Projects & Products / 3D Autoscan System
2019— · Shenzhen University

3D Autoscan System

A scanner + rotary platform pipeline that captures arbitrary objects to ~0.1 mm in under 5 minutes—built to seed a 3D shape database without per-object parameter tuning.

A scanner + rotary platform pipeline that captures arbitrary objects to ~0.1 mm in under five minutes — built to seed a 3D shape database without per-object parameter tuning.

Full system photo
Figure 1. Scanner + rotary platform setup. Controlled by a single PC running the registration pipeline.

Problem

Building a 3D shape database requires scanning many objects. Off-the-shelf scanning is fine for one model, but at database scale it costs human hours: every object shape requires re-tuning algorithm parameters, and every mis-registration eats into the ~5 min/object budget. The need is a robust pipeline that adapts — per-object parameter tuning has to disappear.

Pipeline

The system fuses three algorithmic stages tuned to handle scale, reflectance, and pose variation without manual intervention.

Frame-to-frame registration
Figure 2. Frame-to-frame point cloud + texture registration.

Outlier removal
Figure 3. Environmental outlier removal.

Global registration
Figure 4. Multi-pose global registration.

Frame-to-frame registration aligns sequential scans on both geometry and texture; an outlier removal stage strips away background noise; finally a global registration pass aligns the multiple platform poses into a coherent model.