GCP Collection module

Name	Caption
Send Email	Email notification settings
Input Scenes	Input scene folder
Output Folder	Output folder
Overwrite Results	Overwrite existing results
Source Background Type	Source background type
Source Background Value	Source background pixel value
Collect GCPs	Whether to collect GCPs
Master Matching Channel	Master matching channel
Math Model	Input math model
RF Math Model Order	RPC adjustment order
DEM Source	DEM tile source
Reference Images	Input reference images folder
Select Reference by Resolution	Select reference images by resolution
Reference Channel for Matching	Reference image channel to use for matching
Sampling Method	GCP sampling method
GCP Samples	Number of GCP samples to collect
Matching Algorithm	Matching algorithm
Collection Strategy	Number of passes for GCP collection
Search Radius	Search radius (pixels)
Minimum Score	Minimum score (percentage)
GCP Priority	Prioritize GCPs for processing
Chip Database File	Input chip database file
Chip Channels	Input chip database channels
Chip Matching Algorithm	Chip Matching algorithm
Chip Search Radius	Chip Search radius (pixels)
Chip Minimum Score	Chip minimum score (percentage)
GCP Priority	Prioritize GCPs for processing
Road Network	Input road network path
Road Width Field Name	Input road width field name
Road Width Scale & Offset	Input road width scale & offset
GCP Priority	Prioritize GCPs for processing
Polygon Vector File	Input polygon vector file
GCP Priority	Prioritize GCPs for processing
GCP Text Files	Input GCP text file or files
GCP Text File Format	File format of GCP text file
GCP Priority	Prioritize GCPs for processing
Water Mask File	Input water-mask file
Prune GCPs	Whether to prune GCPs
Prune GCPs Automatically	Prune GCPs automatically
GCP Retention Percentage	Percentage of GCPs to keep
Refine GCPs	Whether to refine GCPs
Rejection Method	GCP rejection method
Rejection Method Thresholds	GCP rejection method thresholds
Maximum GCPs	Maximum GCPs to accept
Minimum GCPs	Minimum GCPs to accept
Refine GCPs (Block)	Whether to refine GCPs
Rejection Method (Block)	GCP rejection method
Residual Threshold (Block)	Residual Threshold
Residual Threshold Type (Block)	Residual Threshold Type
Maximum Residual Per GCP	Maximum Residual Per GCP
Minimum GCPs Per Image	Minimum GCPs Per Image

Parameter descriptions

Send Email

If necessary, you can set up CATALYST Enterprise to send an email notification on job start and job completion.

With this check box selected, an email message is sent to each address specified in the Email Addresses box after the job starts and on completion.

You can specify one or more addresses, and each must be separated by a comma or a semi-colon. The email address of the user currently logged in displays by default.

Input Scenes

The path and name of a folder containing a valid CATALYST Professional OrthoEngine project file (oe.prj) or valid GDB-supported images.

This parameter can be specified by using any of the following:

Path and name of a folder of input images; for example, C:\data
Path and name of a folder containing an OrthoEngine project file; for example, C:\data\data-ingest_RAW_PAN.prj
Path and name of a folder with a wildcard to filter the imagery by file name or type; for example, C:\data\*_PAN.pix
Path and name of a folder path with a wildcard to filter by OrthoEngine project file; for example, C:\data\*_PAN.prj
You can also specify an MFILE as input. An MFILE is a text file with an .mfile file name extension. For more information on this type of file, see About the MFILE format.

Output Folder

The path and name of the folder to which to write the output files.

Overwrite Results

Select this check box to overwrite the existing output files, if any exist. If this check box is left clear, and an output file exists in the relevant folder, the status of the job displays a message informing you of the existence and name of the output file. The message is also written to the event log of the job.

Source Background Type

The method to use to determine which pixels in the source image to process as background (NoData) pixels. In general, if a pixel is considered NoData, the module processes it in a specific manner.

If the Any option or the All option is selected, a value must be specified for the Source Background Value parameter.

Available options are:

File Metadata, else None: reads the NoData value from the input-file metadata. The module first checks for the file-level metadata tag NO_DATA_VALUE in the source raster. If the tag is present, this value is used as a default for all channels in the file. Next, the module checks for channel-level NoData tags; if one is found, the channel-level value overrides the file-level value for that channel.

If there are channel-level NoData tags, but no file-level tag, a pixel is considered as NoData if each of the channels with a NoData tag corresponds to its NoData value. In this case, channels without a NoData tag are ignored when identifying background pixels.

If the file does not contain NoData tags, all pixels in the source image are considered valid.
None: all pixels in the source image are to be considered valid
All: if the pixel value in all channels matches the values provided for the Source Background Value parameter, the pixel is considered background (NoData).
Any: if the pixel value in any channel matches the value provided for the Source Background Value parameter, the pixel is considered background (NoData).

For specific examples, see the Source Background Value parameter description.

Source Background Value

The source background value or values when the Source Background Type parameter is set to:

The source background value is provided as either a single number (applied to all channels) or as a pixel "stack" (a comma-delimited list of values). If a pixel stack is provided, but the number of values does not equal the number of channels, the list is truncated or the last value is repeated as necessary. The background values provided is truncated to the range allowed by the source image data type.

