Image Processing And Data Analysis Of Maize Disease Based On Deep Learning

In the last few years,corn plays an important role in Chinese grain crops.Due to its high nutrition and medical value,it quickly occupies a big market.Therefore,the detection of corn leaf diseases is particularly important in the whole growth process,aiming at the problems of traditional corn leaf diseases that are not easy to identify and identify errors,this paper studies the detection method based on deep learning environment,changes and optimizes the relevant algorithms,and ports the algorithm load to the Huawei development board Atlas 200 DK to complete the detection of corn leaf diseases.Around the above content,the following work has been done:The deep learning environment was built,the corn leaf disease dataset was expanded,a series of data enhancements were carried out on the dataset,and then the YOLOX-S target detection algorithm was selected,and the algorithm was optimized,and four CBAM attention mechanism modules were added to improve the recognition rate of the target detection algorithm.To investigate the feasibility of the proposed algorithm,we have trained a variety of target detection algorithms,such as YOLOv3,YOLOv4 and Faster R-CNN.Compared with YOLOX-S,the improved YOLOX-S algorithm improves the recognition precision by0.2%,moreover,the improved YOLOX-S method has better recognition and accuracy than the other three algorithms.Build the development environment and operation environment of the development board Atlas 200 DK.Set up the hardware environment,configure the relevant parameters,and then build the running environment of the software to build a platform for the smooth progress of the experiment.Then the improved YOLOX-S object detection algorithm is transplanted to Atlas 200 DK,based on the characteristics of the development board and related computing attributes,the detection of corn leaf diseases is completed,and its effect is tested,which is exactly the same as the GPU development test effect,and the transplantation of object detection algorithm is realized,and the GPU algorithm is transplanted to Atlas200 DK,and it still performs well.