Target detection of banana string and fruit stalk based on YOLOv3 deep learning network

摘要

Manual picking of banana bunch is a dangerous, inefficient and labor-intensive aerial work. To solve this problem, in order to realize the intelligent operation of picking, it is a prerequisite to solve the identification and positioning of banana and handle. Combined with the real-time and high efficiency of yolov3 deep learning network in target detection, this paper uses the framework of keras + tensorflow to build Yolov3 model. The average accuracy of banana stalk and banana string target detection is 88.45% and 97.96% respectively.