摘要
对于当前热点的运动捕获方法存在的一些缺点,提出了一种融合深度图和三维模型的人体运动捕获方法。利用Kinect采集深度图像,经过对深度图去除背景,提取轮廓信息,建立轮廓数据库。从深度图中提取三维人体骨架,建立骨架三维模型数据库。输入1组深度图动作序列,经过去背景、提取轮廓特征后与轮廓数据库中的轮廓进行匹配,计算出最小距离所在的匹配序列,输出相应的骨架作为动作捕获的结果。实验证明了这种方法的有效性和可行性,该方法能较精确的得到运动捕获数据。
The method based on depth and 3Dmodel for motion capture is proposed in this paper for some disadvantages in the current method of motion capture.It collects the depth images by Kinect,then,it removes the background of depth image and extracts the contour information.After,it creates a database of contour.Then it extracts skeleton of human body through depth image so as to create 3D model database of skeleton.Finally,for a set of sequence of depth image with motion as inputting,it will match with the database of contour after the process of removing the background,extracting the contour information.Then it calculates the minimum distance for matching sequence,and outputs the corresponding skeleton as a result of the motion capture.The result of experiment shows the availability and feasibility of this approach,it also can get motion capture data more accurate.
出处
《国外电子测量技术》
2015年第1期19-22,共4页
Foreign Electronic Measurement Technology
基金
国家自然科学基金(61271362)项目
