Describing the problem-solving strategies of expert image interpreters using graphical knowledge elicitation methods

Developing insights from remote sensing data requires a human-in-the-loop. The earliest examples are of air photo interpretation experts working with minimal tools. Today, experts in image interpretation rely heavily on the use of computer displays to assist them in insight generation. The goal of this current research is to develop a method for systematically describing the problem-solving tasks completed by experts when faced with the challenge of mapping land cover changes from remote sensing imagery. Knowledge elicitation (KE) methods are used to derive the interpretation processes of experts. Two KE methods, self-constructed external representation (SER) and think-aloud protocol (TAP), are used to develop a descriptive model of expert image analysis. A set of cognitive tasks are identified using qualitative coding of the SER, TAP, and reflection interviews. From these codes, a generalized interpretation model is generated to reflect the use of TimeSync, a visual interpretation tool, for land cover change mapping. Two findings emerge from this work. First, the use of SER and TAP in conjunction with qualitative coding is a useful means for generating the general interpretation model. Second, the interpretation of land cover change from remote sensing imagery consists of several cognitive subtasks that support a process of identification and signification process. The main contribution of this work is toward the development of a useful framework for using graphical KE methods in the context of visual work.