Automated fish fry counting and schooling behavior analysis using computer vision

Abstract

This paper presents an automated fish fry counting by detecting the pixel area occupied by each fish silhouette using image processing. A photo of the fish fry in a specially designed container undergoes binarization and edge detection. For every image frame, the total fish count is the sum of the area inside every contour. Then the average number of fishes for every frame is summed up. Experimental data shows that the accuracy rate of the method reaches above 95 percent for a school of 200, 400, 500, and 700 fish fry. To minimize errors due to crowding in the container, schooling behavior analysis is considered. The behavioral effects of different colored lights on milkfish and tilapia are thoroughly investigated. The system's effectiveness, efficiency, possible improvements, and other potential applications are discussed.