This paper investigates the visual search process and the effect of contextual information on the search process in an urban combat environment. High resolution combat simulation models implement a parallel sweeping or "windshield wiper" search process that is not representative of human search behavior. Furthermore, combat models do not account for additional situational awareness in the form of contextual information. A discrete myopic search model is proposed, a study of which provides a statistical model based on human performance data. This model prioritizes search effort where humans believe that targets are most likely to occur. Nineteen volunteers searched 16 static urban scenes with zero to five targets. These data formed the probabilities that a target is located in each cell in each discretized scene. The discrete myopic search model chooses the cell with the highest probability for each discrete look. Hypothesis testing on experimental data revealed a nearly 20% increase in accurately predicting human search patterns using the discrete myopic search model over the windshield wiper model. Further investigation revealed a significant change in search behavior and detection performance based on the addition of contextual information. The major result of this work indicates that combat models need to bias search based on the situational awareness of observer and properties of the observer's environment.
Conference paper: pdf (242KB)