Contour Information Entropy Based Low-Poly Building Model Reconstruction Algorithm

Due to the fact that building models acquired through photogrammerty and remote sensing typically contain a large amount of redundant information, making them difficult to directly apply in large-scale scene rendering software. To reconstruct low-poly building models from dense building models, this paper proposes a contour-based low-poly building model reconstruction algorithm using information entropy. Initially, a metric for evaluating contour complexity is introduced, which is used to select contours at structural changes in the building model. Subsequently, a method for assessing contour discrepancy is presented, enabling the effective selection of necessary contours for reconstruction. Finally, post-processing algorithm is applied to the selected contours to remove those significantly affected by noise and redundancy. Experimental results on the Helsinki dataset demonstrate that the proposed algorithm significantly reduces the number of contours and polygonal faces required for reconstruction while preserving more architectural features in the low-poly building models.