Detection of Megalopa Phase Crab Larvae Using Digital Image Processing

Abstract

Blue Swimmer Crab has high economic value as an export commodity. One of the problems that occur in crab culture is cannibalism during the larval development phase. Megalopa phase crab larvae tend to eat zoea phase crab larvae due to inadequate feeding. To avoid cannibalism in the crab larvae development phase, farmers need to check the development of larvae regularly and separate the megalopa phase crab larvae from zoea phase crab larvae. Checking is done by looking directly at the rearing tank, which is time-consuming and difficult because the phase changes in the crab larvae are unpredictable. This research focuses on the detection process of the megalopa phase crab larvae using a convolutional neural network (CNN) method. This research uses 210 images, and detection results get shown an average accuracy of 87.7%.