The following examples apply to a 3-channel, 8-bit unsigned image:

Source Background Type set to All and Source Background Value set to 0: a pixel is considered background if all three channels are zero.
Source Background Type set to Any and Source Background Value set to 0: a pixel is considered background if any of the three channels is zero
Source Background Type set to All and Source Background Value set to 128,0,0: a pixel is considered background if its value is exactly 128 for the first channel and zero for the second and third channels
Source Background Type set to All and Source Background Value set to 128,0,0,245: a pixel is considered background if its value is exactly 128 for the first channel and zero for the second and third channels. The value 245 is ignored.
Source Background Type set to Any and Source Background Value set to 128,0: a pixel is considered background if the value in channel 1 is 128 OR if the value in channel 2 or 3 is zero.
Source Background Type set to Any and Source Background Value set to 1032,0: a pixel is considered background if the value in channel 1 is 255 OR if the value in channel 2 or 3 is zero. The first value is truncated to the range allowed for 8U data.

Collect GCPs

Selected by default, this check box controls whether to collect the ground control points (GCPs) from various reference sources.

Master Matching Channel

The channel from the input image to use for matching when collecting ground control points (GCPs). If no value is specified for this parameter, channel 1 is used, by default.

Math Model

The math model to use for collection of ground control points (GCPs) and, subsequently, orthorectification.

Available options are:

Rational Function Model (RPC): uses the Rational Polynomial Coefficients (RPC) that are distributed with the raw imagery. With most modern sensors, this is the most robust model to use, as the distributed RPCs are very well calibrated (accurate). Therefore, fewer GCPs are required, generally, and positional improvements are not constrained to areas within the outer bounds of the GCPs collected. Other abbreviations for the Rational Function Model include RF, RFMODEL, RFI, RF0, RF1, and RF2. This is the default value.
Toutin's Satellite Orbital Model (Rigorous): CATALYST Enterprise uses the Toutin rigorous math-model solution, which calculates the position and orientation of the sensor at the time of image capture from an orbital or ephemeris segment. When this information has been identified, it can then be used to accurately account for known distortions in the image. After the sensor orientation and position have been calculated, this information can be used to drive other processes, such as orthorectification. Other abbreviations for Toutin's Satellite Orbital Model include SAT, RIG, SRIT, and TOUTIN.
File Georeferencing Only: available with select CATALYST Enterprise modules, this option uses the file-level georeferencing information, such as the projection and the extents, to match the multispectral data to the panchromatic data. That is, if no math model exists in the input data, this option is applied regardless of which option is selected for this parameter.

The module first attempts to use the selected math model. If required information is missing from the specified math model, the module automatically tries to use another math-model option and, subsequently, a warning message is displayed.

RF Math Model Order

When using the Rational Function with satellite imagery, you can modify the RF math model to better agree with collected ground control points (GCPs) by selecting the correct RF math-model order.

Order 0: allows x,y translations
Order 1: allows x,y translations with rotation
Order 2: allows x,y translations with rotation and increased warping

Typically, performing a first-order transformation is best, except when the GCPs are not well distributed. If your GCPs are clustered together, a first-order transformation may introduce new and significant errors in the image away from the GCPs. If your GCPs are not well distributed, you will probably obtain better results with the zero-order transformation.

DEM Source

The name of a single digital elevation model (DEM) file or a folder containing one or more DEM tiles.

This parameter can be specified by using any of the following:

Name of a single DEM file; for example, C:\data\DEM\dem.pix
Name of a folder containing one or more DEM tiles and an associated index.txt file; for example, C:\data\DEM
Name of an index file; for example, C:\data\DEM\index.txt

The index.txt file lists the DEM files contained in the specified folder and provides information describing each DEM tile. The information in the DEM index file supersedes other DEM parameters in the module; all other DEM-related parameters are ignored. For more information about the format of the index.txt file and specific requirements for the individual DEM tiles, see Format of the DEM index file.

When the value of DEM Source is the name of an existing folder, the module searches that folder for a file named index.txt, and a set of DEM raster tiles. The index.txt file contains a single vector channel that lists the DEM files contained in the specified folder and provides information describing each DEM tile.

If no value is specified for this parameter, the module uses the default global DEM installed with CATALYST Enterprise (gmted2010).

Note: With some modules, the DEM Source parameter is mandatory. That is, if in the interface an asterisk appears beside the parameter label, it is mandatory and you must enter a value; otherwise, the parameter is optional.

Reference Images

The path of a single reference image or a folder containing multiple reference images to be used for automatic collection of ground control points (GCP). Alternatively, you can enter a comma-delimited list of raster-image file names.

To use multiple GDB-supported geocoded reference images for automatic GCP collection, you can specify a reference-image folder (GDB-compatible) and an associated (GDB-compatible) digital elevation model for automatic GCP collection. The specified folder can contain multiple reference images to use for collecting GCPs.

Note: This module uses all of the images in the input folder. Ensure that the specified folder contains only the images destined for use as reference for automatic GCP collection. For example, in some cases, the mosaic tile folder contains a mosaic preview file and scaled tiles; because these are not appropriate for automatic GCP collection, ensure that there are no such files in the folder specified for Reference Images.

The value of Reference Images is an image file that has been orthorectified previously for use with to automatic GCP collection. The GCPs collected from one or more of the reference images are stored in a GCP segment of the PCISDK image. The module also creates an OrthoEngine project that contains the same GCPs for quality- assurance (QA) tasks and manual editing.

When collecting GCPs using reference images, the module searches for ORTHO_X_ACCURACY and ORTHO_Y_ACCURACY metadata tags in the images. When such tags are present, the module uses those values to determine the accuracy of each GCP, thereby weighting the value of that point against others. The ORTHO_X_ACCURACY and ORTHO_Y_ACCURACY metadata tags are set in the Index PIX File Creator module. If no metadata tags are found in the reference image, the GCP accuracy is calculated to be half of the resolution of the reference image.

