Development of an Automatic Wood Ear Mushroom Harvesting System Based on the Collaboration of Robotics, VR, and AI Technologies

Abstract

To realize "smart agriculture," which is a countermeasure to the recent shortage of agricultural labor, this study aimed to develop an automatic harvesting system for wood ear mushrooms. The automatic harvesting of wood ear mushrooms, an agrarian crop with complex shapes and few characteristics, such as color, is difficult. However, our system realizes the automatic harvesting of wood ear mushrooms with a robotic arm by remotely selecting the harvesting target using a virtual reality application. A deep learning-based wood ear mushroom recognition model is used to detect wood ear mushrooms to be harvested. We developed a model with a high recognition accuracy-a detection accuracy of 90% and growth-level differentiation accuracy of 80%-by applying data augmentation methods and developed a system that can harvest wood ear mushrooms automatically.