Scene change detection and description using Optical Flow

Topological mapping and localization using optical flow only.

The nodes in the topological map are at the location of scene changes (in the form of depth discontinuities and large appearance changes in the scene). These are detected at the location of the zero-crossings of the 2nd derivative of the flow response signal.

The scene changes are described using lower and higher statistical moments, location, and distribution of the induced optical flow in the frame.

Matching is carried out by calculating the Mahalanobis and
Chi-square distances and choosing the smallest distance as the match.


Publications:
N Nourani-Vatani, P V K Borges, J Roberts and M V Srinivasan, Extracting Descriptors from Optical Flow for Topological Localization, IEEE International Conference on Computer Vision, Workshop on Challenges and Opportunities in Robot Perception, November 2011, Barcelona, Spain [In Review]

N Nourani-Vatani, P V K Borges, J Roberts and M V Srinivasan, On the Use of Optical Flow for Scene Change Detection, IEEE Transactions on Robotics, Jan 2011 [In Review]

N Nourani-Vatani and C Pradalier, Scene Change Detection for Topological Mapping and Localization, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), October 2010, Taipei, Taiwan

Category:

Science & Technology

Tags:

• Computer vision
• Optical Flow
• Localization and mapping