FBM method

Feature-based matching (FBM) collects tie points (TP) based on features in the imagery. The features—key points (image locations) and descriptors—are read from feature database (FDB) files created during processing. Each key-point location is characterized by a feature descriptor, which is used to match corresponding points across images.

FBM is useful for matching images when the initial models and exterior orientation (EO) are of poor quality, because the matching does not require any information about the orientation of the images.

To automatically collect TPs using FBM

Click the Local exclusion mask list, and then do one of the following according to the mask you want to use, if any:
- To exclude no pixels from point collection, click None.
- To exclude only image pixels covered by the bitmap, click Last bitmap.
  That is, select this option when there are bitmap segments in one or more of the source-image files that define an exclusion region. When one of the source images contains multiple bitmaps, the last one is used.
- To exclude only image pixels enclosed by polygons, click Last vector.
  That is, select this option when there are vector segments in one or more of the source-image files that define an exclusion region. When one of the source images contains multiple vector segments, the last one is used.
- To exclude the pixels of a specific vector segment, click Specific segment, and then in the Segment number box, type or select the segment number you want.
Click the Photo feature mask list, and then do one of the following according to the mask you want to use, if any:
- To exclude no pixels from point collection, click None.
- To exclude pixels near the fiducial marks (scanned-film airphotos), click Fiducial, and then in the Margin distance box, type or select the amount to expand the mask from the fiducial locations.
- To exclude pixels along the edge of each image, click Edge, and then in the Margin distance box, type or select the amount to expand the mask from the image edges.
In the Rejection threshold box, type or select a value, in pixels, by which to eliminate blunders during random sample consensus (RANSAC).
In the Maximum key points box, type or select the number of key points to extract per image.
Note: The actual number of extracted key points may differ due to a lack of features in the image or because of the minimum numbers of key points enforced to improve the distribution, match-ability, or both.
Click the Thin points list, and then do one of the following according to the thinning you want to apply, if any:
- To apply thinning automatically, click Automatic.
- If you do not want to apply thinning, click None.
- To apply custom thinning, click Custom, and then in the Cells per image box, type or select the number of cells you want to use.
  That is, by defining the number of grid cells per image, you can control the amount of thinning that occurs. The greater the value, the greater the number of points retained. The thinning process divides each image into the number of grid cells you specify and keeps the TPs with the most connectivity in each grid cell.
In the Feature database folder box, type the path and file name of the folder to which to write feature database (FDB) files or, to select a folder, click Browse.
Note: The FDB files created during processing are intermediate and cannot be viewed.
Click the Search area list, and then do one of the following according to area of the imagery you want to search for matching features:
- To match all the key points of each image, click Entire image.
  Typically, you select the entire image when the overlapping area between images cannot be calculated with high accuracy, such as when the math models and orientations of images are inaccurate. Points matched erroneously in non-overlapping areas will be eliminated using RANSAC.
- To restrict key-point matching to only the points in the overlap area between any pair of overlapping images, click Overlap area.
  Typically, you select the overlap area only when the area can be calculated with high accuracy; that is, confidence is high that the math models and image orientations are highly accurate.
Click the Overview level list, and then select the overview level to use during point collection.
Note: Decreasing the overview level will increase processing time. Using overview level 8 will usually produce good results, but for smaller images you may need to decrease the overview level used. If FBM does not find many points, lower this value.
In the Matching channel box, enter the channel to use during collection.
Using the green channel, if it exists, usually produces the best results.
Proceed to Selecting images, start time, and run.