J.L. Buessler, J.P. Urban, G. Hermann, and H. Kihl
Colour histogram similarity, colour histogram footprint, multi-coloured objects, object tracking, histogram intersection
This paper evaluates the potential of colour histogram techniques
for the approach phase in the visual servoing of a camera mounted
on a robot arm. We propose two complementary methods to detect,
locate, and estimate the size of a target object in video images in
order to extract basic visual features for robotic control.
The colour histogram similarity method identifies the object
and its location by comparing the colour histogram of a search
window with the histogram of a reference object. It can be used
to estimate the size of the object which varies strongly when the
camera approaches or moves away from the object. However this
estimation is coarse and computationally intense.
We introduce a new technique, the colour histogram footprint
that creates a binary image where the outline of the object stands
out, like a footprint. Using this imprint, it becomes easy to
determine the precise size and location of the target object.
The combination of these two techniques permits tracking of
video target objects in real time.