Select Reference by Resolution

If selected, the selection of reference imagery is based on resolution of input source imagery.

Reference Channel for Matching

The channel from reference image to use for matching when collecting ground control points (GCPs). Channel 1 is the default.

Sampling Method

The method to use to create ground control point (GCP) samples from the source imagery.

Available options are:

Stereo Candidates: uses the GCP stereo candidates created by the GCP Reference Imagery Preparation module. When this option is selected, the GCP Samples and Search Radius parameters are ignored.
Grid: sample points are generated automatically using sample points distributed evenly across the overlap region.
Susan: sample points are generated automatically using the Susan corner detection algorithm

The Grid and Susan options determine how to find the initial candidate positions in one image (the source image) for collecting sample points. The function builds a patch around each candidate position, and searches within that area for corresponding features in the overlapping images.

The Grid option creates candidates in a grid-like pattern. This option does not preprocess the image, so it tends to be faster. However, it also finds fewer matches, because the grid point might fall on a featureless flat patch somewhere in the image that cannot be matched to anything in the overlapping reference images.
The Susan option finds the candidates by running a corner-detection algorithm on the image, which looks for corner-like features to use as candidates.

When collecting GCPs, the Grid option is recommended because the Susan option finds candidates on building corners that may not be represented in the digital elevation model (DEM), leading to GCPs with higher residuals due to height errors.

GCP Samples

The maximum number of ground control points (GCPs) to collect per reference image. For example, when a raw scene overlaps four reference images and you specify the value of this parameter as 50, the maximum number of GCPs collected is four times that value for a total of 200 (50 x 4 = 200).

If you are using the Grid option, available values are:

num_samples: sample points are generated automatically using sample points distributed evenly across the overlap region. The overlap region is divided into a grid, and the candidates are placed in the center of the grid cells. num_samples defines the number of samples to use. The size of the grid cells is chosen automatically, based on the number of samples requested. This is the default value.
num_samples,margin: sample points are generated automatically using sample points distributed evenly across the overlap region. The distribution is similar to the previous Grid option, but the points are distributed starting at the edges of the overlap rather than in the center of the grid cells, which improves the chances of finding matches at the edge of the image. The margin value controls the minimum distance, in pixels, between the edge of the image and the placement of the candidates. Candidates that are too close to the image edge may not be matched properly; however, adjusting the margin value may alleviate this.

If using the Susan option, available values are:

num_samples: sample points are generated automatically using the Susan corner-detection algorithm. num_samples defines the number of samples to use.

Matching Algorithm

The algorithm to use for automated point matching.

Available options are:

Fast Fourier Transform Phase: when two images have a relative shift between them, the result is a phase difference in the frequency domain. Fast Fourier Transform Phase (FFTP) determines the shift between images using this phase difference.
Normalized Cross-Correlation: this method finds the relative shift between two images by finding the one that produces the maximum cross-correlation coefficient of the gray values in the images.

When the two images being matched have similar gray values and appearances, Normalized Cross-Correlation (NCC) generally produces acceptable results. When there is a rotation or image-size error in the initial math models, NCC may produce better matching results than FFTP. Typically, this method also generates faster results, because the template size that NCC uses is smaller than that used by FFTP.

For more consistently accurate results, FFTP is recommended. This method uses a larger template size than NCC and, because it works in the frequency domain, it looks at the patterns of details in the image rather than the gray values in a small neighborhood, which NCC uses. This makes FFTP more robust than NCC when there is a large difference in brightness between images or when a major land-use change has occurred between the images. FFTP also allows for a better match between images of the same area from different sensors or spectral bands.

Collection Strategy

Number of passes in which to collect ground control points (GCPs) from reference images.

The available options are as follows:

Automatic: determines the number of passes based on the input data.
Two Pass collects GCPs in two passes.
The first pass uses a coarse strategy, and the second uses a fine. When the input images are inaccurate, a two-pass strategy is recommended.
One Pass (Fine): collects GCPs in one pass and uses a fine matching strategy.
When the input data is geometrically close to the reference data, and for making final adjustments, a fine strategy is recommended.

While a fine strategy is the most accurate at collecting GCPs, it is slower than a coarse strategy.
One Pass (Coarse): collects GCPs in one pass and uses a coarse matching strategy.
When the accuracy of the input data is of poor quality, and to manually run two or more passes, a coarse strategy is recommended.

A coarse strategy adjusts the input data such that it is geometrically close to the reference imagery without too much overhead.

While a coarse strategy is faster than fine at collecting GCPs, it is less accurate.

Search Radius

The distance in the x and y directions from a starting location on the reference image over which the search for the best match with a fixed point on the input image is conducted.

The search radius is an estimation of error with the raw image's positional information and the digital elevation model (DEM) accuracy. If you know that your image is accurate to within 80 meters, and your DEM is accurate to within 200 meters, set the search radius to 280 meters. A larger search radius requires more processing time, because more locations are evaluated to determine the best match for a ground control point (GCP).

If this parameter is not specified, the function uses a default search radius based on the resolution of input data.

Minimum Score

The threshold value that controls whether a candidate ground control point (GCP) is accepted as a GCP or rejected. This parameter specifies the minimum-match quality that is considered acceptable, with 1.0 indicating a perfect match.

When using the Fast Fourier Transform Phase matching algorithm, this value is converted internally to a minimum acceptable phase-shift peak value.

