Residual tensor analysis for quality assessment of data integration

Fusion of different data sets in the geospatial field is often used in order to complement the advantages and disadvantages of individual data sets, such as integration of image data and vector data, fusion of Photogrammetric data and LiDAR data, or combination of cadastral data and image data etc. An important step in the data fusion is to register these data sets to a common coordinate system. In this paper, we use residuals after registration to develop a novel quality assessment method, namely residual tensor analysis, for fusion of LiDAR and topographic map data. The results show that our residual tensor method is superior to the common fitting error analysis. In particular, our residual tensor is very useful as a quality indicator for automatic building model construction using LiDAR and topographic map data.