Information-theoretic adaptive clustering for robust multi-structure model fitting

doi:10.1007/s10586-026-06563-2

Return

Information-theoretic adaptive clustering for robust multi-structure model fitting

2026-09-27

PRE

DOI：10.1007/s10586-026-06563-2

Original

Original request for help

Save

Abstract

En 中文

In recent years, density-based clustering algorithms have become increasingly popular in the field of model fitting. However, a critical limitation of these algorithms is their reliance on a fixed threshold (MinPts), which often lacks adaptability to varying data characteristics. When input data is severely contaminated with outliers, determining suitable threshold values for clustering becomes challenging, leading to suboptimal model fitting results. To address these challenges, we propose a novel Information-theoretic Adaptive Clustering-based Fitting (IACF) method, achieving accurate and efficient model fitting. First, we construct a reliable neighborhood graph by integrating motion and preference information, which effectively prunes invalid edges and preserves potential local inlier structures. Second, we propose an entropy-based adaptive strategy that quantifies the disorder of local neighborhoods to dynamically determine optimal clustering thresholds (MinPts*) for each specific dataset without prior knowledge. Building on this, we develop an Entropy-Adaptive Density-based Dominant Set Clustering (EADSC) algorithm to autonomously detect model-related subgraphs corresponding to model instances and accurately partition inliers from outliers. Experimental evaluations conducted on several challenging datasets show that the proposed IACF method outperforms several state-of-the-art fitting methods in terms of both fitting accuracy and computational efficiency.

Keywords:

Density-clustering algorithm

Multi-structure model fitting

Information entropy

Adaptive threshold

Model selection