In view of the visual detection ability required by the picking robot in practical application,a mobile citrus detection algorithm based on deep learning was proposed to realize the rapid real-time detection of citrus fruits.Models such as YOLOv5s,YOLOv5s6,and YOLOv5m were trained on self-made citrus datasets through the PyTorch framework,and the trained models were transformed into ONNX models,and the ONNX models were converted into NCNN models,and deployed on Android mobile terminals.The experimental results show that the recognition rate of YOLOv5s6 under the NCNN deep learning framework reaches 92.9％and the recall rate reaches 75.7％,which can better meet the practical application needs of picking robots.
维普
万方数据