Clustering_Methods_Updated presentation..

Clustering_Methods_Updated presentation..

• 1.
  - Sneha B N Clustering Methods
• 2.
  Clustering  Clustering =Unsupervised learning technique  Groups data points into clusters based on similarity Popular Clustering Methods:  K-Means Clustering  Hierarchical Clustering  DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• 3.
  K-Means Clustering:  Dividesdata into k clusters  Assumes spherical clusters  Sensitive to outliers  Needs to predefine k
• 4.
  Hierarchical Clustering:  Buildsa hierarchy of clusters (tree-like structure)  Agglomerative (bottom-up) or Divisive (top- down)  Good for small datasets  Computationally expensive for large data
• 5.
  What is DBSCAN? Density-Based Spatial Clustering of Applications with Noise  Groups together points that are closely packed (high density)  Separates low-density regions (noise/outliers) Key Concepts in DBSCAN  Epsilon (ε): Radius to search for neighbors  MinPts: Minimum number of points to form a dense region  Points classified as:  Core Points (dense enough)  Border Points (less dense, but close)  Noise Points (outliers)
• 8.
  Why DBSCAN?  Real-worlddata isn't always spherical or even!  Clusters of arbitrary shape can exist  No need to predefine number of clusters (unlike k-means)  Can detect outliers (noise)
• 9.
  DBSCAN Advantages  Canfind clusters of arbitrary shape  No need to specify number of clusters  Robust to outliers  Works well with spatial data  Efficient for Large Datasets
• 10.
  DBSCAN Disadvantages  Hardto choose good ε and MinPts  Struggles with clusters of varying densities  Performance drops on high- dimensional data
• 11.
  DBSCAN in SpaceRobotics and Terrain Mapping  🚀 Used in NASA/ESA space missions for 3D mapping using LiDAR or stereo vision.  🧠 Clusters 3D point cloud data to identify terrain features: Rocks, cliffs, craters (non-spherical shapes).  🤖 Supports autonomous navigation of rovers by segmenting obstacles and traversable paths.  🌌 Ideal for unstructured, noisy environments like Mars or the Moon.
• 12.
  Research on DBSCANwith Point Cloud Data  📚 Efficient Clustering of LiDAR Data for 3D Object Detection - IEEE Paper using DBSCAN for 3D object detection from LiDAR data.  📚 DBSCAN in SLAM (Simultaneous Localization and Mapping) - Improves mapping precision by filtering noise in point clouds.  🌕 Lunar/Martian Terrain Segmentation - DBSCAN helps classify terrain types and supports hazard detection and mission planning.