When using the Normalized Cross-Correlation matching algorithm, this value specifies the minimum match score value required to accept a local match between the input and reference images as a GCP. The default value is 0.75.

GCP Priority

With this feature, you can set the priority of processing for one or more of your specific data aspects: your reference imagery, chip database, reference vectors, or GCPs you imported.

Each data aspect—reference imagery, chip database, reference vectors, or imported GCPs—has a GPC Priority parameter. The one with the highest selected value has the highest priority during processing. For example, if you set the GCP priority for your reference imagery to High, and the remaining aspects to Medium, the reference imagery is processed with the highest priority. Similarly, if you set the GCP priority for your reference imagery to Medium, and the remaining aspects to Low, the reference imagery is processed with the highest priority.

The Locked option retains the GCPs generated during the GCP collection process. GCP collection has two phases: collection and refinement. In the refinement phase, two separate subprocesses occur: GCP distribution and GCP refinement. During GCP distribution, the GCP priority applies. During the second phase, however, the GCP priority does not apply. Use this option when you want to retain the integrity of GCPs from a particular data aspect.

You can also set the GCP priority of more than one aspect to use the same value. That is, you can perhaps set a value of High to your reference imagery and your imported GCPs, as applicable.

Chip Database File

The path and file name of the chip database to use for automatically collecting ground control points (GCPs).

The module collects the GCPs from the chip database, and then automatically saves them to a GCP segment of the PCIDSK image associated with the GCP.

To improve the spatial accuracy of the collected GCPs, the chip database should contain high-quality image chips from a comparable sensor taken in similar conditions and be well-distributed over all of the scenes to be corrected.

Chip Channels

The number of the channels in the chip database to use for automated ground control point (GCP) matching.

Chip Matching Algorithm

The algorithm used for automated ground control point (GCP) matching.

Available options are:

Fast Fourier Transform Phase
Normalized Cross-Correlation

Chip Search Radius

The distance in the x and y directions from a starting location on the reference image over which the search for the best match with a fixed point on the input image is conducted.

The search radius is an estimation of error with the raw image's positional information and the accuracy of the digital elevation model (DEM). For example, if you know your image is accurate to within 80 meters and your DEM is accurate to within 200 meters, set the search radius to 280 meters. A larger search radius requires more processing time, due to more locations being evaluated to determine the best match for a ground control point (GCP).

If no value is specified for this parameter, the module uses the same value specified for the Search Radius parameter.

Chip Minimum Score

The threshold value that controls whether a candidate ground control point (GCP) is accepted or rejected as a GCP. This parameter specifies the minimum match quality that is considered an acceptable match, with 1.0 indicating a perfect match.

When using the Fast Fourier Transform Phase matching algorithm, this value is converted internally to a minimum acceptable phase shift peak value.

GCP Priority

With this feature, you can set the priority of processing for one or more of your specific data aspects: your reference imagery, chip database, reference vectors, or GCPs you imported.

You can also set the GCP priority of more than one aspect to use the same value. That is, you can perhaps set a value of High to your reference imagery and your imported GCPs, as applicable.

Road Network

The path to a vector file that contains the road-network layers to use for automatic collection of ground control points (GCP).

GCP collection from the road network extracts GCPs by matching an optical image with rasterized lines.

The GCPs are collected automatically by matching patches in the image with roads. Matching is performed in the spatial-frequency domain using Fast Fourier Transform (FFT) to transform and match image patches and rasterized lines. The matching algorithm is based on the Kuglin and Hines (1975) paper cited in the References section later in this topic.

In a typical application, between 100 and 200 GCPs are extracted, with most of them being correct. The GCPs can then be used in automated image-orthorectification workflows. This method is well-suited to midresolution images from 5 meters through 15 meters.

Road Width Field Name

The field name of an attribute in the vector segment. The field must contain one or two numerical values. The values are converted to a line width, in meters. Float, double, and integer fields are supported.

If this parameter is not specified, then the following field names are checked:

NBRLANES
LANES
ROADLANES
ROADS
ROADWIDTH

If none of these fields exist, then the Road Width Scale & Offset parameter defines the width of all lines in the file.

If the field does exist, and it contains a single value, then the scale factor part of the Road Width Scale & Offset parameter is used in the computation of the road width:

road width (m) = fieldValue * roadWidthScale

If the field exists, and it contains two values, then the two values in the Road Width Scale & Offset parameter are used to compute the road width:

road width (m) = fieldValue * roadWidthScale + roadWidthOffset

Road Width Scale & Offset

The scale and offset, in meters, used in conjunction with the Road Width Field Name parameter to compute the road width in meters. This parameter contains one or two values. The first value, the road width scale (roadWidthScale), must be greater than zero. If specified, the second value, the road width offset (roadWidthOffset), must be non-negative.

If no value is specified for the Road Width Field Name parameter, the road width scale defines the width of all lines (in meters) in the file and the second value is ignored.

road width (m) = roadWidthScale

If a value for the Road Width Field Name parameter is specified, the scale and offset values are used to convert attribute values in the field to road width values, in meters. For more information, see the description of the Road Width Field Name parameter.

GCP Priority

With this feature, you can set the priority of processing for one or more of your specific data aspects: your reference imagery, chip database, reference vectors, or GCPs you imported.

You can also set the GCP priority of more than one aspect to use the same value. That is, you can perhaps set a value of High to your reference imagery and your imported GCPs, as applicable.

Polygon Vector File

The path to a file containing the polygon layers to be used for automatic collection of ground control points (GCP). The path may also specify a folder containing multiple polygon files.

Note: In general, geo-morphologically stable lake and river polygon layers can be used as reference for GCP collection.

