Non-Contact Measurement Technologies

Study Notes

Non-contact devices collect point data without physical interaction, enabling rapid data acquisition.
Laser scanners are the most widely used non-contact measurement tools, notable for speed and accuracy in data collection.
There has been significant improvement in the accuracy of laser scanning equipment, resulting in increased industrial use.

CT scanners analyze material composition and internal geometry of components, crucial in both medical and industrial applications.
Medical CT imaging allows visualization of the human body, critical for surgical implant design.
High-power industrial CT scanners can safely obtain data regarding the internal shape, porosity, and density of industrial items.

Unique to the medical field, reverse engineering creates precise 3D models from CT or MRI scans, typically in DICOM format.
Applications include surgical planning, implant design, and educational demonstrations.
Initial steps involve processing CT or MRI scans to extract anatomical details, which are segmented based on pixel values.

Scans are segmented according to x-ray attenuation coefficients, isolating the areas of interest for further processing.
A mask is generated to represent the selected anatomy, leading to the creation of point cloud data.

Voxel models depict 3D objects by dividing them into a three-dimensional grid of voxels, aiding in visualizing scanned anatomy.
Point cloud data is transformed into STL (Stereolithography) files, which are essential for 3D printing, focusing on object surface geometry.

STL files facilitate the 3D printing of medical models, assisting in surgical preparations, implant designs, and patient education.
The reverse engineering process requires specialized software and expertise to ensure accuracy and detail.

Noise Reduction: Algorithms like statistical outlier removal and moving least squares filter unwanted noise from scanned data.
Registration: Multiple scans are aligned and merged using algorithms such as Iterative Closest Point (ICP) for complete geometry representation.
Down-Sampling: High-resolution point clouds are reduced using voxel grid filtering and random sampling to ease computational demands.
Normal Estimation: Algorithms like PCA compute surface normals, critical for mesh generation and reconstruction tasks.
Feature Extraction: Algorithms like FPFH and SIFT identify edges, corners, and planes in point clouds for object recognition and segmentation.
Segmentation: Segregation of various objects in point clouds is essential for focused processing and analysis.