Research On Detection Method Of Grape Leaf Disease Based On Internet Of Things Data Fusion

Grape is one of the most common fruits in China.With the improvement of the quality of life,the sales of wine are also increasing.However,the harm of diseases and insect pests is the biggest fear of large-scale planting.The traditional detection method of grape leaf disease is that fruit growers rely on their own long-term accumulated experience to judge the grape disease category.This method is subjective,poor diagnosis effect and easy to misjudge.In this study,grape leaves were taken as the research object,and five common diseases were taken as the research objects.The morphological operation and feature extraction were combined.The appropriate threshold was selected to remove the background of grape leaves under complex background.The feature fusion and disease spot matching were combined.Under the supervision of adaptive fusion strategy,the detection of grape leaf diseases was completed.The main research contents are as follows:(1)A multi properties fusion algorithm for disease spot segmentation is proposed.The features of the disease spots were extracted,and the feature factors which had a great influence on the segmentation of the disease spots were selected.The adaptive factors were fused and the monochromatic background diseased leaves were used in the image segmentation.The trial outcomes indicate that the error ratio of the way is lower than 1.90,which is clearly better than the single properties fusion partitioning algorithm.(2)From the point of view that the same type of disease has similar matching degree,an adaptive multi properties fusion strategy is studied,Defined Acc＿top indicators to evaluate performance.Finally the three basic properties and SIFT properties are fused and their matching degree is reckonion.The trial outcomes indicate that the ACC of the image after multi feature fusion＿Top1 reaches 90.84%,much higher than the original image and single feature segmentation image,slightly higher than the improved k-means segmentation image,and the matching time is also significantly improved.The recognition rate of grape powdery mildew was 92.60%.The recognition rate of Anthracnose and blackspot were higher than 80%,85.30% and 84.12% respectively.