Research on Plant Image Identification Based on Deep Learning

In recent years, increasing attention was paid to the methods from researchers that about the intelligent identification of plants and their diseases based on deep learning algorithms. In this paper, plant images were as the study object. Firstly we listed the Research results on traditional methods and deep learning methods of machine learning, and summarized the classification features of plant images and the general procedure of plant identification, Simultaneously we introduced the general algorithms for deep learning, and studied the structural features of convolutional neural networks, and described the classical model of convolutional neural networks, At the end, we compared experimentally the identification efficiency of VGG16+SVM classifier and VGG16+Softmax classifier on plant images. Experiments have shown that under the same conditions, the SVM classifier has a higher identification rate for plant images with single backgrounds, but the identification rate for plant images with complicate backgrounds is close to that of the softmax classifier, and the VGG16 algorithm needs improvement further in the identification rate on fine-grained plant images with too similar leaf shapes. This also proved that the identification and classification of plant images with complicated background and fine-grained is a major constraint in achieving intelligent identification on plant.