FBMTIE

Description

FBMTIE performs feature-based matching to collect tie points (TP). The features are read from feature database (FDB) files created by the FBMEXTRACT algorithm.

Parameters

Name	Type	Caption	Length	Value range
OEPROJI *	String	Input OE project	1 - 192
FDBDIR	String	Folder of extracted feature database files	0 - 192
FILO *	String	Output file of collected TPs	1 - 192
IMGSET	String	Set of working images	0 - 192
OLEVEL	Integer	Downsample level to use	0 - 1	Default: 1
REJECT	Float	RANSAC rejection threshold	0 - 2	Default: 50.0
MINPTS	Integer	Minimum number of matched points	0 - 1	0 - Default: 10
LOCLMASK	String	Local exclusion mask	0 - 192
PAIRGEN	String	Pair-generation method	0 - 192	ALL | SEQUENTIAL Default: ALL
PAIRSEL	String	Pair-selection method	0 - 192	MODEL | ALL | LOWRES Default: MODEL
MTCHCONN	String	Match connected pairs	0 - 192	YES | NO Default: YES
SEARCHAR	String	Search area	0 - 192	ALL | OVERLAP Default: ALL

* Required parameter

Parameter descriptions

OEPROJI

The OrthoEngine project file from which to collect TPs.

FDBDIR

The path and file name of the folder containing one or more extracted feature database (FDB) files.

If you do not specify a value, the program scans the source-image folder for FDB files.

FILO

The path and file name of the text file to which to write the collected TPs.

The file can be used as input with the FBMIMPORT algorithm to import the points to an OrthoEngine project file.

IMGSET

The processing method to use to process the image pairs.

OLEVEL

The level of downsample to use.

The level must exist in the FDB file of each image.

REJECT

The rejection threshold based on random sample consensus (RANSAC), in pixels, for blunder elimination.

MINPTS

The minimum number of matched points required to consider a set of TPs between two images as acceptable.

If after RANSAC fewer points remain than the number you specify, the images are considered non-overlapping and all TPs are discarded.

LOCLMASK

Whether to apply a local mask to define the area over which valid features are found.

The mask must be defined such that it will exclude any inscribed text, markings, or both in the imagery.

For example, with airphoto imagery, "FID:MARGIN=500" bases the margin on the fiducials plus a 500-pixel margin. "EDGE:MARGIN=500" is based on the image edge plus a 500-pixel margin.

PAIRGEN

The method to use to generate image pairs.

PAIRSEL

The method to use to select image pairs.

MTCHCONN

Whether pairs of images that are already connected by tie points are matched.

Matching connected pairs can be useful after running sequential pair-generation (and matching); that is, sequential matching followed by exterior-orientation approximation of each connected set of photos can generate a basic exterior orientation (EO) and model. The basic EO and model can then be used to collect TPs across flightlines.

To match only unconnected photos, specify NO.

Additional uses of this parameter include trying to collect TPs in pairs that were unmatched previously by using other methods.

SEARCHAR

The area to search for point matching.

If the models are not very accurate, or the downsample level is coarse, ALL may produce better results. Points matched erroneously in non-overlapping areas will be eliminated using RANSAC (per the REJECT parameter).

Details

FBMTIE collects tie points (TP) by using feature-based matching. The key points (image locations) and descriptors are read from the FDB files generated by running the FBMEXTRACT algorithm.

Each key-point location is characterized by a feature descriptor, which is used to match corresponding points across images.

FBMTIE 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.

When you run FBMTIE, a text file containing the collected TPs is created. You then use the file as input with the FBMIMPORT algorithm to merge the points into multi-ray TPs and import them to an OrthoEngine project file.

Example

In the following example, TPs are collected from the input project file by using the features extracted at full resolution. All possible pairs between images are generated, but only overlapping pairs are matched, as determined by the existing models in the project.

oeproji = "airphoto.prj"
fdbdir = ""
filo = "airphoto_tps.txt"
imgset = ""
olevel = 1
reject = 50.0
minpts = 10
loclmask = ""
pairgen = "ALL"
pairsel = "MODEL"
mtchconn = "NO"
searchar = "ALL"

run fbmtie