公开(公告)号：KR101261797B1

公开(公告)日：2013-05-07

申请号：KR1020110115430

申请日：2011-11-07

Applicant: 한국과학기술원

Inventor： 김대식 ,  임재현 ,  유재헌

IPC: G05D1/00 ,  G05D3/00

CPC classification number: G05D1/0214 ,  G06T7/11 ,  G06T7/20 ,  G06T2207/10016

Abstract: PURPOSE: A self-recognition method of an autonomous system is provided to implement accuracy by finding a boundary between a robot body and a background. CONSTITUTION: A visual input of an optical flow shape is obtained(S1). A velocity vector is extracted from an optical flow(S3). A vector to estimate the movement of a system body is extracted. A local feature set is extracted by selecting local features. A body is extracted from the local feature set(S4). [Reference numerals] (S1) Obtain a visual input of an optical flow shape expressed as a vector group of a local picture; (S2) Predict a joint position by using vectors and a free angle of a robot joint corresponding to the vectors; (S3) Extract a velocity vector group matched with prediction values for all frames of video stream; (S4) Extract a main body from the velocity vector group; (S5) Apply a segment algorithm

Abstract translation: 目的：提供一种自主系统的自我识别方法，通过发现机器人身体与背景之间的边界来实现准确性。 构成：获得光流形状的视觉输入（S1）。 从光流中提取速度矢量（S3）。 提取用于估计系统体的移动的向量。 通过选择本地特征来提取局部特征集。 从本地特征集提取一个身体（S4）。 [参考数字]（S1）获取表示为局部图像的矢量组的光流形状的视觉输入; （S2）通过使用与向量对应的机器人关节的向量和自由角度来预测关节位置; （S3）提取与视频流的所有帧匹配的预测值的速度矢量组; （S4）从速度矢量组提取主体; （S5）应用段算法