GCP Priority

With this feature, you can set the priority of processing for one or more of your specific data aspects: your reference imagery, chip database, reference vectors, or GCPs you imported.

You can also set the GCP priority of more than one aspect to use the same value. That is, you can perhaps set a value of High to your reference imagery and your imported GCPs, as applicable.

GCP Text Files

The path and file name of a text file, or a folder of text files, that contains ground control points (GCPs) from other sources.

Each text file must have the MAPUNITS parameter specified in its required format. You can also specify up to two additional parameters, ELEVREF, and ELEVUNIT. The following table shows the supported values, a description, and an example for each parameter.

Parameter	Supported values	Description	Example
`MAPUNITS`	`PIXEL`	Pixel and line	`MAPUNITS=PIXEL`
	`UTM`	Universal Transverse Mercator	`MAPUNITS=UTM 32 T D000`
	`SPCS`	State Plane Coordinate System	`MAPUNITS=SPCS`
	`LONG/LAT`	Longitude and latitude	`MAPUNITS=LONG/LAT D000`
	`METER`	Image along-row and along-column, in meters	`MAPUNITS=METER`
	`FEET`	Image along-row and along-column, in feet	`MAPUNITS=FEET`
`ELEVREF`	`MSL`	Mean sea level	`ELEVREF=MSL`
	`ELLIPS`	Ellipsoid	`ELEVREF=ELLIPS`
`ELEVUNIT`	`METER`	Meters	`ELEVUNIT=METER`
	`US_FEET`	U.S. feet	`ELEVUNIT=US_FEET`
	`FEET`	Feet	`ELEVUNIT=FEET`

When the path points to a folder, each text file must follow a naming convention, <filename>*GCP.txt, where <filename> is the file name of the ingested image.

For example:

Name of the raw image file: image1_RAW_PAN.pix
Name of the GCP text file: image1_RAW_PANGCP.txt

GCP Text File Format

The format of the file or files specified for the GCP Text File parameter, if specified.

Note: Your GCPs must be the same units as the map units (MAPUNITS) in the text file you specify.

The available formats, including a link to an example of each, are as follows.

Format	Example
2D: ID P L X Y S
2DERR: ID P L X Y eP eL eX eY S
3D: ID P L X Y Z S
3DERR: ID P L X Y Z eP eL eX eY eZ S
Bulk GCP File

Note: The standard formats support only points from a single image in each text file. The Bulk GCP File format includes all GCPs from a group of images in a single text file.

GCP Priority

With this feature, you can set the priority of processing for one or more of your specific data aspects: your reference imagery, chip database, reference vectors, or GCPs you imported.

You can also set the GCP priority of more than one aspect to use the same value. That is, you can perhaps set a value of High to your reference imagery and your imported GCPs, as applicable.

Water Mask File

The path to a file that contains the polygon water-mask layer to be used for refinement of ground control points (GCPs). The path can also specify a folder that contains multiple water-mask files.

Prune GCPs

Selected by default, this check box controls whether to prune the ground control points (GCPs).

Prune GCPs Automatically

Selected by default, this check box controls whether to automatically prune the ground control points (GCPs) spatially. Pruning helps to achieve a complete coverage of image extents by GCPs, without overweighting any particular area with too many points. The spatially uniform GCP coverage ensures that the math model refined with the points is stable and accurate.

You can use this parameter in conjunction with GCP Retention Percentage to set how to prune the GCPs.

GCP Retention Percentage

Available when the Prune GCPs Automatically check box is unselected, the entered value determines the percentage of GCPs to be retained. The process adapts to the density of GCPs in different parts of the image.

GCP Retention Percentage is a value between 1 and 100, with larger values retaining a higher fraction of available GCPs. Blunder and duplicate GCPs are always eliminated; therefore, even with 100 percent requested, some GCPs may be eliminated.

Refine GCPs

Selected by default, this check box controls whether to refine the ground control points (GCP). Refinement is to systematically eliminate GCPs that have large errors. To retain the integrity of the GCPs you have imported or otherwise referenced in a text file associated with the project, clear the Refine GCPs check box.

Rejection Method

The method used to reject ground control points (GCP).

Available options are:

Automatic: selects the most appropriate rejection method, based on the sensor of the input scene
RMS Error: RMS error-rejection method, measured in pixels
Standard Deviation: sigma (standard deviation) rejection method, measured in pixels
Percentage: percentage of points that can be rejected, starting with the one with the highest residual
Absolute Distance: absolute distance rejection, in terms of the residuals, measured in pixels
Absolute Number: absolute number of points to reject, starting with the ones with the highest residual

You can specify various values for this parameter, depending on the method selected.

Rejection Method Thresholds

The rejection threshold values for the value selected for the Rejection Method parameter.

This parameter is defined using two values (THRESH1, THRESH2); these differ in meaning depending on the selected rejection method:

Automatic: selects the most appropriate rejection method thresholds, based on the sensor of the input scene
RMS Error: rejection starts from the point with the largest residual error for any point, then recalculates the math model and RMS error. If the RMS error is still above the specified thresholds, the point with the next highest residual is removed and so on, until the x-RMS and y-RMS errors are equal to or less than THRESH1 pixels or THRESH2 pixels.

For example, a value of 2,2 rejects points with the highest residuals until the x-RMS and y-RMS are both less than two pixels.
Standard Deviation: THRESH1 and THRESH2 represent the minimum standard deviation values of the x and y residuals to be rejected, respectively.

