Estimation of milkfish physical weighting as fishery industry support system using image processing technology

The Milkfish prices in many marketplaces influenced by physical weight. The weighting is determined using an analog scale and must be delivered in real time for good marketing for the fishery industry, restaurant and to the final consumer. This paper presents a computer vision system with image processing technology that allows Milkfish scales data to be directly stored in computer databases. The system designed in the form of measuring instruments of physical weight and directly presented on the user's computer. The computer vision approach implemented by determined the measurement pattern to find the parameters (HSVHS) and using contour detection to obtaining object edges as a bounding box. Each bounding box edges measured using pixels scale and formulation using Weight Fish Estimation equation. Accuracy performance of image processing was measured using ROC approach. The results showed the accuracy of weight estimation using 20 fish samples reached by 86.66% and the accuracy of object detection by 95% with error rate measurement by 1 centimeter (cm) from the actual length. This result highly recommended to applied to the fishery field and it can be a tool for fishpond farmers or fishermen to detect fish weight in fish marketing.