The Classification Of Stored Grain Pests Based On Deep Learning

During food storage,the food loss caused by pests is very serious.Therefore,the control of stored grain pests is one of the key technologies for food security.The identification of stored grain pests is a key issue in the prevention and control of stored grain pests.At present,the methods for identifying pests in stored grains at home and abroad include manual identification,conductivity coefficient testing,image recognition,and near-infrared spectroscopy.Because image recognition method is easy to operate,has a high recognition rate,can save manpower,material resources and have other good qualifications,this method has become the main technology for the identification of stored grain pests.In recent years,due to the development of deep learning technology,image recognition technology based on deep learning has become a hot topic at home and abroad.Based on deep learning and image recognition technology,this paper studies the method of deep learning based pest recognition and classification of stored grain.The main research contents and work of this article are as follows.(1)Aiming at the characteristics of pest identification and classification in stored grain,a method of deep learning based pest recognition and classification for stored grain is proposed.And the image preprocessing method of stored grain pests is given.With ReLU as an activator and Softmax as a classifier,a neural network model of stored grain pests recognition based on an 8-layer convolutional neural network model is designed and a corresponding classifier is constructed.The experimental results show that the highest recognition rate of the stored grain pest neural network model reaches 98.3%.Compared with the traditional method,the accuracy rate increased by 3% to 9%.(2).To eliminate the limitations of the ReLU activation function and the Softmax classifier,a neural network model based on the ELU activation function and the dropout classifier is introduced.Based on the original 8-layer convolutional neural network model,four layers were added.The problem about ReLU's robustness to noise disturbances has been overcome;through the introduction of the dropout classifier,the overfitting of contaminated convolutional neural networks in stored grain pests is reduced,and the accuracy of recognition is improved to 98.8%.