So far I am able to get the fuzzy member value of different classes for each pixel. I want to know the steps for producing a hard classified fine resolution image using psa. Do I have to first create a fine resolution grid and then randomly fill those sub pixels based on their fractions and then to maximize the energy function for optimisation ? I need a basic overview about srm technique and how such objective functions can be programmed.
Thanks
Since pixel swapping is a methodology for extracting sub-pixel information out of mixed pixel response, the spatial arrangement is to be changed in such a way that the spatial correlation between neighboring sub-pixels is maximized. The initialization of sub-pixel locations is important. Instead of random initializing, define an ‘attractiveness’ of sub-pixels of identical class as summation of distance-weighted function of neighboring sub-pixels. Two sub-pixels of less attractiveness are selected and their class values are swapped, if attractiveness is thus increased. Otherwise no change is made. This attachment may be helpful.
I am actually struggling in the implementation part for the program. If have some sort of working code that would be a great help.
- Badges
- Science topic
- Similar topics
- Geoscience
- Geoinformatics