Abstract
It is becoming important for working robots to be able to identify and pick objects in various tasks. As in the recent Amazon Picking Challenge, using a marker for the picking task is a more practicable approach. However, a common marker code for working robots does not exist so far. Conventional marker codes as represented by QR code or ARToolKit marker cannot be reliably detected from various viewpoints. Thus in this paper, we propose a new encoded marker which is flexibile to the marker's positions and blur. The proposed marker can be detected by an approach based on the scale space theory independent from such conditions. In addition, the representation of data by M-sequence makes the encoded marker robust to blur. Experimental results showed the effectiveness of the proposed marker compared to the ARToolKit marker. Since the marker is more robust against ground clutter noise, various positions of markers and blur, it is more practicable.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have