Compact Binary Map


Abstract-This paper is concerned with the problem of mobile robot localization using a novel compact representation of visual landmarks. With recent progress in lifelong map-learning as well as in information sharing networks, compact representation of a large-size landmark database has become crucial. In this paper, we propose a compact binary code (e.g. 32bit code) landmark representation by employing the semantic hashing technique from web-scale image retrieval. We show how well such a binary representation achieves compactness of a landmark database while maintaining efficiency of the localization system. In our contribution, we investigate the cost-performance, the semantic gap, the saliency evaluation using the presented techniques as well as challenge to further reduce the resources (#bits) per landmark. Experiments using a high-speed car-like robot show promising results.

 

Acknowledgement This work was partially supported by MECSST Grant in-Aid for Young Scientists (B) (19700192, 21700221), by KURATA grants and by TATEISI Science And Technology Foundation.

 

Members Tanaka Kanji, Saeki Kenichi, Ikeda Kouichirou, Kondo Kensuke

 

Relevant publications:

 

Visual robot localization using compact binary landmarks

Robotics and Automation (ICRA), 2010 IEEE International Conference on, 4397-4403
Kouichirou Ikeda, Kanji Tanaka

Bibtex source, Document PDF


How well compact binary landmark representation works in mobile robot localization? (a) Experimental environment and robot’s trajectory. (b) A high-speed car-like mobile robot. (c) Binary landmark representation using the semantic hashing technique. (d) Sequences of images and binary codes. Top, Middle: Two similar locations. Bottom: A dissimilar location. Note that the codes are similar/dissimilar only at a few bits.


Binary maps. Individual binary codes are viewed as different type independent measurements. We employ K different binary maps for K bit code and then record i-th bit measurements on i-th map. We also discuss how many K′(< K) binary maps are required for successful localization.

Frequency of visual words. Top: Average number of landmarks per word. Words are sorted in terms of the number of landmarks they contain and then grouped into 7 groups. Each datapoint from left to right respectively corresponds to the word as well as their near neighbors in terms of Hamming distance 1, 2 and 3. Bottom: Images corresponding to the central words (ID: 1, 5, 50 and 500) of each group.

Acknowledgement This work was partially supported by MECSST Grant in-Aid for Young Scientists (B) (19700192, 21700221), by KURATA grants and by TATEISI Science And Technology Foundation.

Members Tanaka Kanji, Saeki Kenichi, Ikeda Kouichirou, Kondo Kensuke