Apple Maturity and Defect Detection Using Improved YOLOv5

Unmanned vehicles with arms are currently widely employed for apple picking in orchards. The primary operation is the detection of apple maturity and surface defect. An improved Y 010v5 is proposed to solve the problems of low detection rate and slow model reasoning speed. Firstly, a lighter MobileNetv3_Block network is used to replace the original network backbone. Secondly, the knowledge distillation is introduced to improve algorithm precision, where the yolov5x network acts as a teacher model. Numerical experiments are carried out to validate the improved YOLOV5. Comparisons show that the proposed method improves the detection accuracy and reduces the calculation.