mpkuse / edge_alignment

Ros package for Edge Alignment with Ceres solver

Edge Alignment - CERES

Reimplementation of edge alignment [1] using CERES solver. Look at the folder standalone for an easy to understand standalone implementation (abeit inefficient).

Required Packages

• Eigen3
• CERES
• OpenCV
• ROS

[1] Kuse M., Shen S. “Robust Camera Motion Estimation using Direct Edge Alignment and Sub-gradient Method“. In Proc. of IEEE International Conference on Robotics and Automation (ICRA), 2016 in Stockholm, Sweden.

[2] CERES http://ceres-solver.org/

Sample Alignment

Note

This repo was an attempt for me to have a standalone implementation of edge alignment. A basic standalone implementation can be accessed in the folder standalone. It is not the most efficient implementation, but it works. It uses Ceres's cubic_interpolation. Contributions welcome!

I believe, Yongyen https://github.com/ygling2008/direct_edge_imu in his repo had fixed it and is usable. You are better off trying to run his code. The paper relating to yongyen's work is:
[3] Yonggen Ling, Manohar Kuse, and Shaojie Shen, “Direct Edge Alignment-Based Visual-Inertial Fusion for Tracking of Aggressive Motions“, submitted in Springer Autonomous Robots.

You may look at https://kusemanohar.wordpress.com/2015/09/14/icra2015-submission/ for more resources on this project.

If you are keen on using this, I suggest you implement the gradient decent yourself for this. Once you have a point cloud it is really not that hard.

edge_alignment's Issues

alignment not good.

Hi, Manohar.
Thanks for your open source code. I am also interested in edge_alignment. I have run your edge_alignment code on TUM RGBD dataset. But the result was not good enough. Even with 100 iteration the cost still very high. Do you have some data that I can try? or some thing I need to pay attention?

Thanks very much.