Real Time Hand Based Robot Control Using 2D/3D Images

摘要: In the interaction between man and machine, an efficient, natural intuitive commanding system plays a key role. Vision based techniques are usually used to provide such system. This paper presents new solution using 2D/3D images for real time hand detection, tracking classification which is as interface sending commands industrial robot. images, including low resolution range data high color information, provided by novel monocular hybrid vision system, called MultiCam, at video frame rates. After region extraction applying some preprocessing techniques, segmented unsupervised clustering approach. The image then mapped corresponding 2D image. Due setup of mapping 3D information trivial does not need any complicated calibration registration techniques. Consequently, segmentation becomes simple fast. Haar-like features extracted from input AdaBoost classifier find in track it each frame. found improved through postprocessing finally posture (palm fist) classified on very fast heuristic method. proposed approach has shown promising results application, even under challenging variant lighting conditions was demonstrated Hannover fair 2008.