Development Of Corn Disease Recognition And Diagnosis System Based On Convolutional Neural Network

Disease affects the quantity and quality of corn,which is regarded as one of the most important crops in our country.In terms of corn disease management,the traditional corn disease diagnosis system mainly relies on the personal experience and pathological knowledge of plant protection experts to identify leaf disease through visual inspection,which has various problems like low work efficiency,strongly subjective identification,etc.With the popularization of artificial intelligence and the rapid development of computer vision,image processing technology is increasingly used in the field of agricultural disease identification,whereas there are a range of problems including low stability and poor effect in the complex field.As a result,in recent years,traditional vision technology based on characteristics of disease identification like texture and color has been replaced by a more stable,adaptable,and accurate Convolutional Neutral Network.In view of the bottlenecks in the current corn disease diagnosis system and its situation,this paper takes the Convolutional Neural Network of transfer learning as the core function of the corn disease diagnosis,and develops four modules,including user function module,corn disease diagnosis module,encyclopedia of corn disease module,and prevention and treatment of corn disease module.Firstly,methodological research on the existing corn disease diagnosis system is conducted;secondly,based on the problems existing in the corn disease diagnosis system,its bottlenecks are analyzed via a combination of questionnaire,interview and brainstorming with the improvement group to draw a “fish-bone diagram”;additionally,its bottlenecks are eliminated through replacing the artificial diagnosis with the Convolutional Neural Network of transfer learning and using questioning technique---“5W1H” and 5 principles---“ECRSA”;finally,the feasibility of optimized system for corn disease diagnosis is analyzed,designed,and tested.Compared with the traditional corn disease diagnosis system with plant protection experts as the core,according to the results of the research,the accuracy rate of the optimized system which uses the transferred MobileNetV2 as the core is up to 97.23% and the weight is only 8.69 MB.The entire corn disease diagnosis system in this paper is tested through functional and compatibility tests.Furthermore,considering that corns take around 52 days from planting seeds to growing fruits,effective treatment to early disease is the key point.The corn disease diagnosis system can identify and diagnose corn disease quickly and accurately,curbing the corn disease more effectively.