Object-based Classification of an Urban Area through a Combination of Aerial Image and Airborne Lidar Data

This paper studies the effect of airborne elevation information on the classification of an aerial image in an urban area. In an urban area, it is difficult to classify buildings relying solely on the spectral information obtained from aerial images because urban buildings possess a variety of roof colors. Therefore, combining Lidar data with aerial images overcomes the difficulties encountered with regard to the heterogeneous appearance of buildings. In the first stage of this process, building information is obtained and is extracted using the normalized Digital Surface Model, return information derived from the airborne Lidar data, and vegetation information obtained through preclassification. In the second stage of this process, the aerial image is segmented into objects. It is then overlaid with building information extracted from the first step in the process. By applying the definite rule to the resulting image, it is possible to determine whether or not the object is a building. In the final stage, the aerial image is classified by using the building object as ancillary data extracted from the prior stage. This classification procedure uses elevation and intensity information obtained from the Lidar data, as well as the red, green, and blue bands obtained from the aerial image. As a result, a method using the combination of an aerial image and the airborne Lidar data shows higher accuracy and improved classification, especially with regard to building objects, than results that rely solely on an aerial image.