Development of a Raw Oyster Volume Estimation System Based on U²-Net

This study proposes an advanced image processing algorithm and a volume estimation system for automating the oyster process. Using top and side cameras, our algorithm effectively captures object features under varying moisture and lighting conditions, mitigating light reflection issues. The image processing algorithm consists of object extraction, capturing, feature extraction, and width calculation algorithms. These components accurately extract objects from multiple camera inputs, store the images, and calculate object width using the side camera, providing essential data for volume estimation based on the disc method. Experiments conducted on model and raw oysters demonstrated the system’s robust performance, achieving classification accuracies of 98.5% and 100%, respectively. The system’s high accuracy and reliability across various environmental conditions make it suitable for size classification tasks.