A Fast Connected-Component Labeling Algorithm for Robot Vision Based on Prior Knowledge

  • Authors:

  • Jun Liu;Guang Lu;Binbin Tao;Fang Chen;Haitao Gao;Zhisheng Zhang

  • Affiliations:

  • School of Mechanical Engineering, Southeast University,;School of Mechanical Engineering, Southeast University,;School of Mechanical Engineering, Southeast University,;School of Mechanical Engineering, Southeast University,;School of Mechanical Engineering, Southeast University,;School of Mechanical Engineering, Southeast University,

  • Venue:
  • ICIRA '09 Proceedings of the 2nd International Conference on Intelligent Robotics and Applications
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Machine vision is now a major technique for intelligent robot system to sense the outside world. Connected-component labeling is a simple and efficient way to help robot identify a specific region of interest (ROI). In this paper, the improvement of a two-scan algorithm based on prior knowledge is presented: (1) the rule of label assignment in the mask is improved and ROI orientation is introduced; (2) two cases of label equivalence in the mask were extracted to optimize the strategy of scanning; (3) two fast-connect ways were proposed to reduce the times of scanning. After the algorithm was implemented, parameters of each connected component are calculated to identify the ROI. In addition, this algorithm was also implemented in DSP platform on a service robot to identify a water cup. The experiment results demonstrated the efficiency of the algorithm is enhanced using the above strategies.