An Efficient and Regularized Modeling Method for Massive Scattered Data Combining Triangulated Irregular Network and Multiquadric Function
Abstract
:1. Introduction
2. Methodology
2.1. Modeling Method of MQ Functions with Stability
Algorithm 1. Pseudocode for the iterative solution of linear equations using the damped least squares conjugate gradient method |
INPUT: (coefficient matrix), (vector), (damping factor), (stopping criterion) OUTPUT: (solution vector) 1. FOR DO 2. 3. 4. 5. 6. 7. IF BREAK 8. 9. END FOR |
2.2. Constructing the Concave Hull Boundary of Discrete Points According to the TIN Model
- Based on the planar coordinates of discrete points, the maximum convex hull containing the Delaunay triangulation network (i.e., the TIN model, as shown in Figure 1a) [33] is constructed using the divide-and-conquer algorithm. During the construction of the TIN model, the topological relationships of the triangulation network are recorded, including the indices of triangles, the indices of triangle vertex coordinates, and the triangles adjacent to each vertex. Additionally, the average distance between adjacent discrete points in the triangulation network is calculated.
- Triangles that belong to the convex hull boundary (as illustrated in Figure 1b) are identified and added to the boundary triangle queue T.
- A triangle is retrieved from T, and the following conditions must be met: ① if one edge AB is on the boundary, and the length of AB exceeds (the pre-set maximum length), then the two triangles adjacent to the internal edges BC and CA of are added to T, and is removed from T; ② if two edges AB and BC are on the boundary, and the length of AB or BC exceeds and the number of triangles connected to vertex B is greater than 1 (indicating that point B is not an isolated point), then the triangle adjacent to the internal edge CA is added to the T, and is removed from T. This process is repeated, extracting triangles from T and checking if they meet conditions ① and ② until no triangles in T satisfy either condition, ultimately resulting in a triangle queue T that forms a concave hull boundary (as shown in Figure 1c).
- Boundary edges are extracted from the T that forms the concave hull, and these edges are sequentially connected in either a counterclockwise or clockwise direction to construct an ordered list for the concave hull boundary (as shown in Figure 1d).
2.3. Efficient Modeling Approach of MQ Function Combined with TIN Model
- Initially, the connectivity relationships between triangles in the TIN model are stored in the array TriNet, and the topological depth of the subdomain triangulation network, TopoDepth, is set. In practice, this can be chosen based on the uniformity of the distribution of discrete points (e.g., between 3 and 9). When a value of 5 is selected, the area covered by the subdomain is as shown in Figure 2.
- In the TIN model, a triangle k is selected and used as the center to gradually expand outward (as shown in Figure 2). The triangles obtained are sequentially placed into the container LocalTriangles until the topological depth, TopoDepth, is satisfied, at which point the subdomain is determined. The pseudocode for the recursive function of the topological process is presented in Algorithm 2, where is the index of the vertices of the triangular element.
- All vertices of the triangles within the subdomain ( vertices) are extracted and placed into the container LocalVertex. Based on Equation (8), a local MQ surface function is constructed as
- The processes (2) and (3) are repeated until the local modeling of all subdomains is completed. By stitching together the interpolation results of the central triangles of all subdomains, the modeling of the entire research area is ultimately completed, and the modeling results are stored in the container ModelData.
Algorithm 2. Pseudocode for the topological process recursion function |
INPUT: TopoDepth, k, TriNet OUTPUT: LocalTriangles function GetLocalTriangles (TopoDepth, k, TriNet, LocalTriangles) 1. IF TopoDepth==0 RETURN 2. FOR DO 3. 4. IF LocalTriangles.push_back(ik) 5. ELSE CONTINUE 6. GetLocalTriangles (, , TriNet, LocalTriangles) 7. END FOR END FUNCTION |
3. Parallel Implementation of Modeling Method
- Load discrete data records . Display the approximate distribution of discrete points in the program window and show the range of distribution for and the minimum and maximum limits for the modeling result z.
- Construct the TIN model. Record the topological information of the triangular network and calculate the average distance between adjacent discrete points in parallel.
- Build the convex hull boundary of the TIN model. Adjust the maximum boundary limit length with reference to the average distance of adjacent discrete points, perform multiple calculations, check whether the convex hull boundary is reasonable, and output the boundary data file (*.bln).
