RECOGNITION OF CROP DISEASE AND INSECT PESTS BASED ON DEEP LEARNING IN HARSH ENVIRONMENT

Abstract

Insect and disease pests in agriculture are among the most important factors that may drastically 
reduce crop yields. Identifying and detecting pests early on may eliminate economic losses 
caused by them. In this study, we provide a method for automatically diagnosing crop illnesses 
using convolutional neural networks. The dataset incorporates 10 different crops and contains 
twenty-seven images of illnesses. It was retrieved from the public data set of the 2018 AI 
Challenger Competition. This study trains the Inception-ResNet-v2 model. The cross-layer direct 
edge and the multi-layer convolution are two parts of the model's residual network. To activate 
it, the ReLu function is used after the combined convolution operation is completed. The results 
of the testing show that this model is helpful, with a total recognition accuracy of 86.1%. 
Following the training of this model, we developed a Wechat applet with the ability to detect 
agricultural diseases and insect pests. As a further step, we gave the test. The results prove that 
the system can accurately identify crop diseases and give suitable suggestions.