Segmentation and classification of range images

Area of Science:

• Computer Vision
• 3-D Object Recognition
• Computational Geometry

Background:

• Object recognition in three-dimensional (3-D) space is crucial for computer vision systems.
• Range images provide direct 3-D surface coordinate data, making them suitable for this task.

Purpose of the Study:

• To develop and present a procedure for detecting connected planar, convex, and concave surfaces of 3-D objects.
• To segment and classify surface patches from range images for object reconstruction.

Main Methods:

• A three-stage procedure involving surface patch segmentation using clustering, classification of patches (planar, convex, concave) via statistical tests and eigenvalue analysis, and edge classification for patch merging.
• Utilizes surface points, surface normals, curvature values, and eigenvalue analysis.

Main Results:

• Successfully segmented range images into surface patches.
• Classified patches as planar, convex, or concave with high accuracy.
• Classified boundaries as crease or noncrease edges to merge patches into object faces.
• Demonstrated effectiveness on both real and synthetic images.

Conclusions:

• The developed procedure reliably detects and classifies 3-D object surfaces from range images.
• This method facilitates the reconstruction of object geometry for computer vision applications.
• The approach is robust and applicable to diverse imaging scenarios.