The top image has been reclassified (supervised)by manually selecting pixels that represent certain features. A few selections were made per feature type (water, residential, agricultural etc...) Common feature classes were then recoded to combine them into more general groups. These groups were then color coded to match their feature type (i.e. water=blue etc....).I thought the procedure went smooth, however, the results were rather confusing. There seems to be a problem with my classifying of grass and residential areas. The problem may be the Region Growing Properties, specifically the Spectral Euclidean Distance. I am still a bit confused about this function.
.JPG)
Yes, the spectral euclidean distance is what the program uses as a sample for growing the region. So if you set the distance two high it may include areas that you may not want and vice versa. You can experiment with different distances and see what the difference is (if you want!). It is trying to find the balance that is the tricky part...a little beyond this lab!
ReplyDelete