Font Size: a A A

Study On The Visual And Olfactory System About The Head Of Humanoid Robot

Posted on:2007-04-09Degree:MasterType:Thesis
Country:ChinaCandidate:Y H LiuFull Text:PDF
GTID:2178360182485346Subject:Mechanical Manufacturing and Automation
Abstract/Summary:PDF Full Text Request
Research and applications on robotic field have been extended continuously, thereinto the research and applications on humanoid robot come to front especially and become one of the most active study points in intelligent robot field. The human head plays a very important role in perception and expressioa Aiming at the input and output functions of human head, this thesis studies the perception system of humanoid robot head, which enables it to imitate vision and olfaction of human being, and react accordingly. With research over the visual and olfactory system of humanoid robot head, we hope to simulate the perceptive abilities of human being over the external world, and construct an independent learning platform based on vision skill.In the study of humanoid vision, automatic face detection and tracking are performed. Several colorspaces are mentioned, and skin models and segmentation arithmetic in these colorspaces have been described in the paper. Three linearity colorspace models: RGB,HSI, YCbCr and their interrelation are described. The differences between these skin models are revealed by the experiment. Using corresponding colorspaces in different illumination condition, the best project for segmenting skin and non-skin region from color images input is put forward, and the project provides a reliable and strong foundation for the face detection that based on skin infonnation. Aiming at the influence of speed by large mount of information after skin segmentation, the paper brings out an average partition method. This method reduces the quantitiy of information that needs to be processed and get rid of the noise in images. Invariant Moments is used and has played an important role in judging the ellipse model and obtaining the center of human face. We obtain a satisfying speed (8 frames per second) when real-time tracking. Matching units and classes are mentioned. Based on the character of images in stereo vision, the paper selects the edge points that can represent character of scenery and adjacent character based on gray degree as matching units. At the same time, rapid correlation matching arithmetic such as rank arithmetic and census arithmetic are introduced.In the study of humanoid olfaction, several kinds of typical gases are selected as the object of humanoid robot olfaction for recognization. By using high-quality gas sensors array simulate nose, gas information is collected and then magnified, converted, A/D transformed,and transmited to computer. Finally, these gases are distinguished by computer in order to achieve humanoid olfactioa...
Keywords/Search Tags:humanoid head, vision, human face tracking, olfaction
PDF Full Text Request
Related items