Research on an Extracting Method to Assist Manual Identification of Targets - Advances in Image and Graphics Technologies - page 117

Fig. 5. Double Window Binarization Fig. 6. Erosion

Fig. 7. Extracted Target Fig. 8. Dilation

Though the observation of Fig. 8, one can find that the useless information is re-
moved and the suspected targets have been extracted. Photographic interpreter can
find out targets in the remaining suspected targets. The workload is greatly reduced
and the interpretation precision is improved.

4.3 Applicability Analysis

In order to verify the applicability of this method, we did experiments on a series of
pictures. According to the actual operation, some of the parameters used in the expe-
riment are adjusted.
The experimental results show that the method can achieve rapid and effective ex-
traction through the remote sensing image owns complex background and objects area
with homogeneous gray value. The missing rate tends to be zero. Especially, for the
linear targets, such as airport runways and roads, it works the best (as shown in Fig. 9
and Fig. 10). For targets, which have non-uniform inside, miss rate is high (as shown
in Fig. 11).