Postharvest Strawberry Bruise Detection Using Deep Learning

Abstract. Strawberries are popular fruit worldwide for their special aroma, sweet and acid taste, and rich in vitamins and minerals. However, strawberries are delicate and easy to be bruised. Strawberries can be bruised by compression, impact, or vibration forces during harvesting and transportation. These bruises make strawberries more vulnerable to rot or disease, thus shortening their shelf life. Therefore, detecting strawberry bruises is important for farmers to guarantee the quality of strawberries that are sent to the market. A color camera was used to quickly collect strawberry bruise images under incandescent light. A state-of-the-art deep learning method, Mask R-CNN, was applied for strawberry bruise detection. The result shows that Mask R-CNN can correctly detect whether a strawberry is bruised with an F1 score of 0.99. A severe bruise can also be detected by Mask R-CNN with an F1 score of 0.90.