We want to classify multiple images using training sets and then a random forest classifier. We are planning to mosaic these images and then get the training sets and the classification. We would like some suggestions to where this approach would work or if you had a suggestion to improve our strategy. We are working with 1-m resolution imagery and 4 bands (red, green, blue, and near-infrared). Thanks a lot in advance for your comments and suggestions!
In the world of satellite imagery, mosaics are large collections of overlapping satellite images and data that are merged to create one larger image. If this is what you want you may need to enhance images using as an example Histogram matching which is the process of determining a lookup table that converts the histogram of one image to resemble the histogram of another. Histogram matching is useful for matching data of the same or adjacent scenes that were scanned on separate days, or are slightly different because of sun angle or atmospheric effects. This is especially useful for mosaicking or change detection.
After that, you can mosaic images and perform a Random Forest classification.
one way of conducting Mosaic is using the following
An Efficient Mosaic Algorithm Considering Seasonal Variation: Application to KOMPSAT-2 Satellite Images
Article An Efficient Mosaic Algorithm Considering Seasonal Variation...
It is not obvious what you mean by image mosaic. You could compute similarity of images and try to arrange these images in a 2D grid so that images with a low mutual distance cluster together in the plane. That's topology preserving mapping, see Kohonen networks.
In order to compute a distance between images, you could partition each image into a grid of tiles and determine the average RGBI value for each tile. Then you could sum up the squares of the color differences between every two corresponding tiles of different images.
Regards,
Joachim
In the world of satellite imagery, mosaics are large collections of overlapping satellite images and data that are merged to create one larger image. If this is what you want you may need to enhance images using as an example Histogram matching which is the process of determining a lookup table that converts the histogram of one image to resemble the histogram of another. Histogram matching is useful for matching data of the same or adjacent scenes that were scanned on separate days, or are slightly different because of sun angle or atmospheric effects. This is especially useful for mosaicking or change detection.
After that, you can mosaic images and perform a Random Forest classification.
one way of conducting Mosaic is using the following
An Efficient Mosaic Algorithm Considering Seasonal Variation: Application to KOMPSAT-2 Satellite Images
Article An Efficient Mosaic Algorithm Considering Seasonal Variation...
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training