This paper presents a system for object detection and counting from transparencies digitized using a scanning microdensitometer. Classification and clustering methods are applied sequentially to digitized transparencies of an orange tree and a low altitude color infrared aerial photograph of a citrus orange. The objective is to count oranges on the digitized ground transparency and to count fruit trees on the digitized infrared aerial photograph. This general flow can be applied to any type of spectral data for object detection and counting.

Date of this Version