introduction about the research on Honda's humanoid robot ASIMO
including some of the latest videos demonstrating ASIMO's capabilities.
Over the past 16 years Honda has been developing a series of biped,
humanoid robots. ASIMO, which is the latest model of Honda's humanoid
robots, is 120 cm tall, weighs around 50 kilos and has 26 DOF.
The vision system of ASIMO consists of a pair of color CCD cameras and
a Pentium III platform for image processing. ASIMO is capable of
recognizing gestures, avoiding obstacles and identifying faces.
In the second part I will present a hierarchical approach for
detecting objects in images which is about to be implemented on ASIMO.
On the first level of the hierarchy component classifiers locate parts
of the object in the image. On the second level, a combination classifiers
checks if the located parts are in the proper geometrical configuration. The main advantages of the component-based approach compared to the global approach, in which the object is detected
by a single classifier, are higher robustness against pose changes
and partial occlusions. The main difficulty in component-based detection is the proper choice of a set of components.
I will discuss an algorithm which iteratively learns a set of
components based on an error bound of the component classifiers.
In experiments I will show the results of the algorithm applied
to face detection and recognition.