Experiments

 

In the current version of the system, images are obtained from a customized platform featuring a camera, an electronic compass, and a differential GPS system, all mounted on a tripod. The compass provides absolute orientation measurements; once calibrated, accuracy is plusmn0.5 degree. A differential GPS device obtains ground-truth measurements of position, with accuracy of 3 meters. The camera is calibrated [12], and errors of plusmn2 degrees are introduced by calibration inaccuracies. The images are stored on tape and played back when testing the system; every image is tagged with six dimensional ground-truth.

We have run the estimation procedure with data obtained in Pittsburgh using sequences of images as in the last two panoramas of Figure 4. Two mountain peaks are detected in the first panorama, three in the second. The bearings, in degrees, are {177.6, 131.1, 92.4, 102.3, 65.3}. The system estimates position with an error of 87 meters. We have also run the system on data from Dromedary Peak, Utah; the panoramas are shown in Figure 4. The five bearings detected automatically are (in degrees): { 224.5, 198.8, 163.7, 107.2, 100.8, 96.3 }. A square area of 6km by 6km was used the Dromedary Peak quadrangle, to make the test similar to the Pittsburgh tests. The system estimated position with accuracy of 95 meters, which can be compared to the 71,700m² obtained by Thompson [17]. We benefit greatly from our reliance on all available features, large and small, present in the image.