Design and Implementation of Sprouting Potato Eye Recognition Using YOLOv8

The distribution of potatoes' surface sprouts greatly impacts the seed potato growth after sowing, and the rapid and accurate recognition of potato eyes can promote the automotive cutting of seed potatoes and benefit the sowing quality. However, the small potato eye area, few extracted features, and complex background of the potato surface lead to low accuracy in sprouting potato eye detection. The present paper aims to strengthen the sprouting potato eye detection by learning image features and enhancing feature extraction with a trained recognition model using the latest YOLOv8 network. The potato images in the present study are taken in an actual field and under complex circumstances, where the soil, clods, and ridges are noisy and interfere. The results show that the final precision rate is 98.9%, the recall rate is 95%, the test speed is 42 FPS, and the total recognition accuracy is 90.1%. The proposed model is effective and feasible for detecting sprouting potato eyes in an actual scenario, providing technical support for intelligent potato seed cutting.