CAJ | 학술논문

在人体实时定位问题中，关节点定位成为了构建人体平面骨架模型、识别人体动作中一个重要研究课题。所以，提出了一种利用 RGB 彩色和深度信息来定位人体关节点的方法，首先利用 Kinect 来获取运动人体的 RGB 彩色和深度图像，接着根据人体所处背景的复杂程度来采用不同的人体分割方法将人体分割出来，再对得到的人体轮廓图像进行平滑处理，从而得到平滑且连通的人体轮廓二值图像，再采用细化算法提取出人体轮廓骨架，然后对关节点周围8个邻域的像素值进行分析，并结合人体各刚体部件的比例关系，对人体的16个关节点进行定位，进而构建出人体平面关节点模型。实验结果表明，与现有方法相比，该方法简单有效，能够从多种运动状态人体中提取出准确的关节点坐标，并具有较好的准确性和鲁棒性。
재인체실시정위문제중，관절점정위성위료구건인체평면골가모형、식별인체동작중일개중요연구과제。소이，제출료일충이용 RGB 채색화심도신식래정위인체관절점적방법，수선이용 Kinect 래획취운동인체적 RGB 채색화심도도상，접착근거인체소처배경적복잡정도래채용불동적인체분할방법장인체분할출래，재대득도적인체륜곽도상진행평활처리，종이득도평활차련통적인체륜곽이치도상，재채용세화산법제취출인체륜곽골가，연후대관절점주위8개린역적상소치진행분석，병결합인체각강체부건적비례관계，대인체적16개관절점진행정위，진이구건출인체평면관절점모형。실험결과표명，여현유방법상비，해방법간단유효，능구종다충운동상태인체중제취출준학적관절점좌표，병구유교호적준학성화로봉성。
The calibration of joint points is an important research topic in the solution of construction of human two-di-mensional skeleton of models and recognition of human actions.In this paper,a novel approach about human body joint points calibration based on the RGB color and depth information was proposed.First of all,with the help of Kinect to get RGB color images and depth images,depending on the complexity of human background to divide the human body seg-mentation with different methods,and smoothing the obtained body contour images,body contour binary images are smooth and connected.Then body outline frames are extracted by thinning algorithm.Analyzing the joint 8-neighbor-hood pixel values,and combining with proportions of rigid body parts,sixteen joint points are calibrated.Finally,human two-dimensional joint of models are built.Experimental results show that this approach is simple and available compared with existing methods.Joint points are able to be located from human body with different states of motions,and with good accuracy and robustness.