Generation of synthetic imagery is not generally the end of the process. The imagery needs to be viewed, either by humans or by machine depending upon the application. Our work is primarily concerned with making use of synthetic imagery to assess target conspicuity.

Out imagery is generated by CAMEO-SIM which is a physics based image rendering system and therefore allows parametric studies to be undertaken (see the presentation by Marilyn Gilmore for more detail). That is, once a particular scenario has been created it can then be used to predict effects under different conditions such as diurnal variations and weather conditions. The same scene geometry can be used and different material properties assigned for different times of the year.

Three methodologies are being developed to assess target signatures.

Observer Trials

This is a traditional psychophysical approach. Images of static or moving target are displayed to subject and performance is measured, usually in terms of probability of detection and detection time. We are currently looking at methods to achieve more realistic light levels within the laboratory (typical monitor luminances are 10 – 50 cd m-2, whereas light levels in the outside world can be 3-4 orders of magnitude higher). One method being considered is direct viewing of images produced by an LCD projector which ahs the potential to generate considerably higher light levels. Another issue that arises in these experiments is how to map the very large dynamic range values produced by CAMEO-SIM into the limited range available on displays.

Contrast Metrics

This is a computational method of assessing target signatures that uses a multiscale contrast calculation (Peli), and in some sense approximates early stages of vision.

Vision Modelling

This third method uses a fully developed image based vision model – GTV developed at the Georgia Tech research Institute. The model simulates preattentive and attentive vision processes and includes processing of colour and simple motion. DERA is currently working with Georgia Tech on validating and developing the model with the aim of applying it to analyse synthetic imagery.

Each of these topics will be discussed within the context of assessing synthetic imagery and examples using images of moving aircraft and static land vehicle target detection will be presented.