Abstract | ||
---|---|---|
Clustering techniques can be adopted to analyze 3D model database and improve the retrieval performance. However, 3D model database lack valuable prior knowledge. Thus, it becomes difficult for the clustering methods to pre-decide the appropriate parameter's value. Moreover, clustering methods are short at handling outliers by treating outliers as "noise". The paper introduces a robust hierarchical clustering algorithm for analyzing 3D model database. The proposed algorithm stops automatically by utilizing outlier information and adopts the concept of core group to reduce the influence of parameter on the clustering result. Core group refers to the data that are always clustered together. After discussing some desirable properties of the new algorithm, the paper conducts a series of experiments on Princeton Shape Benchmark and 2 real-life datasets from UCI. Comparative study demonstrates advantages of our algorithm. |
Year | DOI | Venue |
---|---|---|
2006 | 10.1109/IMSCCS.2006.167 | IMSCCS (2) |
Keywords | Field | DocType |
core group,comparative study,artificial neural networks,hierarchical clustering,clustering algorithms,computer science,robustness,shape,solid modeling,data mining,information retrieval,data analysis,image retrieval | Data mining,Fuzzy clustering,CURE data clustering algorithm,Computer science,Artificial intelligence,Cluster analysis,Canopy clustering algorithm,Data stream clustering,Correlation clustering,Algorithm,Constrained clustering,Machine learning,DBSCAN | Conference |
Volume | Issue | ISBN |
2 | null | 0-7695-2581-4 |
Citations | PageRank | References |
0 | 0.34 | 7 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lv Tianyang | 1 | 33 | 8.49 |
Shaobin Huang | 2 | 11 | 7.93 |
Xi-zhe Zhang | 3 | 38 | 8.94 |
Zhengxuan Wang | 4 | 47 | 13.93 |