Research On The Construction And Application Of Grape Disease And Insect Pest Identification Model Based On Deep Learning

Grape has important nutritional value,medicinal value and economic value.China is the largest producer and consumer of table grape in the world,but grape will be infected with various diseases and insect pests in the process of growth.Efficient identification of grape diseases and insect pests is one of the keys to the prevention and control of grape diseases and insect pests.In this paper,a grape pest recognition algorithm based on improved YOLO V3 is proposed to study the grape pest image recognition.The main work is as follows:(1)The construction of grape diseases and insect pests data set.Relying on the State key Laboratory for Biology of Plant Diseases and Insect Pests,Chinese Academy of Agricultural Sciences,79 categories and 8442 original images of grape diseases and insect pests were screened and screened,and then a total of 58874 data sets were constructed by marking and data preprocessing.(2)Based on the improved YOLO V3 grape pest identification algorithm(Eff-B3-YOLO V3).Aiming at the problems of large number of model parameters and large number of Darknet-53 convolution layer in backbone network of YOLO V3,combined with the characteristics of grape pest image recognition and inspired by "barrel theory",an improved grape pest recognition algorithm(Eff-B3-YOLO V3)was proposed.The algorithm replaces the Darknet-53 backbone network with the Efficient Net network to effectively balance the image resolution and the depth and width of the training network,and improves the prediction network of YOLO V3 by canceling the output of the 26 × 26 feature layer and adding a residual edge,so that the recognition accuracy of the trained model is significantly improved and the number of parameters of the model is reduced.The feasibility and superiority of the algorithm proposed in this paper are proved by comparing with YOLO V3,Faster R-CNN and Retina Net.(3)Based on the improved YOLO V3 intelligent identification system of grape diseases and insect pests.According to the method of software engineering,an intelligent recognition of grape diseases and insect pests We Chat Mini Programs based on improved YOLO V3 is developed.The software can recognize 79 kinds of grape diseases and insect pests online through the intelligent recognition model of grape diseases and insect pests on the cloud server,with a recognition rate of 97.29%.The recognition results,characteristics of diseases and insect pests,causes of disease,geographical distribution of diseases and insect pests and suggestions for prevention and control will be fed back to the user.At the same time,the grape pest data set can be dynamically updated and optimized.The research of this paper shows that in the process of identifying grape diseases and insect pests,the model not only has a high recognition rate,but also covers most common grape diseases and insect pests at present,which fills the blank of domestic image recognition of many kinds of grape diseases and insect pests under natural conditions.It can not only assist fruit farmers to quickly identify grape diseases and insect pests and prescribe the right medicine,but also help relevant researchers to further study grape diseases and insect pests.To achieve the purpose of science and technology to help farmers has an important guiding value for the prevention and control of grape diseases and insect pests.