Research And Application Of Maize Leaf Disease Recognition Based On Image Super-resolution

Maize is one of the three major food crops in my country.Due to the impact of climate change,maize leaf diseases have shown an increasing trend year by year in recent years,causing serious economic losses.Therefore,timely diagnosis of maize leaf diseases is of great significance to reduce the occurrence of diseases.Aiming at the problem that common corn leaf diseases are difficult to identify in the field environment under complex background,this thesis uses image recognition technology and image super-resolution technology to carry out related research,and develops an application program for corn leaf disease identification,which has high practical value.The main work of this thesis is as follows:(1)Image super-resolution model fused with wavelet attention.In order to solve the existing problems in image reconstruction,such as image blur and lack of details,this thesis proposes an image super-resolution model that incorporates wavelet attention,and combined with the idea of HAN,a layer attention module is added to capture the Long distance dependencies between layers.In addition,a channel spatial attention module is added to incorporate channel and contextual information.These two attention modules work together in multi-level features to better capture feature information.Then by constructing a data set and comparing it with three super-resolution models.The experimental results show that the reconstruction effect of the WAHAN model is better than that of RCAN,SAN,and HAN,and has higher visual effects and image quality.(2)Maize leaf disease recognition based on image super-resolution.Aiming at the problem of accurate identification of maize leaf diseases in the field environment under complex background,on the basis of preprocessing the fuzzy images by using image superresolution,the target detection model is further used for disease identification.Based on the data set collected by myself,this thesis conducts comparative experiments on various image super-resolution models and target detection models.The experimental results show that when using the WAHAN image super-resolution model and the YOLOv5 target detection model,the recognition accuracy can reach more than 90% at 2,3,and 4 times scaling factors.This method provides an effective solution for maize leaf disease identification in field environments with complex backgrounds.(3)Development of corn leaf disease identification APP.This thesis develops a mobile APP for corn disease identification based on Flutter and MLKIT framework.Aiming at the problem of network instability that may occur in the field environment with complex backgrounds,this thesis realizes the dynamic identification function of maize leaf diseases that does not depend on the network,and improves the stability of the program.At the same time,in order to use the function of the image super-resolution model more conveniently,the APP specially implements the zoom function,thereby improving the resolution of the blurred image.