Iterative relabeling using spectral neighborhoods
A first image is received. An initial label is assigned to at least some pixels in the first image, including by assigning a first label to a first pixel. A determination is made that the first pixel's label should be replaced with a different label. The first pixel's label is updated with the different label.
1. A system, comprising:
a processor configured to:
receive a first image comprising an observation of a first area and receive a second image comprising an observation of the first area;
assign a first set of labels to at least some pixels in the first image, including by:
assigning an initial label to a first pixel in the first image;
determining that the first pixel's initial label should be replaced with a different label; and
updating the first pixel's initial label with the different label;
assign a second set of labels to at least some pixels in the second image; and
generate as output a classification map, at least in part by selecting, for a given pixel in the classification map, a label determined for the given pixel as it either: (1) appears in the first image, or (2) appears in the second image; and
a memory coupled to the processor and configured to provide the processor with instructions.
2. The system of claim 1 wherein the determination that the first pixel's initial label should be replaced with a different label is based at least in part on a majority vote of spectral neighbors of the first pixel.
3. The system of claim 2 wherein the spectral neighbors of the first pixel comprise nearest neighbors of the first pixel.
4. The system of claim 1 wherein the first pixel's initial label is assigned based at least in part by a classifier trained using a training set.
5. The system of claim 1 wherein the first pixel's initial label is assigned based at least in part by a classifier that takes as input the received first image and a boundary map corresponding to the first area.
6. The system of claim 1 wherein the first pixel's initial label is assigned using a historical classification associated with the first area.
7. The system of claim 6 wherein the historical classification associated with the first area is determined from crop data associated with the first area.
8. The system of claim 1 wherein the first image and second image are included in a time series of observations of the first area.
9. The system of claim 1 wherein the selecting is performed based at least in part on a first weight associated with the first image and a second weight associated with the second image.
10. The system of claim 8 wherein the first weight and second weight are based at least in part on a respective rate of convergence associated with the respective first and second images.
11. The system of claim 8 wherein the first and second images were captured by respective different first and second sources having respective different first and second weights.
12. The system of claim 1 wherein labels included in the classification map are represented as numeric values.
13. The system of claim 1 wherein the classification map comprises land use classification labels.
14. The system of claim 1 wherein the processor is further configured to provide the output to a yield predictor.
15. The system of claim 1 wherein the processor is further configured to refine the classification map using a boundary map.
16. A method, comprising:
receiving a first image comprising an observation of a first area and receiving a second image comprising an observation of the first area;
assigning a first set of labels to at least some pixels in the first image, including by:
assigning an initial label to a first pixel in the first image;
determining that the first pixel's initial label should be replaced with a different label; and
updating the first pixel's initial label with the different label;
assigning a second set of labels to at least some pixels in the second image; and
generating as output a classification map, at least in part by selecting, for a given pixel in the classification map, a label determined for the given pixel as it either: (1) appears in the first image, or (2) appears in the second image.
17. A computer program product embodied in a tangible computer readable storage medium and comprising computer instructions for:
receiving a first image comprising an observation of a first area and receiving a second image comprising an observation of the first area;
assigning a first set of labels to at least some pixels in the first image, including by:
assigning an initial label to a first pixel in the first image;
determining that the first pixel's initial label should be replaced with a different label; and
updating the first pixel's initial label with the different label;
assigning a second set of labels to at least some pixels in the second image; and
generating as output a classification map, at least in part by selecting, for a given pixel in the classification map, a label determined for the given pixel as it either: (1) appears in the first image, or (2) appears in the second image.
18. The method of claim 16 wherein the determination that the first pixel's initial label should be replaced with a different label is based at least in part on a majority vote of spectral neighbors of the first pixel.
19. The method of claim 18 wherein the spectral neighbors of the first pixel comprise nearest neighbors of the first pixel.
20. The method of claim 16 wherein the first pixel's initial label is assigned based at least in part by a classifier trained using a training set.
21. The method of claim 16 wherein the first pixel's initial label is assigned based at least in part by a classifier that takes as input the received first image and a boundary map corresponding to the first area.
22. The method of claim 16 wherein the first pixel's initial label is assigned using a historical classification associated with the first area.
23. The method of claim 22 wherein the historical classification associated with the first area is determined from crop data associated with the first area.
24. The method of claim 16 wherein the first image and second image are included in a time series of observations of the first area.
25. The method of claim 16 wherein the selecting is performed based at least in part on a first weight associated with the first image and a second weight associated with the second image.
26. The method of claim 25 wherein the first weight and second weight are based at least in part on a respective rate of convergence associated with the respective first and second images.
27. The method of claim 25 wherein the first and second images were captured by respective different first and second sources having respective different first and second weights.
28. The method of claim 16 wherein labels included in the classification map are represented as numeric values.
29. The method of claim 16 wherein the classification map comprises land use classification labels.
30. The method of claim 16 further comprising providing the output to a yield predictor.
31. The method of claim 16 further comprising refining the classification map using a boundary map.