Remove False Alarms
INTRODUCTION
However some false alarms remain. Thus we use an image of markers to remove most of them. We have kept the image which corresponds
to the minimum value of conditional variances. We classify this image by
Kmeans with the number of classes equal to the one obtained by the modified
Fuzzy Cmeans algorithm. Then we filter the classified image with a median filter to remove noise. The resulting image gives
the interior limits of urban areas and marks the existence of urban areas. The segmented urban areas that have a null intersection with the
result of median filter are considered to be false alarms.
RESULTS
The combination of the Gaussian data attachment and the image of markers
give accurate results i.e. precise limits and few false alarms.
(a): Superposition of the filtered image (red)
and the segmented image (village of Toreilles).
(b): Superposition of the filtered image (red) and the segmented
image (village of Canet en Roussillon).
 |
(c) and (d): After removing false alarms
 |
 |
| c |
d |
Last
modified: Tue May 5 09:49:05 MET DST 1998