Estimating (green-forest) land cover in satellite images of Rutgers' Cook-Douglas campus This experiment illustrates how the eye-tracker can be used to evaluate the effectiveness of different visual features in conveying information. Here we simply use different colors to represent different land-uses and compare the impact of simultanteous presentation of different colors. Are we more accurate in estimating these colors when presented in isolation? Looking at the four conditions below, it looks as though the magnitude of the estimations may decrease as more colors are presented simultaneously. Also, the relative saliency of the colors is likely to have a substantial influence on our judgments with the obvious case below of red overwhelming the green, blue and yellow information. Another comparison illustrated here assesses the impact of satellite background information (bottom set of images) vs. isolating relevant regions on a white background (top images). One general prediction for all these conditions is that relative estimates of target areas are likely to decrease as more information is presented in the same image. I also expect that scanning behavior will become more erratic in these conditions where there is much more competing and thus distracting information. Sample data are presented below (but don't expect any meaningful trends as I'm the only subject so far & a very biased one at that!) |
|||||
1  |
2  |
3  |
4  |
||
|
 |
 |
 |
 |
White Background (images linked to results) |
 |
 |
 |
 |
Satellite Background |
|
Predictions |
|||||||||
| 1 | Scanned & estimated green land cover will decrease w/ # simultaneous colors. i.e., Fewer fixations and saccades in green regions. (This prediction is based on my simple observation that irrelevant colors seem to mask target colors (i.e.,green in this case). |
||||||||
| 2 | Scanning behavior will be concentrated in the central region of the display | ||||||||
| 3 | Satellite Background (more irrelevant information) will generate more erratic scanning behavior. ie., more fixations & saccades outside of defined regions & more variability in time-series |
||||||||
| 4 | An dditional goal here is to evaluate fractal properties of the time series across the different background goal. | ||||||||
Analyses To follow the analyses , be sure to look at Color &Interest Areas for definition of the different conditions |
|||||||||
|
|||||||||
1 |
|||||||||
3 |
|||||||||
DJ Aks 8/23/06
|
|||||||||