Counting and Weight Estimation of Penaeus Vannamei Through Image Processing

Shrimp farming stands as a prominent and rapidly expanding sector of aquaculture in the Philippines, yet its sustainability is hindered by substantial feed costs, which represent a significant portion of total expenses for both small-holder and commercial farms. This study aims to address this challenge by developing an algorithm that accurately counts and estimates the weight of Penaeus Vannamei shrimps, thereby optimizing feeding patterns and fostering healthier growth. The research implements a measurement station with fixed lighting and camera height to generate a comprehensive dataset of shrimp images and corresponding weights. Utilizing YOLOv8 for detection and counting, coupled with supervised machine learning incorporating Area-based, Perimeter-based, Color-based, and GLCM-based estimation methods across multiple image batches. The study identifies Exponential GPR as the most effective regression model for the dataset. Furthermore, it reveals that Area-based and GLCM-based estimations exhibit the highest accuracy, with respective MSE, MAE, and R^2 values of 1.6814, 0.9005, 0.96 and 1.0638, 0.62359, 0.68. These findings offer valuable insights into optimizing efficiency and cost reduction strategies within shrimp farming operations, based on accurate counting and weight estimation algorithms.