The tomato sorting process still heavily relies on manual quality assessment, which is inefficient and costly. The diversity in size, color, and defect morphology makes it difficult for traditional algorithm-based inspection equipment to meet the requirements for accurate sorting. A solution is needed to automatically and accurately identify and quantify various appearance defects (such as mottling, cracks, and dark spots) and distinguish between similar features (e.g., green fruit vs. mottling) to achieve automated and precise grading.