Developments in cultural feature extraction have resulted in a number of disparate processes some based primarily on photogrammetric approaches and some based primarily on image processing approaches. Almost all the practical implementations to date have resulted in tools that assist an operator in feature extraction but are not fully automatic. A major shortcoming in these systems is the ability to reliably extract bare earth digital elevation models. Largely lacking from prior efforts are methods of utilizing other geospatially referenced data and data components to draw inferences about photogrammetrically derived elevation data. The proposed research explores iterative extraction processes that use the output of previous processing to filter subsequent results. The approach envisioned is similar to Kalman filtering used for calculating positions. In most positioning applications, both system error and geometric dilution of precision can affect the accuracy of the position. Kalman filters use iterative processing to define exclusion areas, i.e. areas where the target cannot be located. Eventually the position solution converges on the true position. The solution is found by using context cues of where the target cannot be. A similar exclusionary process could be implemented by iteratively classifying extracted coplanar features as exclusion areas from determining ground level
