Fig. 1. Flowchart of proposed method

Fig. 2. Illustration of point cloud segmentation based on graph-cuts model. (a) Aligned result of two heterologous point clouds; (b) segmentation results

Fig. 3. Illustrations of guided point cloud filtering with blended point clouds. (a)(b) Terrestrial laser scanning (TLS) point cloud used only; (c)(d) proposed method

Fig. 4. TLS point cloud and dense matching point cloud of a laboratory building. (a) TLS point cloud; (b) dense matching point cloud; (c) aligned result of two different point clouds; (d) point cloud of area A1; (e) point cloud of area A2；(f) point cloud of area C; (g) point cloud of area B1; (h) point cloud of area B2; (i) point cloud of area D

Fig. 5. Fusing results of two cross-source point clouds by using proposed method. (a) Dense matching point cloud processed by using proposed segmentation method; (b) dense matching point cloud segmented and smoothed by using proposed method; (c) segmented dense matching point cloud and TLS point cloud; (d) final dense matching point cloud and TLS point cloud processed by proposed method

Fig. 6. Comparison between facade LiDAR point cloud and dense matching point cloud of a laboratory building. (a) Aligned result of LiDAR point cloud and dense matching point cloud; (b) blended point clouds after segmentation; (c) comparison between original dense matching point cloud and smoothed point cloud based on progressive migration method; (d) comparison between original dense matching point cloud and smoothed point cloud based on proposed method; (e) smoothed dense matching point cloud based on progressive migration method; (f) smoothed dense matching point cloud based on proposed method; (g) blended point clouds with overlapped area; (h) comparison between original dense matching point cloud and smoothed point cloud based on progressive migration method; (i) comparison between original dense matching point cloud and smoothed point cloud based on proposed method; (j) comparison between two smoothed dense matching point clouds based on proposed method and progressive migration method

Table 1. Detailed description of three datasets

Table 2. Parameters used in experiment

Table 3. Quantitative evaluation of smoothing effect of dense matching point cloud on TanksandTemples benchmark at evaluation threshold of 10 cm