Satellite image classification across multiple resolutions and time using ordering constraint among instances

1. A method comprising:

classifying low-resolution pixels of a low-resolution satellite image of a geographic area to form an initial classification map;

selecting at least one physically-consistent classification map of the low-resolution pixels based on the initial classification map; and

using a water level associated with at least one of the physically-consistent classification maps and an elevation map that provides an elevation for each high-resolution pixel to identify a set of high-resolution pixels as representing a perimeter of water in the geographic area.

2. The method of claim 1 wherein each of the physically-consistent classification maps of the low-resolution pixels provides a set of low-resolution pixels representing a perimeter of the water in the geographic area that is different from the perimeter of the water represented by the high-resolution pixels.

3. The method of claim 2 wherein the perimeter of the water represented by the high-resolution pixels passes within a low-resolution pixel.

4. The method of claim 1 wherein using the water level associated with at least one of the physically-consistent classification maps comprises using a lowest water level associated with any of the selected physically-consistent classification maps.

5. The method of claim 1 wherein each selected physically-consistent classification map is associated with a respective threshold number of high-resolution pixels such that a low-resolution pixel in the physically-consistent classification map is designated as land unless a water level associated with physically-consistent classification map will cause more than the threshold number of high-resolution pixels in the low-resolution pixel to contain water.

6. The method of claim 5 wherein the at least one selected physically-consistent classification maps comprise at least two physically-consistent classification maps with different associated threshold numbers of high-resolution pixels.

7. The method of claim 1 further comprising:

using a second water level associated with at least one of the selected physically-consistent classification maps to identify a set of high-resolution pixels as representing a perimeter of land in the geographic area.

8. The method of claim 1 wherein selecting at least one physically-consistent classification map of the low-resolution pixels based on the initial classification map comprises selecting a physically-consistent classification map based on a number of differences between the physically-consistent classification map and the initial classification map.

9. The method of claim 8 wherein a plurality of physically-consistent classification maps of the low-resolution pixels have the same number of differences when compared to the initial classification map.

10. A system comprising:

a memory having computer instructions; and

a processor executing the computer instructions to implement:

a classifier receiving a low-resolution image of a geographic area and classifying each pixel of the image to form an initial classification map;

a comparison module comparing the initial classification map to a plurality of physically-consistent classification maps to select a physically-consistent classification map; and

a high-resolution classifier classifying high-resolution pixels of the geographic area based on the selected physically-consistent classifier map and based on an elevation map that provides an elevation for each high-resolution pixel.

11. The system of claim 10 wherein the high-resolution classifier classifies the high-resolution pixels by selecting a water level associated with the physically-consistent classification map and determining which high-resolution pixels would contain water at that water level.

12. The system of claim 10 wherein the comparison module selects a plurality of physically-consistent classification maps and the high-resolution classifier classifies the high-resolution pixels based on the plurality of selected physically-consistent classifier maps.

13. The system of claim 12 wherein at least one of the physically-consistent classification maps is associated with multiple water levels.

14. The system of claim 13 wherein the high-resolution classifier classifies the high-resolution pixels by identifying a lowest water level associated with any of the selected physically-consistent classifier maps and determining which high-resolution pixels would contain water at that lowest water level.

15. The system of claim 14 wherein the high-resolution classifier classifies the high-resolution pixels by identifying a highest water level associated with any of the selected physically-consistent classifier maps and determines which high-resolution pixels would not contain water at that highest water level.

16. A method comprising:

receiving a low-resolution image of a geographic area and classifying each pixel of the image to form an initial classification map;

comparing the initial classification map to a plurality of physically-consistent classification maps to select a physically-consistent classification map; and

classifying high-resolution pixels of the geographic area based on the selected physically-consistent classifier map and based on a high-resolution elevation map that provides an elevation for each high-resolution pixel.

17. The method of claim 16 wherein classifying the high-resolution pixels comprises classifying the high resolution pixels by selecting a water level associated with the physically-consistent classification map and determining which high-resolution pixels would contain water at that water level.

18. The system of claim 16 wherein at least one of the plurality of physically-consistent classification maps is associated with multiple water levels.

19. The system of claim 18 wherein classifying the high-resolution high-resolution pixels comprises identifying a lowest water level associated with any of the selected physically-consistent classifier maps and determining which high-resolution pixels would contain water at that lowest water level.

20. The system of claim 19 wherein classifying the high-resolution pixels comprises identifying a highest water level associated with any of the selected physically-consistent classifier maps and determining which high-resolution pixels would not contain water at that highest water level.