Fast Globally Optimal ICP - A CUDA Implementation of Go-ICP

University of Pennsylvania, CIS 5650: GPU Programming and Architecture, Final Project

• Authors: Mufeng Xu & Zhaojin Sun
• Tested on:
  • Windows 11/Ubuntu 24.04, i9-13900H @ 2.6GHz 32GB, RTX 4080 Laptop 12GB (Personal Computer)

Demo

Introduction

Algorithm

The key algorithm in the globally optimal ICP problem is Branch-and-Bound (BnB), global minimum is searched in the $SE(3)$ space (3D Rigid Rotation and Translation), which is 6-dimensional.

Instead of directly searching over $SE(3)$, it is proposed to launch an outer-BnB, which searches over the $SO(3)$ space (3D Rotation) with BnB algorithm, and for each branch, an inner-BnB function, which searches over the $\mathbb{R}^3$ space, will be invoked.

Dependencies

• OpenGL Mathematics (GLM)
• TOML++
• tinyply
• nanoflann

Reference

1. Jiaolong Yang, Hongdong Li, Dylan Campbell and Yunde Jia. Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2016.
2. Jiaolong Yang, Hongdong Li and Yude Jia. Go-ICP: Solving 3D Registration Efficiently and Globally Optimally. International Conference on Computer Vision (ICCV), 2013.
3. Go-ICP (GitHub)