Metallic Material Image Segmentation by using 3D Grain Structure Consistency and Intra/Inter-Grain Model Information
Abstract
In this project, we conducted research on developing new methods and software tools to automatically segment various microscopic images of materials (especially metallic materials) to accurately extract their micro-structures, which determine mechanical and other important properties of the materials. The major accomplished work includes: 1) a general multi-label segmentation propagation framework to preserve the shape, appearance, and topology properties of the segments from slice to slice for 3D material image segmentation, 2) new algorithms for enforcing specified topology in image segmentation, 3) an interactive segmentation tool by allowing minimal and simplistic interactions for more accurate 3D material image segmentation, 4) a new clustering method to effectively and robustly segment the super alloy grains from 3D multichannel super alloy images, where each channel corresponds to a specific microscope setting, 5) faster clustering algorithms based on Edge-Weighted Centroid Voronoi Tessellation model by using propagation of the inter-slice consistency constraint for large-scale material image segmentation, and 6) a fully-automatic method to detect cracks from pavement images that can be used for pavement road maintenance.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 05, 2015
- Accession Number
- ADA617033
Entities
People
- Song Wang
Organizations
- University of South Carolina