In order to efficiently use imagery for mapping, geopositioning, interpretation, DEM / feature extraction, and visualization, it must be "georeferenced" along with other related images, and data. That is we must establish a relationship between each pixel in the image and its locus in the object space. Determining the mathematical description of these geometric relationships is called sensor modeling and platform modeling. Determining the actual numerical values of the parameters of the models is done through a process called triangulation, block adjustment, strip adjustment, block triangulation, aerial triangulation, "AT", etc. Other, related terms include registration, georegistration, georeferencing, and restitution. Rectification (as you will remember) is a special case of georeferencing in which we modify the geometry of the image to make the relation between pixel and object locus simpler.
Block triangulation can be: small (single image), large (thousands of images), homogeneous (same sensor, same scale, same acquisition time) or heterogeneous (different sensors, different scales, different acquisition times), vertical, oblique, convergent, close-range, terrestrial, airborne, satellite, active, passive, controlled, uncontrolled, overcontrolled, minimally controlled, heavily redundant, minimally redundant, or any combination of the above.
We will gain some experience with a representative aerial block, a terrestrial block, and a staellite image block, using our own software, plus possible some commercial software tools. Interesting aspects to consider are accuracy, precision, reliability, error propagation, blunder detection, speed, ease of operation, flexibility, editing, presentation of results. Triangulation is overwhelmingly a point-based operation, although it can also be feature-based. Following triangulation, stereo pairs are often selected for intensive exploitation, terrain and feature extraction, and many other applications.
Other topics that we will try to cover include matching (space domain, frequency domain, correlation, least squares, constrained, unconstrained), DEM extraction, automated and semi-automated feature extraction (points, lines, segments, higher level features), visualization (synthetic imagery, perspective view generation, environemnts, VRML), radar and SAR imagery.