Identification Of Strawberry Leaf Diseases And Insect Pests Based On Computer Vision

With the rapid development of information technology,image processing technology is more and more widely used.With the rapid development of strawberry planting in Taigu County,Jinzhong City,Shanxi Province,and the increasing of strawberry planting area,the impact of diseases and pests on Strawberry growth results in the decline of strawberry output,which in turn affects the economic income of fruit farmers.It is found that different kinds of diseases and insect pests will have different effects on the growth of strawberry leaves,and different diseases will have different characteristic information.It is one of the important problems to effectively identify and control the diseases and insect pests to ensure the economic income of fruit farmers.But at present,the identification of strawberry leaf diseases and insect pests mainly depends on the experience of the observer,which can not guarantee the real-time and accurate judgment.In this paper,the common powdery mildew,gray mold,red central column root rot and Lepidoptera pests in strawberry leaves were identified and studied.The image processing technology was used to preprocess the strawberry leaves diseases and pests images: image graying,binarization,image denoising,image lesion segmentation.The severity of diseases and insect pests in strawberry leaves was studied by using image segmentation.Through the support vector machine,the strawberry leaves with 15 feature information of texture,color and shape of diseases and insect pests are recognized and classified automatically,and the recognition rates of different kernel functions of SVM classifier are compared.The results show that RBF function has the best recognition rate of strawberry leaf diseases.In the complex background,the recognition rate of strawberry leaf diseases and insect pests is 76% through the fusion of color features and shape features,because the color and shape of the spots are different.In this paper,we use deep learning of alexnet,vggnet16,resnet50,densenet121 and mobilenetv2 to classify and identify diseases.The experimental results show that the highest recognition rate of densenet121 is 92.88%,that of alexnet is 87.19%,that of vggnet is 83.99%,that of resnet50 is 87.69%,and that of mobilenetv2 is 86.21%.The experimental results show that: for the strawberry leaf disease and pest images with complex background,the effect of using densenet121 convolution neural network is the best,the recognition time is the shortest,and the efficiency is the highest.