主要是对运动员动作的识别问题进行的研究.提出了基于MeanShift和肤色相结合的分割方法以及运动轨迹合成方法,创新性的将PCA算法中的K-L变换应用于对图像序列的运动轨迹的特征提取当中.以乒乓球运动为例,对运动员的接挡球和扣杀球这两个动作进行了实验验证.实验结果证明了K-L变换在最大程度的保留原有信息的同时,降低了原始特征数据的维数,减小了计算量,增大了原始数据的类间距离,提高了分类的准确性.
This paper research on the pattern recognition problem about athletes' action.A segmentation method based on MeanShift and skin color,and trajectory synthesis method is presented.This paper innovatively applied K-L transform of PCA algorithm to the trajectory feature extraction of image sequences.Take table tennis as an example,experiment on blocking the ball and smashing the bali.The experiment shows that K-L transform retains the original information in maximum.At the same time,it reduces the original dimension of feature data and reduces the amount of calculation.