For example, a value of 2,1 rejects points that have a standard deviation higher than two of the resX mean, and rejects points that have a standard deviation higher than one of the resY mean.
Percentage: THRESH1 represents the percentage of the number of points to be rejected and THRESH2 represents the ratio weighing between the x and y residuals. For example, if you set a rejection weight of 2, you are giving twice the weight to the x-residual (resX) as to the y-residual (resY). By default, the residual in x and y have the same weight. Therefore, if you have a point with a resX of 0.4 and a resY of 0.5, the point is given a resX of 0.8 and a resY of 0.5 for the rejection.

For example, a value of 5, 2 rejects five percent of GCPs, and gives the x-residual (resX) twice the weight as that of the y-residual (resY).
Absolute Distance: THRESH1 and THRESH2 represent the minimum absolute x and y pixel residuals to be rejected. The rejection starts from the point with the largest x or y residual distance.

For example, a value of 2,2 rejects points with a resX of greater than two pixels, and rejects points with a resY of greater than two pixels.
Absolute Number: THRESH1 represents the number of points to be rejected and THRESH2 represents the ratio weighing between the x and y residuals. For example, if you set a rejection weight of 2, you are giving twice the weight to the x-residual (resX) as to the y-residual (resY). By default, the residual in x and y have the same weight. Therefore, if you have a point with a resX of 0.4 and a resY of 0.5, the point is given a resX of 0.8 and a resY of 0.5 for the rejection.

For example, a value of 10, 0.5 rejects 10 GCPs, and gives the x-residual (resX) half the weight of the y-residual (resY).

Maximum GCPs

The maximum number of ground control points (GCPs) to accept.

After the module performs an initial collection of GCPs using the reference data, it refines the collection to ensure that only the most accurate points are retained.

If there are more GCPs collected than the specified maximum value, the module performs a second refinement, keeping only the GCPs with the lowest RMS error, up to the specified maximum number of GCPs. The second refinement uses the following rejection method:

RMS Error: for the second refinement, the rejection threshold value is set to the specified Maximum GCPs value.

Minimum GCPs

The minimum number of ground control points (GCPs) to accept.

After the module performs an initial collection of GCPs using the reference data, it refines the GCPs, and then verifies that the number of remaining GCPs is greater than or equal to the minimum number of GCPs.

When there are fewer GCPs remaining than the specified minimum value, the module performs a second refinement of the original GCPs, using less stringent rejection-threshold values. These values, specified for the Rejection Method Threshold parameter, differ based on the value selected for the Rejection Method parameter:

RMS Error: for the second refinement, both rejection threshold values (THRESH1, THRESH2) are quadrupled.

For example, an initial threshold of 1,1 is relaxed to 4,4.
Standard Deviation: for the second refinement, both rejection threshold values (THRESH1, THRESH2) are quadrupled.

For example, an initial threshold of 2,1 is relaxed to 8,4.
Percentage: for the second refinement, the first rejection threshold value (THRESH1) is divided in half; the second (THRESH2) is left as is.

For example, an initial threshold of 10,2 is modified to 5,2.
Absolute Distance: for the second refinement, both rejection threshold values (THRESH1, THRESH2) are quadrupled.

For example, an initial threshold of 2,2 is relaxed to 8,8.
Absolute Number for the second refinement, the first rejection threshold value (THRESH1) is doubled; the second (THRESH2) is left as is.

For example, an initial threshold of 20,0.5 is modified to 40,0.5.

If there are still too few GCPs after the second refinement, the module aborts and displays an error message.

Refine GCPs (Block)

Selected by default, this check box controls whether to refine the ground control points (GCP) on the final block of images/GCPs. Refinement is to systematically eliminate GCPs that have large errors. To retain the integrity of the GCPs you have imported or otherwise referenced in a text file associated with the project, clear the Refine GCPs check box.

Rejection Method (Block)

The method used to reject GCPs.

Available options are:

Progressive:Residuals greater than 1,000, 500, 100, 20, and 10 pixels are removed iteratively before applying the Block Residual Threshold
Direct:Residuals greater than the value specified by Block Residual Threshold are removed immediately

Residual Threshold (Block)

The job removes GCPs until the Block Residual Threshold is reached. The method in which GCPs are removed is defined by the Rejection Method (Block).

Residual Threshold Type (Block)

The statistic used to reject GCPs.

Available options are:

Standard Deviation: sigma (standard deviation) rejection method, measured in pixels
Absolute Distance: absolute distance rejection, in terms of the residuals, measured in pixels

Maximum Residual Per GCP

The maximum residual a GCP can have after refinement.

Minimum GCPs Per Image

Refinement stops removing GCPs from an image if the image does not contain the specified amount of GCPs.

Details

General job details

Preprocessing requirements

Before running this module, the following requirements must be met to ensure the job processes successfully and produces accurate results:

Reference imagery: with this module, you can automatically collect ground control points (GCP) from reference imagery for the raw imagery being imported. To use reference imagery for automated GCP collection, ensure that the reference imagery provided is accurately georeferenced and in a GDB-compatible format. For high-quality GCP collection, good practice dictates that the reference images have the highest level of accuracy possible, and be within a reasonable difference in spatial resolution to the raw imagery for which GCPs will be collected. While this requirement is not mandatory, it is important to make note of, to ensure a higher degree of accuracy in math-model refinement.