- Set the topological depth, TopoDepth, of the subdomain triangular network. Select from the range of 3 to 9, with a default choice of 5.
- Set the number of modeling nodes and in the x and y directions. To ensure high-quality modeling, try to make the scale of the modeling unit smaller than the average distance between adjacent discrete points.
- The modeling nodes contained within each triangle are determined based on the relative position relationship between the modeling nodes and the TIN model (as illustrated in Figure 3). For modeling node P and (as shown in Figure 3), if conditions , , and are met, it is indicated that point P is located on the left side of each edge of the triangle, either within or on the edge of the triangle. Otherwise, tracking continues in the direction of the side where the cross-product is less than 0 (to the right side of this edge). If the tracking reaches the boundary without locating the triangle to which point P belongs, this indicates that the point lies outside the concave hull boundary, and no interpolation is performed for nodes outside this boundary. The modeling nodes contained in the triangle numbered k are placed into the container ModelNodesInTri, and the number of model nodes contained in the k-th triangle is denoted as .
- Parallelization of modeling. The modeling tasks of subdomains are independent of each other and can be dynamically allocated to multiple threads to improve the computational efficiency. The pseudocode for parallel computation is shown in Algorithm 3.
Algorithm 3. Pseudocode for parallel modeling computation |
2. #pragma omp parallel for private (LocalTriangles, LocalVertex, w) schedule (dynamic) 3. FOR DO 4. IF CONTINUE 5. IF () THEN 6. GetLocalTriangles (, i, TriNet, LocalTriangles) 7. ELSE 8. GetLocalTriangles (TopoDepth, i, TriNet, LocalTriangles) 9. END IF
13. 14. FOR DO 15. xx = LocalVertex[i].x-ModelData[ModelNodesInTri[k][j]].x 16. yy= LocalVertex[i].y-ModelData[ModelNodesInTri[k][j]].y 17. 18. END FOR // Constrain the interpolation result . 19. IF THEN 20. ELSE IF THEN 21. END IF 22. ModelData[ModelNodesInTri[k][j]]. 23. END FOR 24. END FOR |
4. Modeling Experiment
4.1. Accuracy Validation on a Synthetic Model
4.2. Stability Validation in the Seabed Terrain Modeling of a Fjord in Norway
4.3. Efficiency Validation of Terrain Modeling for Massive Digital Elevation Data
5. Discussion
6. Conclusions
- An efficient and regularized modeling method integrating the TIN model and MQ functions, referred to as TIN-MQ, is proposed. It is demonstrated in Section 4.1 and Section 4.3 that the modeling error of TIN-MQ is halved compared to the kriging method, and its modeling efficiency exceeds that of kriging by several tens of times. The minimal increase in the computational time with an increase in the number of modeling nodes is identified as the most significant advantage of the TIN-MQ method. This method is noted for its excellent modeling quality and high efficiency, offering broad prospects for application in massive data modeling.
- The stability of MQ function modeling is significantly enhanced by the construction of constrained residual MQ functions and a damped least squares linear equation (Equation (7)), which is solved using the conjugate gradient method, known for its strong anti-ill-conditioning capabilities. Section 4.2 indicates that the stability of the TIN-MQ method surpasses that of the MQ-RBF method in the commercial software Surfer.
- The integration of the TIN model significantly facilitates the determination of concave hull boundaries for discrete points and modeling within these boundaries, effectively shielding against redundant information from outside the boundaries. Additionally, the adjacency of triangles in the TIN model is utilized to construct modeling subdomains, which is key in enhancing the overall modeling efficiency of the TIN-MQ method.
- The dynamic distribution of all subdomain modeling tasks to all threads, enabling the independent execution of assigned tasks by each thread, is achieved using the OpenMP technology. This is a key technology that significantly enhances the modeling efficiency of the TIN-MQ method in this paper.
