Accurate Detection And System Implementation Of Disease Spore Germination Rate Based On Deep Learning

Fruits and vegetables rot in the process of picking,transportation,and storage have brought great losses to the world’s fruit and vegetable production.The rot of fruits and vegetables is mainly due to the pathogenic effect of pathogenic fungi.In order to obtain the drug resistance and activity level of disease fungi,researchers need to collect and count information on fruit and vegetable fungi.The traditional methods of manual detection of diseases and insect pests rely on the experience of farmers,or ask for expert guidance,which is slow,inefficient and expensive.High and subjective.In order to meet the needs of the development of modern and large-scale fruit and vegetable planting industry,it is urgent to develop a smart and efficient automatic identification method for fruit and vegetable diseases and insect pests.Curvuria leaf spot,black spot,and anthracnose are three types of fruit and vegetable diseases with high incidence.A disease spore is used as the research object.The improved instance segmentation algorithm Mask Scoring R-CNN is used to detect and segment these three types of spores and germinated spores under the microscope.By comparing the results of multiple network models,the aim is to propose an efficient and A high-precision method is used to identify fruit and vegetable diseases,and the improved model is integrated with the Web system to build a fruit and vegetable disease spore detection system.The main work of this study is summarized as follows:Collect thousands of images of fruit and vegetable disease fungal spores under the microscope,manually label the type,quantity,specific location and germination of disease spores in each image,and build a fruit and vegetable disease fungal spores dataset.It has a certain reference effect.An improved Mask Scoring R-CNN algorithm was proposed to identify fruit and vegetable disease fungal spores in complex backgrounds,which solved the problem of difficult identification of fruit and vegetable disease fungal spores in complex backgrounds.This research introduces ResNeSt and Involution in Mask Scoring R-CNN,replaces the original network backbone part with ResNeSt,and replaces the Convolution of FPN in the original network with Involution.The labeled dataset is applied to six different instance segmentation models,and it is verified by experiments that the improved model has the best performance on the labeled dataset.When the intersection ratio is 0.5,mAP reaches 0.710,compared with the original Mask Scoring R-The mAP value of the CNN network model has increased by about 3.3 percentage points.Experiments show that the improved Mask Scoring R-CNN algorithm has better detection accuracy,the trained model has better segmentation effect,and can better identify and segment this dataset spores.A fruit and vegetable disease spore detection system based on the Flask framework is built.The improved model in this paper is deployed to the Web system,so that the system can perform batch detection on the uploaded microscopic images of fruit and vegetable disease spores and return the detection results.In summary,the improved Mask Scoring R-CNN model is used in this paper to achieve better results,and the improved model is deployed to the Web system,so that agricultural workers can easily use the model to detect disease spore images under the microscope,and obtain the detection result map and The germination rate and number of various spores in the figure have a clear understanding of the current impact of diseases and insect pests on fruits and vegetables and the control effect of currently used pesticides on disease spores,so as to realize the precise treatment of pesticides and fertilizers.The work done in this paper has played a positive role in the prevention and control of fruit and vegetable diseases.