Research on classification and recognition method of plant leaves based on deep learning

Abstract

With the development of global science technology and economy, the coordination of social relations and ecological relations has become an important problem that needs to be solved urgently in contemporary development. Tree resources are the foundation of the ecosystem. In order to make better use of tree resources and implement strategies tailored to local conditions, it is necessary to identify the types of trees. Deep learning is an important method to distinguish tree species by analyzing leaves. Deep learning provides incomparable advantages of traditional object detection methods for accurately extracting the deep features and classification of leaf images. In order to achieve accurate and rapid classification and identification of plant leaves, this paper uses two network models, YOLOv5 network and image recognition ResNet-50 based on faster regions with convolutional (Faster-RCNN), to make a comparative study on the constructed 10 types of common plant leaf datasets. The experimental results show that yolov5 network can quickly and accurately identify plant leaves, and its lightweight model can be easily deployed on mobile terminals. Faster-RCNN network model can accurately extract the multi-layer feature images of plant leaves to obtain a average accuracy that is 1.1% higher than that of the YOLOv5, but the recognition speed is slower. It can be used for fine classification and leaf identification in subsequent plant experiments.