Citrus pest recognition method based on deep convolutional network and transfer learning

Abstract

The invention discloses a citrus insect pest recognition method based on a deep convolutional network and transfer learning, and the method comprises the following steps: step 1, collecting citrus insect pest image data information, and making a corresponding training data set; step 2, performing data enhancement and expansion on the acquired image data set; step 3, constructing an improved InceptionV3 model based on a deep convolutional network and transfer learning; and step 4, carrying out model weight parameter loading initialization. Compared with an existing identification method, the citrus insect pest identification method based on the deep convolutional network and the transfer learning has the advantages that a large amount of manpower and material resources are saved, and the whole collection process is simple and easy to operate; a multi-scale feature fusion method is added to a training model based on an InceptionV3 model and transfer learning, semantic information contained in image features is further improved, and the problem that the recognition rate is low due to the fact that insect colors and textures specific to citrus insect pests are complex is effectively solved.

Description

technical field [0001] The invention relates to the related technical field of insect species identification in agricultural planting, in particular to a citrus pest identification method based on deep convolutional network and transfer learning. Background technique [0002] The prediction and control of diseases and insect pests is an important content in the development of modern agriculture. As one of the important origins of citrus in the world, the prediction and control of citrus diseases and insect pests is an important agricultural research topic; most of the identification of crop diseases in my country is realized by manual methods. The limitations of the system restrict the large-scale pest control of crops; the efficiency of traditional identification is too low and the cycle is too long, which is not suitable for the identification of agricultural pests with relatively high requirements for intelligence; citrus, as the main agricultural product grown in Sichuan P...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a citrus pest identification method based on deep convolutional network and transfer learning to solve the problems of waste of manpower, complicated collection process and low recognition rate for most citrus pest identification proposed in the background technology

Images