Home / Rigel SDK / 2D/3D Inspection

MODULE 03 — INSPECTION

2D/3D Inspection

Building blocks for inspection automation — image segmentation, feature extraction,
matching- and registration-based alignment, and classification for both 2D images and 3D volumes.

Segmentation

Image segmentation that isolates the regions to inspect.

Graph-based Segmentation

Graph-based segmentation algorithms such as GraphCut and Random Walker.

Region-based Segmentation

Region-expansion segmentation such as Watershed and Region Growing.

Level Set

Level-set segmentation that converges by propagating a boundary surface.

Feature Extraction

Extracts the shape and feature information that inspection decisions are based on.

Blob Analysis

Blob detection, feature extraction, and separation of touching blobs.

Contour Analysis

Contour extraction and shape feature analysis.

Feature Point Detection

Feature point detection and descriptor computation.

Alignment

Matching and registration — aligns inspection images to reference shapes and volumes.

Matching

Template matching, geometric shape matching, and feature point matching.

Image 2D-3D Registration

Registers a 3D volume to 2D images by projecting it through the X-ray geometry (DRR), using geometric parameters such as projection offsets, SID/SDD, and rotation directly as optimization variables.

Transform Models

Transform models selected step by step — from 2D/3D rigid, similarity, and affine to B-spline and displacement fields (free-form deformation).

Similarity Metrics

Similarity measures matched to modality and image characteristics — mutual information, normalized cross-correlation (NCC), mean squared error, and more.

Optimization

Optimizers including L-BFGS-B, Levenberg-Marquardt, and Powell, with multi-start, robust loss (Huber, IRLS), and multi-resolution strategies.

Point Set Registration

2D/3D point-set registration with phantom pose estimation, and non-rigid point-set registration such as board warpage.

Demons · Phase Correlation

Demons non-rigid registration, fast FFT-based estimation of shift, rotation, and scale, and sub-pixel registration based on feature point patches.

Classification

Inspection judgment and classification using the extracted features.

Feature-based Classification

Judgment and classification using segmentation and feature extraction results.

DL-linked Classification

Judgment linked with deep learning classification models. Details are covered in the AI & Deep Learning module.

Need 2D/3D Inspection technology?

We will guide you to the right configuration and an evaluation SDK for your system or software.

Contact us