Pests Diseases Of Crop Detection System Based On Computer Vision

Pest diseases of crops are one of the common agricultural disasters in China.It may not only have a serious impact on the growth environment of crops in arid regions,but may also cause economic losses that are difficult to accurately measure for China’s agricultural production and economic development.In order to effectively improve the quality and efficiency of regional agricultural production and economic activities,it is necessary to detect the pest diseases of crops.Therefore,how to optimize the detection methods of pest diseases has positive social significance for agricultural production activities.Traditional recognition methods of pest diseases mainly rely on manual detection and past experience.Human subjective factors occupy a large influence,and the actual recognition effect is not accurate enough;in recent years,computer vision technology has gradually developed,which can be more objective and real-time image for online identification.Therefore,based on the analysis of existing research results and literature,this paper summarizes the key issues of computer vision technology in pest diseases recognition of crops,and proposes a new detection method: A variety of feature fusion methods based on convolutional neural network CNN and feature package model.Extracting four key features such as the color and texture of each crop leaf image,and further fused these four features as training samples for CNN to achieve real-time pest diseases Detection.In the experimental stage,this paper selected a total of 17,624 crop leaf images as the original data set,and simultaneously selected a support vector machine(SVM)model,a recurrent neural network(RNN)model,and a convolutional neural network(CNN)model under the same batch of feature conditions to perform comparative experiments.The experimental results show that the effect of multi-feature fusion algorithm based on CNN model on pest disease recognition is better than the multifeature fusion algorithm of other models and the performance of single-feature algorithms of all models.Finally,based on the above-mentioned multi-feature fusion algorithm of the CNN model,this paper designs and implements a WeChat applet for pest disease detection.Users can upload images of crop pests leaves in real time,and after a brief analysis of the model,they can provide users with identification results which can better solve the problems that pest diseases detection require higher realtime,feasibility,accuracy and other aspects.This paper proposes a pest disease detection of crops system based on CNN and multi-feature fusion.Its innovations are: through the fusion of multi-features,a more comprehensive feature is formed,and the prediction accuracy is greatly improved;the system is based on a WeChat applet,farmers can use it without threshold,in real time and conveniently;provide users with a communication platform in the field of crops,users can search or publish content about crops in small programs to promote experience exchange;the system asynchronously captures agricultural portal news Class data and no requirement in background maintenance.The application prospect of the system is also very broad.It is expected to improve the accuracy and identification efficiency of pest diseases detection,thereby improving the timeliness and accuracy of pest disease detection and reporting,and it is of great significance to the detection and control of crop pest diseases,and agricultural production.