Real-time object tracking based on colour feature and perspective projection

Abstract

This paper proposes a fast and robust real-time object tracking technique in a ball and beam system following the concept of rotating frame (to be called as rotating ball and beam system). The technique uses a standard contact free video camera which is placed at a static position and allows the sensor to capture the entire dynamic motion of the system. Since the captured image is in the world coordinates, instead of the rotating beam coordinates, the position of the object is easily extracted using either of the two projective transform methods such as Hue, Saturation and Value (HSV) or Hue, Saturation (HS). Experimental results on a prototype ball and beam system demonstrate that performance of HS based object localization, which ignores the light intensity, is better compared with HSV based method.