STEREO VISION HEAD VERGENCE USING GPU CEPSTRAL FILTERING

• DOI: 10.5220/0003319406650670
• Publication Date: 2011-06-28T08:23:14Z
• Authors: Luís Almeida; Paulo Menezes; Jorge Dias
• Keywords: Vergence (optics)

Vergence ability is an important visual behavior observed on living creatures when they use vision to interact with the environment. The notion of active observer equally useful for robotic systems tasks like object tracking, fixation and 3D environment structure recovery. Humanoid robotics are a potential playground such behaviors. This paper describes implementation real time binocular vergence using cepstral filtering estimate stereo disparities. By implementing filter graphics processing unit (GPU) Compute Unified Device Architecture (CUDA) we demonstrate that robust parallel algorithms used require dedicated hardware now available common computers. algorithm speed up more than sixteen times current CPU. overall system implemented in IMPEP (IMPEP Integrated Multimodal Perception Experimental Platform) illustrate performance experimentally.