3D Computer Vision Scanner

A turntable-based 3D scanner built as part of a university project in geometric and 3D computer vision. This repository provides the code, documentation, and data to calibrate, drive a laser + camera setup, scan objects, and reconstruct 3D geometry.

Overview

3D scanning is accomplished by combining a rotating turntable, a laser (planar line laser), and a camera. As the object rotates, the laser projects a line onto the surface, and the deformation of the line observed by the camera encodes depth. By sweeping many views and triangulating, a 3D point cloud or mesh is reconstructed.

This project was developed for the Geometric and 3D Computer Vision course at Ca'Foscari University of Venice, and demonstrates:

• Camera calibration
• Laser pose estimation
• Turntable control and synchronization
• Point cloud reconstruction

---

Features

• Camera calibration (intrinsics, distortion) via checkerboard images
• Extrinsic calibration to locate laser plane relative to camera
• Automated turntable & scanning loop
• Reconstruction of 3D points from laser sweep
• Utilities and helper functions for IO, transformations, visualization

Getting Started

Installation

1. Clone this repository:

   git clone https://github.com/Coalessence/3D-Computer-Vision-scanner.git
   cd 3D-Computer-Vision-scanner

2. Install dependencies with pip.

3. Place calibration or scan images under the G3DCV_turntable_scanner_data/ folder (or configure your own paths).

Calibration

1. Capture multiple checkerboard images for intrinsic calibration.

2. Run:

   python calibration.py path/to/checkerboard_images/

3. For extrinsic / laser calibration:

   python calibration.py path/to/laser/calibration_images/

4. This will generate calibration parameters (camera matrix, distortion coefficients, laser plane parameters) saved to files.

Scanning Workflow

Once calibration is done:

python scanner.py --config path/to/calib_params.json --out output_folder

This script will:

• Rotate the turntable step by step
• At each step, project the laser, capture an image
• Detect the laser line, triangulate depth points
• Aggregate the points into a full 3D point cloud

You can visualize the result using a point cloud viewer (e.g. via PyntCloud, Meshlab).

You can adjust parameters (e.g. number of steps, motor delay) in the scanner config or within scanner.py.