Detecting the Honeydew of Common Pistachio Psylla Pest by Using Image Processing Technique

Pests and disease control have always been one of the main concerns and challenges of farmers and growers. The use of machine vision and image processing has greatly helped growers in pest management. The purpose of this study was to use image processing technique to detect honeydew produces by pistachio psylla and find the relation between the percent of leaf coverage by honeydew and pistachio psylla infestation. The leaves were collected from the research orchard with various infestation rates. Leaf samples were imaged by three cameras with 7, 13 and 20.7 MP resolutions at the same exposure conditions in imaging chamber with controlled lighting conditions. Images were processed in the Matlab R2019a using Watershed and Otsu segmentation algorithms to find the percentage of leaf surface covered by honeydew. The covered area was calculated using predefined functions in the image processing toolbox. A graphical user interface (GUI) was also designed to make the program more user friendly. Considering TPR mean value of 0.95 and total accuracy of 0.88 for watershed segmentation method showed its acceptable performance in discriminating honeydew out of other objects in images. Coefficient of determination and regression equation between pest population (obtained from manual count by expert) and percentage of leaf area covered by honeydew were obtained for different cameras. Camera with 20.7 MP resolution achieved the best performance with coefficient of determination 0.93 and regression equation y=1.03 x. The results from other cameras were also satisfactory.