The parameters of the line-of-sight model can be adjusted to fit a set of ground control points (GCPs). A GCP consists of a 3-dimensional ground coordinate (x, y, h) and its estimated accuracy (sx, sy, sh). An image point (IP) consists of a position in the satellite image (column, row) and its estimated accuracy (scolumn, srow). Marking the position of a GCP in the satellite image consequently gives a constraint on the model parameters. A number of such measurements give rise to a system of constraint equations that can be solved by least-squares adjustment.

An optimal least-squares solution should employ the “generalised least squares” approach. This makes full use of all a priori information, including start values for the parameters and their estimated accuracies. Details on how the generalised least-squares method is applied to the satellite line-of-sight problem are found in [Westin, 1990] and [Westin, 1991].

Tie-points (TPs) are image measurements that connect locations in different images. In the “generalised least squares” approach, they are treated as IPs with small (zero) variance measured for GCPs with high (infinite) variance. In this way TPs can conveniently be accommodated in the same generalised least-squares solution.

The positioning of GCPs in an image (IP measurement) can be achieved using different tools:

• Manual pointing in an image viewer. Here it must be simple to colour balance the image so that the GCP feature is easy to identify. It is also important that the image can be properly zoomed so that the sub-pixel position of the feature can be measured.
• Automated pointing using image chips. The chip is a small section from another image, extracted around the GCP. The position of the chip in the raw image is found by normalised cross-correlation of image pixel grey levels. It is necessary to correctly re-sample in order to achieve a common geometric space for the chip and raw image. It must also be possible to estimate the sub-pixel location of the best fit. In the case of ground control, the chip is extracted from a previously orthorectified image, while in the case of tie-points, the chip is extracted from the raw image to which it should be tied. It must be possible to handle both these types of chips.