral. So far, this visual technique can only be done by the expert and experienced researchers. Therefore, in this research, the implementation of backpropagation artificial neural network is done to identify the genus image of corals as the application of visual technique in computer system. Hopefully, this application can help and make it easier for the early-stage researchers to identify corals from the features of photography results. The genus which belong to the family of pocilloporidae. The feature extractions; colors,textures and shapes are used as the input to the network. The methods are RGB and HSV for color, moment invariant for shape and three methods which are statistical moment, gray-level co-occurrence matrix (GLCM) and local binary patterns (LBP). The experiments towards the combination of color and shape features on each texture feature, show that the identification level of color, texture (statistical moment) and shape is better than other feature combinations. The result of the recognition rate is 96% and the elapsed time ìs 10.94 second. The recognition rate of color, texture (GLCM) and shape is 96% and the elapsed time is 20.849 second. While the recognition rate on color, texture (LBP) and shape is lower, that is 68% and the elapsed time is 0.465 second.