Adaptive Density Spatial Clustering Method Fusing Chameleon Swarm Algorithm.

Area of Science:

• Data Science
• Machine Learning
• Computer Vision

Background:

• Density-based spatial clustering of applications with noise (DBSCAN) is effective for arbitrary datasets.
• DBSCAN's performance is highly sensitive to neighborhood radius (Eps) and noise point identification.
• Existing DBSCAN methods struggle with rapid and accurate parameter optimization.

Purpose of the Study:

• To develop an adaptive DBSCAN method (CSA-DBSCAN) for improved clustering accuracy and efficiency.
• To optimize DBSCAN's Eps parameter and noise point identification using the chameleon swarm algorithm (CSA).
• To enhance image segmentation capabilities by integrating superpixel information.

Main Methods:

• Utilized the chameleon swarm algorithm (CSA) to iteratively optimize DBSCAN's clustering evaluation index.
• Introduced deviation theory in spatial distance for refined nearest neighbor search and noise point assignment.
• Incorporated color image superpixel information to improve segmentation performance.

Main Results:

• CSA-DBSCAN demonstrated rapid identification of accurate clustering results on synthetic and real-world datasets.
• The algorithm effectively addressed the over-identification of noise points.
• Significant improvements were observed in color image segmentation tasks.

Conclusions:

• CSA-DBSCAN offers a practical and effective solution for accurate and efficient clustering.
• The adaptive approach enhances DBSCAN's robustness and applicability in image segmentation.
• The method shows considerable clustering effectiveness and real-world practicality.