Extraction of urban 3d features from lidar data fused with aerial images using an improved mean shift algorithm

摘要

An approach is proposed for the extraction of the urban three-dimensional features efficiently and accurately. In this method, firstly, both the LIDAR data and the aerial images are respectively pre processed and matched using the affine transformation model .In order to exploit the spectral data and classify the LIDAR data with high accuracy, a data extraction procedure is employed which extracts the converted pixel values of the aerial image to LIDAR data. Then, an improved Mean Shift algorithm is employed to classify the LIDAR data fused with reflected intensity and spectrum attribute into groups by kinds of feature, such as buildings, vegetation, water etc. The classification accuracy is evaluated by space accuracy and confusion matrix evaluation. Finally, the 3D models of interested regions are quickly constructed based on the classified points and the aerial-image by SketchUp. Using this method, the 3D models of urban objects could be easily extracted and constructed.