Biography

Yu Chen is a SLAM algorithm engineer in Segway Robotics. He obtained his M.S. degree in Electronic Engineering and Computer Science School, Peking University, where he joined the 3D reconstruction group of EECS, he is advised by Prof. Yisong Chen and co-supervised by Shuhan Shen. His research interests mainly include 3D reconstruction, which include Structure-from-Motion (SfM), distributed optimization, Simultaneous Localization and Mapping (SLAM), semantic reconstruction. He used to be a web developer when he studied in BUAA. His dream is to build the world by computer vision techniques. His career in the future is classified…

Interests

  • Structure-from-Motion
  • Distributed Optimization
  • Simultaneously Localization and Mapping
  • Computer Vision

Education

  • M.S. in Computer Software and Theory, 2017.09 - now

    Peking University

  • B.S. in Software Engineering, 2013.09 - 2017.06

    Beihang University

Experience

 
 
 
 
 

Research Intern

TuSimple

Apr 2019 – Jul 2019 Beijing
Responsibilities include:

  • I researched in the globally optimal guaranteed rotation averaging optimization approaches. I was supervised by Dr. Ji Zhao.
  • I improved global rotation averaging approach by x100 times in efficiency, without precision loss compared with state-of-the-art in large scale datasets.
  • I make the global rotation averaging optimization approach practicable in SLAM backend.
 
 
 
 
 

Algorithm Intern

Megvii

Sep 2018 – Dec 2018 Beijing
Responsibilities include:

  • I implemented a mobile-based real-time 3D reconstruction framework utilizing the raw images and depth images as input.
  • I applied depth fusion-based approach for front-end reconstruction, multi-view based approach for backend texture mapping.

Recent Posts

Robust Optimization in 3D Vision

1. 凸优化和非凸优化

凸优化问题的一般形式

$$ \min\quad f(x) \qquad\qquad\qquad\qquad \
s.t.\quad h_i(x) = 0, \forall i = 1,\cdots, n\
\qquad\quad g_j(x) …

Survey of Global Structure from Motion

1. Global Structure-from-Motion by Similarity Averaging

之前所有的global SfM方法都分为四步:Rotation Averaging, Translation Averaging, …

三维重建中的旋转(Rotation)

旋转在三维重建中是比较重要的,这里主要对旋转的性质及应用做一些总结。

1. 旋转矩阵

设某个单位正交基 $(e_1, e_2, e_3)$ 经过一次旋转变成了 $(e_1^{'}, e_2^{'}, e_3^{'})$ 。那么, …

View Graph

首先我们来看看什么是view graph以及为什么view graph在重建中有什么作用?

A view graph is a graph which nodes represent images and edges represent the number of …

Introduction to Structure from Motion

1. Incremental Structure from Motion

第一次提出 Retriangulate 的概念来尝试解 …

Contact