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
References
- Xin, Q.; He, Z. Three dimensional stratum interpolation and visualization based on section and borehole data from jointing the moving least square method and poisson reconstruction method. Earth Sci. Inform. 2020, 13, 1341–1349. [Google Scholar] [CrossRef]
- Zhang, J. Optimization of interpolation parameters based on statistical experiment. Open Geosci. 2022, 14, 880–905. [Google Scholar] [CrossRef]
- Qin, S.; Dai, Z. Interpolation Technique for the Underwater DEM Generated by an Unmanned Surface Vessel. CMES Comput. Model. Eng. Sci. 2023, 136, 3157–3172. [Google Scholar] [CrossRef]
- Yan, L.; Tang, X.; Zhang, Y. High Accuracy Interpolation of DEM Using Generative Adversarial Network. Remote Sens. 2021, 13, 676. [Google Scholar] [CrossRef]
- Franke, R. Scattered data interpolation: Tests of some methods. Math. Comput. 1982, 38, 181–200. [Google Scholar] [CrossRef]
- Myers, D.E. Spatial interpolation: An overview. Geoderma 1994, 62, 17–28. [Google Scholar] [CrossRef]
- Pouladi, N.; Møller, A.B.; Tabatabai, S.; Greve, M.H. Mapping soil organic matter contents at field level with Cubist, Random Forest and kriging. Geoderma 2019, 342, 85–92. [Google Scholar] [CrossRef]
- Vasileios, B.; Maria, M.; Nikolaos, D. Comparison between different spatial interpolation methods for the development of sediment distribution maps in coastal areas. Earth Sci. Inform. 2023, 16, 2069–2087. [Google Scholar] [CrossRef]
- Sekulić, A.; Kilibarda, M.; Heuvelink, G.; Nikolić, M.; Bajat, B. Random Forest Spatial Interpolation. Remote Sens. 2020, 12, 1687. [Google Scholar] [CrossRef]
- Fonseca, C.R.A.F.; Costa, J.F.C.L.; Hundelshaussen, R.; Bassani, M.A.A. Kriging parameter optimisation: Global versus local search strategies. Appl. Earth Sci. 2021, 130, 185–196. [Google Scholar] [CrossRef]
- Vigsnes, M.; Kolbjørnsen, O.; Hauge, V.L.; Dahle, P.; Abrahamsen, P. Fast and Accurate Approximation to Kriging Using Common Data Neighborhoods. Math. Geosci. 2017, 49, 619–634. [Google Scholar] [CrossRef]
- Jun, F.U. Kriging Interpolation with Neighborhood Constraints. Master’s Thesis, Zhejiang University, Hangzhou, China, 2022. [Google Scholar] [CrossRef]
- Hardy, R.L. Multiquadric equations of topography and other irregular surfaces. J. Geophys. Res. 1971, 76, 1905–1915. [Google Scholar] [CrossRef]
- Liu, J.; Wang, F.; Nadeem, S. A new type of radial basis functions for problems governed by partial differential equations. PLoS ONE 2023, 18, e0294938. [Google Scholar] [CrossRef] [PubMed]
- Ku, C.Y.; Xiao, J.E.; Liu, C.Y. A Novel Meshfree Approach with a Radial Polynomial for Solving Nonhomogeneous Partial Differential Equations. Mathematics 2020, 8, 270. [Google Scholar] [CrossRef]
- Chen, Y.T.; Li, C.; Yao, L.Q.; Cao, Y. A Hybrid RBF Collocation Method and Its Application in the Elastostatic Symmetric Problems. Symmetry 2022, 14, 1476. [Google Scholar] [CrossRef]
- Bawazeer, S.; Baakeem, S.; Mohamad, A. New Approach for Radial Basis Function Based on Partition of Unity of Taylor Series Expansion with Respect to Shape Parameter. Algorithms 2020, 14, 1. [Google Scholar] [CrossRef]
- Tavaen, S.; Kaennakham, S. Numerical Comparison of Shapeless Radial Basis Function Networks in Pattern Recognition. Comput. Mater. Contin. 2022, 74, 4081–4098. [Google Scholar] [CrossRef]
- Rippa, S. An algorithm for selecting a good value for the parameter c in radial basis function interpolation. Adv. Comput. Math. 1999, 11, 193–210. [Google Scholar] [CrossRef]
- Beatson, R.; Levesley, J.; Mouat, C. Better bases for radial basis function interpolation problems. J. Comput. Appl. Math. 2011, 236, 434–446. [Google Scholar] [CrossRef]
