Computer vision systems that attempt to extract cultural features from aerial imagery are often forced to interpret segmentations where the actual features are broken into numerous segments or fragments. For example, reads and road-like features are difficult to completely segment due to occlusions, poor contrast with their surroundings, and changes in surface material. Often the nature of the segmentation process is designed to err toward oversegmentation of the image, since the joining of feature descriptions is believed to be simpler than their decomposition. No matter what the cause, it is necessary to aggregate these incomplete segmentations, filling in missing information, in order to reason about the overall scene interpretation. This paper describes a method to select sets of such fragments as candidates for alignment into a single region, as well as a procedure to generate new linear regions that are linked composites of the original sets of fragments. Portions of the composite region that lie between pairs of the original fragments are approximated with a spline. The resulting composite region can be used to predict the areas in which to search for missing components of the cultural feature. Copyright © 1985 David M. McKeown, Jr. and John F. Pane This research partially sponsored by the Defense Advanced Research Projects Agency (DOD), ARPA Order No. 3597, monitored by the Air Force Avionics Laboratory Under Contract F33615-81 -K-1539 and by the Defense Mapping Agency Under Contract DMA 800-85-C-0009. The views and conclusions contained in this document are those of the author and should not be interpreted as representing the official policies, either expressed or implied, of the Defense Advanced Research Projects Agency, of the Defense Mapping Agency, or the US Government. To be presented at the IEEE Conference on Computer Vision and Pattern Recognition, June 9-13, 1985 in San Francisco, California.