Automatic interpretation of infrared aerial color photographs of citrus orchards having infestations of insect pests and diseases

Abstract

This paper describes a system designed to analyze aerial color infrared photographs of citrus orchards. The input to the system is a 35-mm transparency of the infrared aerial photograph mounted in a three-color film digitizer which is controlled by an XDS 930 computer. An algorithm locates the outline of each individual tree in the photograph and then cluster analysis is applied. For the trees of a transparency, the center of gravity (CG) of the set of trees is determined and then the distance of each individual tree from CG is computed by using an empirical distance function which we have derived. Sorting, performed on the basis of the computed distances, clusters the trees. The cluster identifying algorithm separates out the individual clusters which are then grouped together into a smaller number of classes by applying the same algorithm. Fifteen transparencies having mealybugs, brown soft scale, gummosis, and root rot infestations were analyzed and experimental results show very satisfactory classification of the infestations.