Detection of shrimp feed with computer vision

Abstract

In smart shrimp farming system development, while many works have been focusing on developing an effective water quality monitoring system, little attention has been paid on an automated feeding system. Ideally, an efficient feeding system should not only be able to feed automatically, but also should be able to determine and adjust the suitable amount of food at each feed. This is to save the cost from overfeeding and labor usage, as well as to achieve high shrimp growth rate. This paper proposed a simple and low-cost shrimp food pellet detection algorithm that utilized the technique of 2D-histogram and color space analysis to detect the amount of unconsumed feed left on the feeding tray. The result provided useful information on how to adjust the amount of food in the next feed. The algorithm was developed using color segmentation on three different color spaces: HSL, LAB, and YCrCb. Experimental results confirmed that the proposed algorithm can effectively determine the amount of food pellets on the feeding tray.