RISAS, A Novel Rotation, Illumination, Scale invariant Appearance and Shape Feature

This is a joint work by Kanzhi Wu, Xiaoyang Li, Yong Liu, Ravindra Ranasinghe, Gamini Dissanayake and Rong Xiong and submitted to The 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2016).

Motivation

Compared with existing prominent 2D image features such as SIFT and SURF, a major limitation of current RGB-D detectors and descriptions[1][2][3] is that they are designed without simultaneously considering effectiveness and both detection and description, which the present paper aims to address.

Method

On descriptor end, we further improve previous work, LOIND[4], into LOIND+ on 2 aspects: 1) more accurate neighbourhood region selection 2) more robust and efficient dominant orientation estimation. On the detector end, based on the design principle of LOIND, we propose an RGB-D keypoint detector which is able to extract information rich region in both grayscale and depth channel. The combination of the proposed detector and descriptor, RISAS, shows state-of-the-art performance under illumination, rotation and scale variations. Furthermore, another contribution of this work is that we construct the first dataset specifically for RGB-D feature evaluation which can be download from link.

Results

risas_demo

Video

RISAS_VIDEO

Code

C++ code is still under preparing

Reference

  • [1] Tombari, Federico, Samuele Salti, and Luigi Di Stefano. “Unique signatures of histograms for local surface description.” Computer Vision–ECCV 2010. Springer Berlin Heidelberg, 2010. 356-369.
  • [2] Rusu, Radu Bogdan, et al. “Fast 3d recognition and pose using the viewpoint feature histogram.” Intelligent Robots and Systems (IROS), 2010 IEEE/RSJ International Conference on. IEEE, 2010.
  • [3] Nascimento, Erickson R., et al. “BRAND: A robust appearance and depth descriptor for RGB-D images.” Intelligent Robots and Systems (IROS), 2012 IEEE/RSJ International Conference on. IEEE, 2012.
  • [4] Feng, Guanghua, Yong Liu, and Yiyi Liao. “LOIND: An illumination and scale invariant RGB-D descriptor.” Robotics and Automation (ICRA), 2015 IEEE International Conference on. IEEE, 2015.