Computer Recognition of Citrus Infestations

A computer software system is described that uses digitized color information from aerial color infrared transparencies to detect the presence of citrus mealybug (Pseudococcus citri Risso), brown soft scale (Coccus hesperidum L.), and Rio Grande gummosis in individual citrus trees. The color coordinates at each spatial point, and color differences at adjacent points are used to locate the trees and to detect the infestations; and compensation is made for the variation in color characteristics between different transparencies. The system requires the input of four parameters: a flag denoting the presence of heavy shadows in the image, nominal tree s'ize and spacing of the citrus trees, and a flag denoting the season of the year when the transparency was taken. An index of recognition, Iq, was defined and used as a measure of recognition effectiveness. For unknown data, Iq ranged from 43% to 81%, with nominal values of 60% to 80% for all three infestations.