A New Dynamic Edge Detection toward Better Human-Robot Interaction

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Robot’s vision plays a significant role in human-robot interaction, e.g., face recognition, expression understanding, motion tracking, etc. Building a strong vision system for the robot, therefore, is one of the fundamental issues behind the success of such an interaction. Edge detection, which is known as the basic units for measuring the strength of any vision system, has recently been taken attention from many groups of robotic researchers. Most of the reported works surrounding this issue have been based on designing a static mask, which sequentially move through the pixels in the image to extract edges. Despite the success of these works, such statically could restrict the model’s performance in some domains. Designing a dynamic mask by the inspiration from the basic principle of “retina”, and which supported by a unique distribution of photoreceptor, therefore, could overcome this problem. A human-like robot (RobovieR-2) has been used to examine the validity of the proposed model. The experimental results show the validity of the model, and it is ability to offer a number of advantages to the robot, such as: accurate edge detection and better attention to the front user, which is a step towards human-robot interaction.