Plant disease and pest identification method based on attention mechanism and multi-level convolution characteristics

Abstract

The invention discloses a plant disease and pest identification method based on an attention mechanism and the multi-level convolution characteristics. The method aims to classify the plant pest and disease damage images in a natural scene by combining the attention mechanism and the multi-level convolutional characteristics and mainly solve the problem that the plant pest and disease damage identification is interfered by the large-area complex backgrounds in the natural scene images, and all convolutional layer characteristics in a network model are fully utilized. According to the method, adeep convolutional neural network containing the attention mechanism is designed, a local position area mask covering a target spatial position in an image is generated by utilizing the attention mechanism and is used for weakening the complex background interference information in a natural scene image, and the semantic information and the detail information are fully utilized in combination with the multi-level convolution characteristics to perform more differentiated characteristic representation on the plant diseases and insect pests. Finally, the obtained model is used for the plant disease and pest image classification in a natural scene, and the obtained local area mask is used for obtaining the position information of the plant diseases and pests.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a method for recognizing images of plant diseases and insect pests in natural scenes, and more specifically, to a method for identifying plant diseases and insect pests based on an attention mechanism and multi-level convolution features. Background technique [0002] Plant pests and diseases are one of the leading causes of harm to commercial agricultural products. Plant pest identification plays a vital role in agricultural plant pest prediction, food security, and agricultural economic stability. Due to the wide variety of plant pests and diseases with subtle differences between species, the identification of plant pests and diseases largely relies on the expertise of agricultural experts, which means that it is costly, time-consuming and labor-intensive. With the development of machine learning and computer vision technology, the automatic identification of plant dise...

AI Technical Summary

Problems solved by technology

However, most of the existing pest images in the datasets verified by these methods are collected in a controlled laboratory environment, while in practical applications, most plant diseases and insect pests appear in natural scenes, usually with complex background information, which has great impact on Whether the identification method of plant diseases and insect pests can reduce the ability of complex background information interference in natural scene images puts forward requirements, and it is also a major difficulty in applying deep learning technology to identification tasks of plant diseases and insect pests

Images