Abstract
This work presents preliminary results of a textureless object recognition system for an indoor mobile robot.
Our approach relies on 1) segmented linear features, and 2) pairwise geometric relationships between features. This approach is motivated by the need for recognition strategies that can handle many of indoor objects that have no little or not textural information on their surfaces, but have strong geometrical consistency within the same object class.
Our matching method consists of two steps. First, we find correspondence candidates between linear fragments. Second, a spectral matching algorithm is used to find the subset of correspondences which is the most consistent. Both matching methods are learnt by using logistic classifiers.
We evaluated the developed recognition system with our own database, which is composed of eight indoor object classes. We also compared the performance of our line feature based recognition approach with a SIFT feature based method. Experimentally, it turned out that the line features are superior in our problem setup - the detection of textureless objects.
Our dataset consists of eight object classes, which are refrigerators, microwave, ovens, monitors, phones, printers, sofas, wall clocks, and wiry chairs. The images are taken in kitchens and hallways at CMU. The database is composed of 40 labeled training images for eight object classes (i.e., five images each object class), and 145 test images. Our recognition system is robust in the variation of object instances and illumination to some extent (Fig.2(a)(c)). It is generally successful for scenes with clutter and with only partial view of the target object (Fig.13(b)(d)). Fig.3 shows what causes the failures of detection of monitors. Apparently, monitors are quite similar to other objects which have rectangular shapes, especially microwave ovens. This kind of confusions occurs with other pairs of objects such as sofas/wiry chairs, and fridge/printers. Another explanation is that monitors are generally located on the table and other objects on the desktop could distract the system. Fig.3(a) is a good example of this.
 |
|
Figure 1. A dataset - 8 object classes
|
 |
 |
|
|
(a) Intra-class variation (b) In the cluttered scene (c)
Illumination variation (d) With only partial view |
 |
|
|
Figure 3. Examples of confusion - Similarly
looking objects.
|
Publication
1. Gunhee Kim, Martial Hebert, and Sung-Kee Park, "Preliminary Development of a Line Feature-based Object Recognition System for Textureless Indoor Objects", in Springer-LNCIS Publication "Recent Progress in Robotics: Viable Robotic Service to Human." S. Lee, I. Hong Suh, M. Sang Kim (eds.) (Selected papers from the 2007 International Conference on Advanced Robotics (ICAR 2007)), 2007.
Funding
- Intelligent Robotics Development Program, a 21st Century Frontier R&D Programs by the Ministry of Commerce, Industry, and Energy of Korea.
Copyright notice
The documents contained in these directories are included by the
contributing authors as a means to ensure timely dissemination of scholarly
and technical work on a non-commercial basis. Copyright and all rights
therein are maintained by the authors or by other copyright holders,
notwithstanding that they have offered their works here electronically.
It is understood that all persons copying this information will adhere
to the terms and constraints invoked by each author's copyright.