Note: To improve performance, ensure that the reference imagery contains overviews (pyramids) by running the Pyramid module on all reference images prior to collecting GCPs.
Digital elevation models: To use reference imagery in this module, you must supply a digital elevation model (DEM). The DEM must coincide with the area of interest (AOI) and the reference imagery, and be in a GDB-compatible format. DEMs can be provided as a single file, or as multiple files or tiles covering the AOI.
Chip database: with the GCP Collection module, you can perform automated GCP collection from reference chip databases. A chip database is a compilation of individual image samples, called chips, usually measuring 256 pixels by 256 pixels or smaller. Each image section contains an accurate geocoded location and metadata, such as the sensor from which it was generated, the date it was acquired, the viewing angle, and so on.

When working with a chip database, it is important to understand the limitations of automatic GCP collection from a chip database, with regards to difference in the spatial resolution between the raw input scenes and the chip database image chips. If there is a difference of more than a magnitude of 5 times, the results from automatic GCP collection from a chip database may be marginal. The chip database must be supplied in the PCI Geomatics chip-database format.
Polygon-vector files: Polygon-vector files can be supplied for automated GCP collection. These can be in a single file or in multiple files covering the region of interest. The polygon-vector file or files should be provided in a GDB-compatible format.
Road-network file: You can provide a vector-road-network file (line) for automated GCP collection. This reference file can be used to automatically collect GCPs for roads in the AOI. The road-network file must be supplied in a GDB-compatible format.
Water-mask file: GCPs collected automatically by running this module can be refined using a water-mask file. This ensures that GCPs accidentally collected on a body of water are rejected. The water-mask file must be supplied in a GDB-compatible format.

Module details

The GCP Collection module, automatically collects GCPs from a variety of resources, such as geocoded images, vectors, and chip databases. This is an optional step in the CATALYST Enterprise workflow for satellite and aerial imagery. A multitude of actions take place, including:

Refining the math model with the GCPs collected from the reference data sources
Providing quality-assurance (QA) metadata for performing QA procedures on the data before proceeding in the CATALYST Enterprise workflow

With satellite imagery, the module supports all valid GDB-format images, whether or not they contain a math model. If no math model or orbit segment can be found in the given image, GCPs will be collected, and a polynomial math model will be generated from the GCPs.

With aerial imagery, the module supprts only, a valid CATALYST Professional OrthoEngine project file (oe.prj) containing exterior orientation (EO) and the camera model. When a valid OrthoEngine project file is unavailable, no GCPs are collected.

The GCP Collection module applies a series of predefined configuration files that define characteristics of the GCP-collection process and the type of output to create. These are defined in the settings.py file located in the PROHOME\exe\PGS\config folder of your CATALYST Enterprise installation.

About GCP collection on road vectors

This module extracts ground control points (GCPs) by matching an optical or SAR image with rasterized lines.

The module is designed to extract GCPs from road vectors in medium-resolution images with resolutions from about 4 meters through 15 meters. It models each road as a ribbon with a sinc-like cross-section. The ribbon is either dark on a uniform bright background, or bright on a dark background. The width of the ribbon is either set explicitly or derived from attributes of each road segment, such as the number of traffic lanes.

This model works very well for mid-resolution images, such as SPOT and Landsat 7 PAN, and provides GCPs accurate to about 1 pixel, or within 5 meters through 10 meters. PCI extended to higher-resolution images by averaging the images down to the matching resolution of 3 meters through 6 meters. The matcher can also be directed to try both bright-on-dark or dark-on-bright matches at every point, and then select whichever one is better.

The enhanced process extracts a sufficient number of GCPs, but their accuracy is still at the 5-meter through 10-meter level. This is due to several factors:

Inherent accuracy of road vectors

This factor alone limits the achievable accuracy of GCPs extracted from road vectors. Canadian experience shows that most road vectors have an accuracy of about 5 meter through 10 meter; for example, the current Canadian national road network claims an accuracy of 6.5 meters. In addition, many road vectors are less accurate, or even in disagreement with the images, due to rapid land-use changes in many areas. The road GCP module can even cope with large inaccuracies, or fail for such GCPs, but the overall achievable accuracy is limited.
Road model used

The "smooth-ribbon" model is well suited to mid-resolution images, particularly in non-urban areas. However, it becomes inadequate at higher resolutions due to other details in the images, such as vehicles, street furniture, road markings, vegetation, and building and tree shadows. Objects such as these introduce significant clutter, and lower the overall accuracy of the matches. These matches are often rejected. If they are not rejected, the matcher settles on nearby, more uniform linear objects, such as bright rooftops of large buildings or their shadows.
Varied appearance of roads

The roads may not be uniformly brighter or darker than their surroundings, as the model assumes. Their appearance depends on the spectral bands of the matched image, type and age of the road-surface material, season (snow versus no snow), and even the moisture content. In many images, roads are mid-gray, and, in such a case, are not handled well by the tool.
Significant building lean in urban scenes

High-resolution satellites fly at lower altitudes and, therefore, introduce a higher degree of building lean. Some images can also be taken in off-nadir orientation, which magnifies this effect even more. In urban areas, leaning buildings often obscure streets, at which point the matcher either fails or matches incorrect features.
Elevated highways

In most instances, the elevations of GCPs are extracted from a bare-earth digital elevation model (DEM), while some of the GCPs can be extracted from elevated highways, bridges, or multilevel highway intersections or interchanges. Such features are matched correctly in the image space, but have incorrect (ground) elevations, which leads to biased GCPs and an inaccurate refined model.

All of the preceding factors contribute to the lowered accuracy of road GCPs in high-resolution images, and limit their accuracy to the 5-meter through 10-meter range. They also introduce a possibility of systematic biases, particularly along the viewing direction of the satellite.

