提出一种基于级联结构的人体动作识别方法:针对Dollar时空兴趣点检测器易受图像噪声、摄像机运动与缩放等因素影响产生伪兴趣点的问题,提出了一种基于轨迹差异度的兴趣点筛选方法,有效避免了引入背景中的伪兴趣点,提高了人体运动特征提取的准确度；采用规范切与最小冗余最大相关(mRMR)准则对词袋模型生成的特征向量进行自动特征选择,同时建立一个用于分类的级联结构,在识别各类不同动作时选择不同的特征子集,使得分类器使用的特征更具区分性.在KTH人体运动测试集上实验,验证了该方法能提高动作识别的准确度.%A human action recognition method based on cascaded structure was proposed in this paper. Firstly, a trajectory-based method was proposed to select the interest points detected by the Dollar detector, which was sensitive to image noise, camera movement and zooming. Therefore, the pseudo interest points in the background could be effectively excluded and the extracted features would be more relevant to action recognition. Secondly, an automatic feature selection method based on the combination of normalized cuts and mRMR criteria was used to determine a subset of the words generated by the Bag-of-Words model and construct a cascaded structure for action recognition. The purpose was to make the feature used by the cascaded structure more distinct. Lastly, the experimental results validate the contribution to the improvement of accuracy in human action recognition.
展开▼
