Tea disease identification method based on deep transfer learning

Abstract

The invention discloses a tea disease identification method based on deep transfer learning, and the method comprises the following steps: a training stage: obtaining pictures of a plurality of varieties of tea leaves, and taking the pictures as a training data set; preprocessing the pictures in the training data set, wherein the preprocessing comprises random flipping, random clipping, random angle rotation, color jitter and noise addition; constructing a picture classification model; introducing SE Block into the DenseNet, introducing a channel attention mechanism through the SE Block, and constructing a feature channel weighted SE-DenseNet network model; inputting the training data set into the constructed SE-DenseNet network model, training the SE-DenseNet network model through transfer learning, and storing the trained model; and a test stage: inputting a to-be-identified tea image, performing classification identification on diseases in the tea image through the SE-DenseNet network model obtained through training to obtain disease features, and outputting a disease classification result through the classifier. According to the method, the recognition accuracy under the conditions of small samples and uneven sample distribution is improved, and the recognition accuracy and speed are higher than those of an original model.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a tea disease identification method based on deep transfer learning. Background technique [0002] Tea is an important economic crop in my country, not only has a large planting area, but also has many varieties. However, tea is easily affected by diseases in the process of growth and planting, which directly affects the quality and quality of tea, causing huge economic and social loss. The original artificial identification of diseases is mainly based on expert experience and visual observation, but this method is often limited by factors such as long cycle, strong subjectivity, poor consistency, high error rate, and difficult quantification. With the development of machine learning, image processing and machine learning methods have been widely used in crop disease identification. Traditional machine vision methods require a large number of image segmentation and feature...

AI Technical Summary

Problems solved by technology

However, DenseNet only simply merges the output feature maps when performing dense connections, and does not take into account the correlation between different feature channels, resulting in the inability to adaptively learn the correlation coefficient between feature channels.

Images