Road GCPs extracted from high-resolution images are still useful if the biases in the original (nominal) model of the scene exceed 10 meters through 5 meters. However, if the nominal model is already accurate to greater than 10 meters, the model refined with the road GCPs may be less accurate than the nominal one.

Images with resolutions outside the optimum range will be processed, but the results may be unsatisfactory: few extracted GCPs for low-resolution images, and many incorrect matches for high-resolution images. The module issues a warning for images with pixel sizes less than 1.9 meters.

Naming convention for GCP IDs

Each GCP collected has an ID. The naming convention of the ID is described in the following table, where the GCP ID is: XXXXXXXX_YYYY_N.

GCP ID element	Description
XXXXXXXX	Hash code of the input image
YYYY	Hash code of the reference data
N	Number of the GCP, one to any number

About GCP collection on polygon vectors

This module extracts ground control points (GCPs) by matching an optical or SAR image with rasterized polygons.

This module also automatically collects GCPs by matching patches in the image with polygons, usually lakes. The matching is performed in the spatial frequency domain using the Fast Fourier Transform (FFT) method to transform and match image patches and rasterized polygons. The matching algorithm is based on the Kuglin and Hines (1975) paper cited in the References section. The number of GCPs extracted depends on the number of polygons in the area used for GCP extraction. The GCPs can then be used in automated image orthorectification workflows.

Job results

For each scene found in the Input Scenes folder, the GCP Collection module creates a linked PCIDSK file.

With satellite imagery, the name of each PCIDSK file is system-generated with the scene ID, a state (includes a _RAW tag in the file name), and a type, such as _PAN, _MS, or _PSH. In addition to the PCIDSK files, a CATALYST Professional OrthoEngine project file is created for each scene type.

When the Input Scenes folder contains a data-ingest_RAW_PAN.prj file, GCPs are collected only on panchromatic scenes, and the GCP Collection module creates only one new OrthoEngine project file: gcp-collection_RAW_PAN.prj.

Should there be any additional OrthoEngine project files in the Input Scenes folder, they are simply copied to the output folder.

When a data-ingest_RAW_PAN.prj file does not exist in the Input Scenes folder, the GCP Collection module collects GCPs on all input OrthoEngine project files, and creates a gcp-collection.prj file for each.

With aerial imagery, the module creates, in addition to the PCIDSK file, an OrthoEngine project file of all the GCPs collected.

For each scene containing collected GCPs, the number of GCP segments created depends on various factors, such as GCP refinement, the number of GCPs collected, and the distribution of the GCPs, among others. With satellite imagery, in most cases, each scene contains at least two model segments (either a rational function model or a satellite model). The first comes from the raw data itself, and the second is created after collecting the GCPs. A new model can then be computed based on the GCPs.

When processing satellite imagery, the GCP Collection module produces:

gcp-collection_RAW_<TYPE>.prj: OrthoEngine project file for each scene type for which GCPs were collected. The module only collects GCPs on panchromatic data when a data-ingest_RAW_PAN.prj file exists in the Input Scenes folder.
data-ingest_RAW_<TYPE>.prj: OrthoEngine project files when no GCPs were collected. The module only collects GCPs on panchromatic data when a data-ingest_RAW_PAN.prj file exists in the Input Scenes folder.
Linked PCIDSK file of the multispectral image bands (when the Input Scenes folder contains raw multispectral data)
Linked PCIDSK file of the panchromatic image band (when the Input Scenes folder contains raw panchromatic data)
Linked PCIDSK file of the pansharpened multispectral bands (when the Input Scenes folder contains raw pansharpened data)
Linked PCIDSK file of the SAR image bands (when the Input Scenes folder contains raw SAR data)
Linked PCIDSK file of any valid raster channels other than the preceding, if existing
PCIDSK file of the footprint of each image
PCIDSK file in MFILE format (.mfile) of each image
Text file of information on the GCPs collected for each scene
Text file summarizing all GCPs collected, including residuals and error information
Text file summarizing all exported math models

When processing aerial-triangulation imagery, the GCP Collection module produces:

gcp-collection_RAW_MS.prj: OrthoEngine project file of all images and collected GCPs
Linked PCIDSK file to the aerial image
PCIDSK file in MFILE format (.mfile) of all images
Text file of information on the GCPs collected for each aerial scene
Text file summarizing all GCPs collected, including residuals and error information
Text file summarizing all exported math models

Note: The output folder also contains a gcp-collection_RAW_<TYPE>.prj.refhash.txt file, which contains a list of hash values of reference data.

The files are created from all of the data in the output folder each time you run the GCP Collection module (even when the Overwrite Results check box is clear).

Important: The summary files retain data from the previous GCP-collection task. That is, if you delete one or more of your input scenes, but not the corresponding output data, the summary files will still contain information on the deleted scenes when you run the module. To avoid having data from a previous GCP-collection task in your summary files, first remove them from your output folder.

You can now open the scenes in CATALYST Professional Focus or open the OrthoEngine project file in OrthoEngine to perform quality assurance (QA) tasks, or proceed to use the next module in the CATALYST Enterprise workflow.

After successfully running this module, you will generally proceed to use one or more of the following modules:

Tie-Point Collection and Refinement module: for tie-point collection, to improve relative accuracy between overlapping scenes
DEM Extraction module: uses the improved math model of stereo images to extract a digital surface model (DSM) or digital terrain model (DTM)
Orthorectification module: uses the updated math model to orthorectify a set of images

GCP Collection module

Description

Parameters

Parameter descriptions

Details

References