Mushroom-YOLO: A deep learning algorithm for mushroom growth recognition based on improved YOLOv5 in agriculture 4.0
Yifan Wang, Lin Yang, Hong Chen, Aamir Hussain, Congcong Ma, Malek Algabri
Abstract
In agriculture 4.0, internet of things is pushing the boundary of smart agricultural applications to assist farmers from production to sale of crops. Mushroom is one of the most economically valuable crops in agriculture production, and widely cultivated all over the world, from China to the United States. Growing shiitake mushrooms requires real-time adjustment of the indoor environment, and statistics on the yield and types of shiitake mushrooms. The traditional planting method is labor-intensive and inefficient. Moreover, the traditional image processing methods have strict requirements on crop background, which also increases the cost of planting. To address this issue, in this paper, a deep learning algorithm for mushroom growth recognition based on improved YOLOv5 is proposed and named Mushroom-YOLO for small targets detection such as mushrooms, and the mean average precision is up to 99.24% and this performance is much better than the original YOLOv5. In addition, a prototype system for the flower shiitake mushroom yield recognition used iMushroom is presented. The prototype and real shiitake mushroom planting case study show the effectiveness, and provide a potential way to control the quality of shiitake mushroom growth without human in indoor farming.