Persistent Homology Meets Object Unity: Object Recognition in Clutter | IEEE Journals & Magazine | IEEE Xplore

Persistent Homology Meets Object Unity: Object Recognition in Clutter


Abstract:

Recognition of occluded objects in unseen and unstructured indoor environments is a challenging problem for mobile robots. To address this challenge, we propose a new des...Show More

Abstract:

Recognition of occluded objects in unseen and unstructured indoor environments is a challenging problem for mobile robots. To address this challenge, we propose a new descriptor, Topological features Of Point cloud Slices (TOPS), for point clouds generated from depth images and an accompanying recognition framework, TOPS for Human-inspired Object Recognition (THOR), inspired by human reasoning. The descriptor employs a novel slicing-based approach to compute topological features from filtrations of simplicial complexes using persistent homology and facilitates reasoning-based recognition using object unity. Apart from a benchmark dataset, we report performance on a new dataset, the UW Indoor Scenes (UW-IS) Occluded dataset, curated using commodity hardware to reflect real-world scenarios with different environmental conditions and degrees of object occlusion. THOR outperforms state-of-the-art methods on both the datasets and achieves substantially higher recognition accuracy for all the scenarios of the UW-IS Occluded dataset. Therefore, THOR is a promising step toward robust recognition in low-cost robots, meant for everyday use in indoor settings.
Published in: IEEE Transactions on Robotics ( Volume: 40)
Page(s): 886 - 902
Date of Publication: 18 December 2023

ISSN Information:

Funding Agency:


I. Introduction

Object recognition is essential to robot visual perception as most vision tasks fundamentally rely on the ability to recognize objects, scenes, and categories. Object recognition in humans is incredibly sophisticated; humans recognize a multitude of objects in unstructured environments regardless of occlusion or variations in appearance, viewpoint, size, scale, or pose. Despite several efforts ranging from classical [1] to modern computer vision methods [2], achieving such performance in robot vision systems with commodity hardware is still challenging [3]. As a step toward addressing this multifaceted challenge, we present a recognition framework closely aligned with how object recognition works in humans [4], [5]. We combine persistent homology, a computational topology tool, with human intelligence mechanisms such as object unity [6] and object constancy [5] to achieve object recognition in cluttered environments.

Contact IEEE to Subscribe

References

References is not available for this document.