A small sample plant disease identification method and system

Abstract

The invention discloses a small-sample plant disease identification method and system. The method includes: randomly selecting a plurality of original pictures containing diseases as the first sample set; expanding the first sample set through an improved deep convolution-based generative adversarial network to obtain a second sample set; performing a second sample set on the second sample set Verifying and constructing a training set together with all or part of the second sample set that has passed the verification and the original picture that does not contain the disease; training the convolutional neural network based on the training set to obtain a classification model; inputting the image of the disease to be identified into the classification model to obtain the disease Recognition results. In the present invention, aiming at the disease pictures of small samples of plants, the improved deep convolution generation confrontation network is used to expand the sample set, so that the positive and negative ratio of the expanded second sample set is roughly 1:1, so that the data is balanced, and the two The number of patients has reached 10,000, and the convolutional neural network is used to classify the diseases on the expanded second sample set, which has a good classification effect.

Description

technical field [0001] The invention relates to a plant disease identification method, in particular to a small-sample plant disease identification method and system. Background technique [0002] In the detection of many quarantine high-risk plant diseases, manual inspection is currently used. For example, citrus canker disease in citrus crops, due to different growth environments, the shape of the lesion itself is changeable; the light and shooting angles when collecting images Many reasons, such as shooting skills, will affect the image quality, so it is very difficult to obtain high-quality lesion images. For quarantine diseases such as citrus canker and citrus huanglongbing, in order to prevent the spread of the disease, once found, corresponding measures (burning, burying) must be taken in time for treatment. [0003] The development of machine learning supported by image recognition is very rapid. The more classic machine learning methods include: Naive Bayesian Clas...

AI Technical Summary

Problems solved by technology

However, in the traditional expansion method, there are often problems such as poor quality of generated samples, insufficient diversity of generated samples, over-fitting problems, and sample imbalance.

Images