Research On Recognition Of Fruit And Vegetable Leaf Diseases Based On Computer Vision

The output of fruits and vegetables in agricultural production occupies an important position.However,due to the complex and ever-changing environmental factors in the growth process,it is highly prone to disease,resulting in a decline in production and a huge resource.In recent years,computer vision technology has developed rapidly,and agricultural disease identification has gradually become a new application field of this technology,which plays an important role in the process of agricultural development.Aiming at this problem,this paper takes the typical downy mildew,powdery mildew and target spot disease of cucumber leaves as a research sample,and proposes a method based on computer vision technology to identify and warn the disease of fruits and vegetables.Based on the analysis of the application scenarios and objectives,the article first gives the main functional implementation requirements,and determines the red rice NOTE 4 as the Android software platform.The development of the disease identification method is completed on the PC side of the existing Windows 10 system.The HTTP protocol implements the transmission of information,deploys the client and server development environment on the PC,and installs and configures the integrated development tool Eclipse and the computer vision library Open CV.Then,100 images of each of the three diseases were preprocessed,the noise and various redundant information in the image were removed,and the gray image and binarized image of the disease image were obtained,combined with the full threshold method and the local dynamic threshold method.The lesion area and the transition area in the image are segmented to prepare for subsequent feature extraction.According to the characteristics of the three diseases of cucumber,13 features of the color,shape and texture of the lesion area and the transition area were extracted by calling the corresponding function in the Open CV library in Eclipse.The LIBSVM tool was used to compare the transition area.Whether or not the feature data has an effect.After repeated test and verification,the radial basis kernel function is used to punish the parameter C=32.When the kernel function parameter γ=1,26 feature data of two divided regions are used as the input set.SVM has the best classification effect,downy mildew and white powder.The optimal recognition rates of disease and target spot disease were 82.5%,75%,and 90%,respectively.After that,the client image acquisition and result display software was developed on the Android mobile phone to realize the interaction with the server-side disease recognition.The client can obtain and display the recognition result in real time and provide information saving function.Finally,the environmental factors affecting the disease of leaf and leaf of fruits and vegetables were analyzed.The causes of the downy mildew of cucumber leaves were analyzed.The monitoring system of downy mildew disease was developed through the monitoring of air temperature and humidity.The system can view environmental information,weather information,and provide historical data queries in real time.The whole system realizes the effect of image collection,result display and disease identification separation,and provides certain early warning functions.It has played a role in the free mobility of Android devices and the efficiency of PC data processing.The recognition results are good and the real-time interaction is strong.It has certain reference significance for the development of disease identification in fruits and vegetables.