全文预览

智能机器人与多模态人机交互的研究-外文文献翻译

上传者:相惜 |  格式:doc  |  页数:37 |  大小:0KB

文档介绍
ng environmental circumstances that, for instance, make the skin color detection difficult or almost impossible. Fig. 3. Multiple-cue approach to user localization. For person localization and the subsequent tracking process, both camera systems (omnidirectional as well as frontally aligned stereo system) are utilized. Contour Modelling: The contour which we refer to is that of the upper part of the body ofa frontally aligned person. First, we generated a statistically determined average head-shoulder contour by collecting views of different people. Then, a model was learned based on this set of training images (see fig. 4). This simple contour shape prototype consists ofa course of orientations along the modelled contour and realizes a piecewise approximation of the upper shape ofa person

收藏

分享

举报
下载此文档