Understanding automatic tie-point collection

A tie point (TP) is a specific location that is recognizable visually in the overlap area between two or more images. The image positions of TPs on the overlap areas can be identified and measured both manually and automatically. CATALYST Professional OrthoEngine technology can automate the collection of TPs by using image-correlation techniques.

To create TP samples from the source image, you use either the Grid or Susan method. To match points on the target images, you use the Fast Fourier Transfer Phase (FFTP) or Normalized Cross-Correlation (NCC) algorithm. For example, the NCC algorithm computes the correlation coefficient of the gray values between the template window and the search window, and determines the correspondence between two image areas according to the similarity of the gray-level values.

The matching process can be accomplished when your project meets one or more of the following criteria:

The exterior orientation of each image was computed based on ground control points (GCPs), TPs, or both.
You collected three TPs between each pair of overlapping images.
You used a Global Positioning System (GPS) to obtain the x, y, and z coordinates for each image center, and you estimated the omega, phi, and kappa rotations, or they were supplied by an Inertial Navigation System (INS).
You used the Satellite Orbital math model and imported the ephemeris or orbit information with the satellite data.
You imported the polynomial coefficients from the satellite data for use with the Rational Functions math model.

Image correlation and search radius

Image correlation uses a hierarchical approach to find matching features in the overlapping area between two or more images using a moving window with a search radius of 100 pixels, by default. The first attempt at correlation is performed on a high-level pyramid of the images. Depending on the resolution of the images or the accuracy of the math model, the predictability of the match can be greater than the size of the search window. For example, if your image is a resolution of 0.10 meters, and your estimated math model is off by 15 meters, then the features that you are trying to match are 150 pixels away from their estimated locations. If you are experiencing a low success rate with the CATALYST Professional automatic tie-point collection, increasing the search radius may improve your results. Increasing the search radius, however, increases the processing time.

Matching candidate points in the source image will be projected to the ground, and then reprojected to the target image. You must have a good initial position for the two correlation windows. With an aerial project, a good initial position can be determined when the exterior orientation (EO) parameters of the images are accurate.

The following factors can affect automatic collection of TPs:

If the contrast of the images is very low or very high, automatic collection of TPs may not be able to determine the correlation. Make sure the images have an appropriate level of contrast.
If the search radius is too small, too few TPs may be collected, and if the initial position information is inaccurate, or there is a large amount of terrain relief, the TPs may be distributed unevenly. Be sure to use a search window of an appropriate size. A larger search window will increase processing time, however.
Tip: To quickly determine an appropriate search radius, generate two coarse orthorectified images by using an average constant-elevation value, and then determine the greatest offset for the same feature point in the overlapping images. You can use the offset, plus a certain amount of buffer, as a reference for the search radius to use in collecting TPs.

The search radius specifies the distance from a starting location on the target image over which the search for the best match with a fixed point on the source image will be conducted.

The search radius is also an estimation of error with the positional information of the raw image and the accuracy of the DEM. If you know that your images are accurate to 80 meters, and your DEM is accurate to 200 meters, set the search radius to 280 meters. A larger search radius—for example, 300 or greater is recommended for hilly or mountainous terrain—will require more processing time, because more locations are evaluated to determine the best match for a TP.