论文部分内容阅读
为解决运动图像中人体关节自动快速定位问题 ,提出了一种体育运动序列图象中的人体关节定位的快速算法 ,从运动序列图象中提取运动信息。运用序列图象前帧图象中关节位置的先验知识对运动关节在前后帧中进行块匹配 ,得到后帧图像中人体关节的位置 ,继而计算出科学训练中教练员所关心的一些运动参数。其中许多前提条件和假设都是基于高速摄像机或摄影机拍摄的图象的。在分析过程中 ,首帧图像的关节位置由人工输入 ,作为先验知识 ,以后的分析过程由计算机自动进行 ,分析结果迅速准确。作为应用实例 ,对棒球运动和跳高运动的技术录象资料的有关参数进行了提取和分析。
In order to solve the problem of automatic rapid positioning of human joints in moving images, a fast algorithm of human joint positioning in sports sequence images is proposed. The motion information is extracted from the moving sequence images. Using the prior knowledge of the joint position in the frame image before the sequence image, the motion joint is block-matched in the front and the back frame to obtain the position of the human joint in the image of the back frame, and then some exercise parameters concerned by the trainer in the scientific training are calculated . Many of these prerequisites and assumptions are based on images taken with high-speed cameras or video cameras. In the analysis process, the joint position of the first frame of image is manually input as a priori knowledge, and the subsequent analysis process is automatically performed by a computer, and the analysis result is prompt and accurate. As an application example, the relevant parameters of technical video recording of baseball and high jump were extracted and analyzed.