Research On Corn Disease Recognition Based On Deep Learning And Growth Monitoring Iot System

With the application of artificial intelligence in agriculture,agricultural intelligence and modernization develop faster,and the realization of precision agriculture is one of its core parts.At this stage,maize planting in my country is facing the erosion of various diseases.There are many types of diseases and intersecting diseases.Symptoms mainly appear on maize leaves.At the same time,due to the lack of effective management of field planting and the inability to obtain timely and accurate information on farmland,production management and yield of maize are affected.At present,there is no public and high-quality data set in the research on maize disease identification through machine learning,and maize disease identification based on deep learning first requires a powerful,professional and efficient data set,secondly,deep learning requires high equipment and complex model structure,etc.The problem is that the self-defined network realizes the low accuracy of small sample data classification.In order to accurately judge the growth status of corn,the Internet of Things technology is used to remotely monitor the growth status and diseases of corn,so as to liberate labor,save resources and optimize management.Therefore,this paper Combining deep learning and IOT technology,a comprehensive IOT intelligent monitoring system for corn disease identification and growth was designed.The main research contents of this subject are as follows:(1)It establishes a corn common disease image dataset(Corn Disease Data Set,CCDS).Obtain corn disease images through website download,web crawler and other technologies,select corn disease images in the Plant Village image dataset,and self-harvested corn disease images,and process the images after manual screening and review to establish a professional corn disease dataset,including a dataset of four categories: corn health,maize northern/southern leaf blight,and rust.In addition to being used for own research,this standard dataset can be publicly released to promote research in the field of corn disease identification.(2)Using the transfer learning method to solve the classification problem of the self-built small sample corn disease data set.The first is to introduce the idea of transfer learning based on the problem of low recognition rate of custom models on small sample data sets,and the second is to solve the problem of low recognition rate of the migrated model and easy to produce over-fitting phenomenon,it is proposed to use data augmentation and Dropout,The regularization method reduces the overfitting of the model,improves the generalization ability of the model,and effectively improves the recognition accuracy.(3)Based on transfer learning,to select a baseline model and to optimize the model,achieve high recognition accuracy,and use it in engineering.First compare the migration effects of the three networks to select the baseline model,and then improve the selected VGG16 baseline model,including introducing a batch normalization algorithm,adding a Convolutional Block Attention Module to the convolution layer to improve feature extraction efficiency,and redesigning the fully connected layer,which aims to reduce the calculation parameters,improve the training speed,and comprehensively improve the recognition and judgment accuracy of the model and the robustness and stability of the model.Finally,thinking from various angles,4types of comparative experiments were designed,the results of each type of experiment were analyzed,and the effectiveness of the improved model was verified from different levels and angles.The average recognition rate of the improved model is 95.6%.At the same time,by improving the Otsu achieve more accurate segmentation of corn disease spots and more accurate judgment of corn disease grades,the images collected by the hardware system are preprocessed and input into the improved VGG16 model to identify corn disease categories and disease grades.(4)Building an integrated corn disease identification and growth monitoring IOT system.It adopts the three-layer architecture model of the Internet of Things,which is composed of sensors and cameras,Raspberry Pi,wireless transmission module,server,monitoring platform,etc.,and contain the corn disease subsystem,The corn images and environmental parameters,displayed through the visual interface of the monitoring terminal,to judge the growth environment status and disease status of corn,to achieve the purpose of monitoring and feedback on corn,and to achieve the complete interconnection of the system.