- Ku, C.Y.; Liu, C.Y.; Xiao, J.E.; Hsu, S.M. Multiquadrics without the Shape Parameter for Solving Partial Differential Equations. Symmetry 2020, 12, 1813. [Google Scholar] [CrossRef]
- Jiang, Z.W.; Wang, R.H.; Zhu, C.G.; Xu, M. High accuracy multiquadric quasi-interpolation. Appl. Math. Model. 2010, 35, 2185–2195. [Google Scholar] [CrossRef]
- Ullah, R.; Ali, I.; Shaheen, S.; Khan, T.; Faiz, F.; Rahman, H. A Mesh-Free Collocation Method Based on RBFs for the Numerical Solution of Hunter–Saxton and Gardner Equations. Math. Probl. Eng. 2022, 2022, 2152565. [Google Scholar] [CrossRef]
- Balazovicova, L. Riverbed response to high flows with installed willow spiling using Multiquadric Radial Basis Function (RBF MQ). Geogr. Cassoviensis 2021, 15, 121–134. [Google Scholar] [CrossRef]
- Mehdi, A.; Azra, K. A Multimethod Analysis for Average Annual Precipitation Mapping in the Khorasan Razavi Province (Northeastern Iran). Atmosphere 2021, 12, 592. [Google Scholar] [CrossRef]
- Van Do, V.N.; Lee, C.H. Nonlinear analyses of FGM plates in bending by using a modified radial point interpolation mesh-free method. Appl. Math. Model. 2018, 57, 1–20. [Google Scholar] [CrossRef]
- Liu, H.; Guo, P.; Liu, J.; Liu, R.; Tong, T. An extension of multiquadric method based on trend analysis for surface construction. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 3435–3441. [Google Scholar] [CrossRef]
- Song, X.; Rui, X.; Ju, Y.; Yang, Y. Improved geologic surface approximation using a multiquadric method with additional constraints. Min. Sci. Technol. 2010, 20, 600–606. [Google Scholar] [CrossRef]
- Zhang, J.; Liu, P. Combining Fitting Based on Robust Trend Surface and Orthogonal Multiquadrics with Application in DEM Fitting. Acta Geod. Et Cartogr. Sin. 2008, 37, 526–530. (In Chinese) [Google Scholar]
- Jordan, G. Adaptive smoothing of valleys in DEMs using TIN interpolation from ridgeline elevations: An application to morphotectonic aspect analysis. Comput. Geosci. 2006, 33, 573–585. [Google Scholar] [CrossRef]
- Li, L.; Kuai, X. An efficient dichotomizing interpolation algorithm for the refinement of TIN-based terrain surface from contour maps. Comput. Geosci. 2014, 72, 105–121. [Google Scholar] [CrossRef]
- Liu, H.; Liu, J.; Ma, C. Inverse Theory and Methodology for Direct Current Induced Polarization Method. Central South University Press: Changsha, China, 2017. [Google Scholar]
- Dwyer, R.A. A Faster Divide-and-Conquer Algorithm for Constructing Delaunay Triangulations. Algorithmica 1987, 2, 137–151. [Google Scholar] [CrossRef]
Modeling Method | Average Error | Maximum Error |
---|---|---|
Kriging | 0.000924 | 0.00986 |
TIN-MQ | 0.000408 | 0.00532 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Liu, H.; Zhang, Y.; Liu, X.; Ahmed, I.; Liu, J. An Efficient and Regularized Modeling Method for Massive Scattered Data Combining Triangulated Irregular Network and Multiquadric Function. Mathematics 2025, 13, 978. https://doi.org/10.3390/math13060978
Liu H, Zhang Y, Liu X, Ahmed I, Liu J. An Efficient and Regularized Modeling Method for Massive Scattered Data Combining Triangulated Irregular Network and Multiquadric Function. Mathematics. 2025; 13(6):978. https://doi.org/10.3390/math13060978
Chicago/Turabian StyleLiu, Haifei, Yuhao Zhang, Xin Liu, Ijaz Ahmed, and Jianxin Liu. 2025. "An Efficient and Regularized Modeling Method for Massive Scattered Data Combining Triangulated Irregular Network and Multiquadric Function" Mathematics 13, no. 6: 978. https://doi.org/10.3390/math13060978
APA StyleLiu, H., Zhang, Y., Liu, X., Ahmed, I., & Liu, J. (2025). An Efficient and Regularized Modeling Method for Massive Scattered Data Combining Triangulated Irregular Network and Multiquadric Function. Mathematics, 13(6), 978. https://doi.org/10.3390/math13060978