Crop disease identification method and system based on DCGAN and RDN

Abstract

The invention relates to a crop disease identification method and system based on DCGAN and RDN, and belongs to the field of agricultural informatization and plant protection, the method comprises the following steps: firstly, carrying out data collection, including a data set disclosed by a network and manually collected data, then, guaranteeing the accuracy and distribution balance of a training data set through the technologies of data visualization, data cleaning, DCGAN data generation and the like; dividing the processed data according to the proportion of 60% of a training set, 20% of a verification set and 20% of a test set; according to the method, a deep residual network (RDN) recognition model is constructed, and after model training parameters are set, a training set and a verification set are loaded for model training; and finally, the trained model is applied to a crop leaf disease identification system for crop disease prediction, and crop disease categories and probabilities are returned by the system. According to the method, various diseases of various crops can be recognized, and particularly, the recognition accuracy under the condition of uneven sample distribution can be improved.

Description

technical field [0001] The invention relates to the fields of agricultural informatization and plant protection, in particular to a crop disease identification method and system based on the combination of DCGAN and RDN. In particular, it relates to a recognition method and system that utilizes DCGAN technology for data enhancement and combines RDN algorithm to improve detection accuracy under the condition of uneven distribution of large-scale crop leaf training data. Background technique [0002] 1. Professional terminology: [0003] (1) DCGAN. DCGAN is the abbreviation of Deep Convolutional Generative Adversarial Networks (Deep Convolutional Generative Adversarial Networks). The present invention mainly uses DCGAN technology for data enhancement, which is used to expand training data, so as to enhance the generalization ability of the deep learning network and increase the accuracy of recognition. [0004] (2) Deep Residual Network. A deep learning model, please refer ...

AI Technical Summary

Problems solved by technology

[0012] (1) At present, most of the published results are experimental data on small data sets, mainly focusing on one crop for research, and some results only divide the data into healthy leaves and diseased leaves for classification and identification. In this case, higher accuracy may be obtained, but it lacks universality, and the practical application value is not high

[0013] (2) Data distribution is not taken into account. When the amount of data is large and there are many types of diseases, the data volume of common diseases may be very large, while some uncommon diseases may have a large amount of data due to the difficulty of data collection and the small amount of data, etc. The reason is that the distribution of training data is unbalanced, which will affect the use effect in this case

[0014] (3) At present, most of the methods are mostly to reuse the existing models, and the model innovation is low

[0023] (4) Aiming at the low degree of innovation of the training model and the low recognition accuracy under the classification of multiple crops, the present invention proposes a crop disease recognition model based on a deep residual dense connection network (Residual Dense Connection Based Network) , further improving the recognition accuracy

Images