Abstract

In human–robot coexisting environment, one of the primary objectives is to equip robots with the facilities of detecting and following a human being in front, using various sensors. Within this genre, developing algorithms for vision sensor-based shoe detection and subsequently following is considered an active problem. Considering that the shoe poses, during the pursuit, undergo different transformations, this paper presents how the vision sensor based shoe detection problem can be treated as analogous to template matching under general conditions. This paper first shows how a popular fast randomized template-matching algorithm, called FAsT-Match algorithm, and its contemporary variant for color images, called CFAsT-Match algorithm, can be implemented in real robots with success for detecting shoes in subsequent frames. Then, this paper proposes a new density-based clustering method, called ordering points to identify the clustering structure (OPTICS), based template-matching algorithm that is specifically developed to overcome several implementation problems associated with the FAsT-Match and CFAsT-Match algorithms, in real life. Experimental results and performance evaluations demonstrate the superiority of the proposed OPTICS-based algorithm for shoe detection during human tracking in human–robot coexisting environments.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call