Comparing the effects of fused and non-fused imagery on objectoriented classificaiton
The effects of fused and non-fused imagery on object-oriented classification were analyzed. Imagery for this study included a four band multispectral Quickbird image and LiDAR bare earth and surface models. Results of a Quickbird only classification and a fused Quickbird and LiDAR classification were compared. The nearest neighbor classifier in Definiens Professional was used. Classes included roads, buildings, rock, bare soil, grass, trees, lakes and ponds, and rivers and streams. The segmentation process was successful in the creation of semantically meaningful objects. The addition of elevation information to the segmentation process, however, produced image objects that better represented our land cover classes. Both the fused and Quickbird classifications had overall accuracies above 90%. No statistical differences between the classifications were found, however. Individual class user's and producer's accuracy generally increased for the fused classification.
