Collecting ground control points automatically

OrthoEngine can use image correlation to identify the pixel and line locations in the raw image that correspond to features in the georeferenced image or to the georeferenced positions on the chips. It is this image correlation that OrthoEngine uses in automatic collection of ground control points (GCP).

Note: When working with radar data, and the input data contains image layers written as complex values, the total power in decibels is computed temporarily and used for matching. When possible, it is also recommended that you calibrate the data in sigma, beta or gamma naught. FFTP matching is also recommended when working with SAR data.

To collect GCPs automatically

  1. Click the Control source list, and then click one of the following:
    • Georeferenced image: to use a georeferenced image as the source for the GCPs
    • Chip database: to use a chip database as the source for the GCPs
  2. In the Reference file box, type the path and file name of the image or chip database or click Browse to select the file.
  3. If you selected a chip database as your source for GCPs, you can click Chip Criteria to specify the sensor and the image's acquisition date.
  4. If you selected a georeferenced image as your source for GCPs, enter the following:
    • In the Match channels box, type the number representing the channel in the georeferenced image that you want to use for the image correlation. To specify multiple channels, use a comma-delimited list.
    • To use a digital elevation model (DEM) to automatically supply the elevation values of the GCPs, type the path and file name of the DEM in the DEM file box, or click Browse to select the file. To define the DEM-file options, click DEM Settings. For more information on the options, see Setting DEM file options.
  5. Under Images, in the Use column of the table, click to select or clear the check marks to choose the images you want to use in the matching process. Only images with check marks in the Use column will be used. To select all of the images, click Use All. To clear all selections, click Use None.
  6. In the Match Channel box, type the number of the channel from the raw image that has the same color band or wavelength as the source so that features on the ground will look the same in both images. Use a channel other than blue, because the blue band tends to saturate and it may not be as sharp as some of the other bands.
    • To use the same channel for all the raw images, type the number of the channel that you want to use for the image correlation, and then click the Apply to all images check box.

      To specify multiple channels, use a comma-delimited list.

    • To select a different channel for each image, in the Image ID column, click the image you want, and then in the Match channel box, type the number of the channel that you want to use for the image correlation. Repeat for each image.
  7. If stereo GCPs are required, select the Stereo GCPs check box, and then select a Reference image sampling method:
    • To sample a target number of points you specify, click Count, and then in the box to the right, type the number of target GCPs you want.
    • To sample points at a spacing interval you specify, click Spacing, and then in the box to the right, type a number corresponding to the spacing you want. From the list to the right, select a unit of measurement.
  8. Under Search Options, do the following, as applicable:
    • Beside Sample source method, select an option:
      • Susan: sample points are generated automatically using the SUSAN corner-detection algorithm
      • Grid: (default) sample points are generated automatically using sample points distributed evenly across the overlap region
      The Susan and Grid options determine how to find the initial candidate positions in one image (the source image) for collecting GCPs. The function builds a patch around each candidate position that it searches for in overlapping 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.
      • The Grid option creates candidates in a grid-like pattern. Because this option does not preprocess the image, it tends to be faster, but 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 images.

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

    • In the Number of GCPs per image box, type the number of points that you want to collect over each image. Points are collected in an evenly spaced pattern over the image, similar to the tie point collection pattern on aerial photographs.
    • Click the Matching method list, and then click one of the following:
      • FFTP (Fast Fourier Transform Phase matching): when two images have a relative shift between them, the result is a phase difference in the Fourier domain. FFTP determines the shift between images using this phase difference.
      • NCC (Normalized Cross Correlation): this method finds the relative shift between two images by finding the shift that produces the maximum cross-correlation coefficient of the gray values in the images. This is the default method.

      If no value is specified for this parameter, the default method, NCC, is used.

      When the two images being matched have similar gray values and appearances, 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. Because the template size that NCC uses is smaller than the one used by FFTP, this method also typically generates faster results.

      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 in cases where there is a large brightness difference between images or when a major land-use change has occurred between the images and allows it to better match images of the same area from different sensors or spectral bands.

    • In the Search radius box, type a number to define the search radius around the kernel.

      Click the list to the right, and then click an appropriate unit for the search-radius value:
      • Pixels (default)
      • Meters
      • Feet
      • US_Feet

      For example, to enter 100 pixels, type 100, and then select PIXEL.

    • In the Minimum acceptance score box, type a threshold value that controls whether a candidate GCP is accepted as a GCP or rejected. This parameter specifies the minimum match quality that is considered an acceptable match, with 1.0 indicating a perfect match. When using the FFTP algorithm, this value is converted internally to a minimum acceptable phase-shift-peak value. When using the NCC matching method, 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.
  9. Click Match GCPs.
  10. To reject matches, under Collected GCPs, click one or more entries in the Use as GCP column of the table to clear the check marks, as necessary. You can also click Use All to select all the matches or click Use None to clear all of them.
  11. To view the residual errors for the GCPs collected automatically, click Compute Model. OrthoEngine calculates a temporary model using any existing data in the project file plus the newly collected GCPs, but the model is not saved in the project file. In the Collected GCPs table, scroll to the right to view additional columns, including residual errors.
  12. To view how the selected GCPs are dispersed over the images, click Distribution. The image footprints appear as gray lines, and the GCPs appear as blue crosshairs.
  13. To accept the matches identified with check marks as GCPs in your project, click Add GCPs to Project.
  14. To save a record of all matched GCPs to a text file, click Print to File.

© PCI Geomatics Enterprises, Inc.®, 2024. All rights reserved.