An automated change detection system

The aim of the research described is to prototype algorithms which would form a suitable basis for an operational, automated map update system. Using forestry in the U.K as an example, algorithms were developed based on edge matching and object analysis which could be applied with or without a ground map and with various levels of user intervention. These were tested using Thematic Mapper data and digitised ground cover maps on the Elchies Forest near Elgin, Scotland. Where possible, results were augmented using evidential reasoning to reduce false alarms based on specific knowledge of forestry practice. Results indicate that even with the minimum of user intervention major areas of change can be detected. Increasing user intervention can reduce processing complexity and time, but is less suitable for an automated system. The paper presents the overall framework in which the techniques might be exploited.