基于复杂性度量与多尺度运动编码的图像动作识别算法

Action recognition algorithm based on complexity measure and multi-scale motion coding

人体动作的识别与理解是人机交互、机器人应用的关键技术之一,为了提高人体各种复杂动作的识别精度与鲁棒性,研究了基于复杂性度量与多尺度运动编码的动作识别技术。通过不同长度的滑动窗口对视频序列获取子序列;通过时间序列复杂性来度量人体运动轨迹,设计了一种多尺度的滑动窗口,从而选择出有效子序列;基于有效子序列,引入k-均值聚类分析算法,对人体运动进行编码,获取运动编码直方图;引入条件随机场对动作分类学习,完成动作识别与理解。所提出的算法在人机交互、智能家居、视频监控等领域具有较好的参考价值。

The recognition and understanding of human motion was one of the key technologies in human-computer interaction and robot applications,in order to improve the recognition accuracy and robustness of various complex motions of human body,a motion recognition scheme based on complexity measurement and multi-scale motion coding is studied.For a video sequence,the subsequence is obtained by sliding windows of different lengths.The trajectory of human body was measured by time series complexity,thus,meaningful subsequences were selected;A meaningful subsequence based on selection,the k-mean clustering algorithm was introduced for motion coding,generative motion coded histogram.Moreover,in order to solve the sensitivity problem of fixed length sliding windows,a multi scale sliding window generation motion coding was designed.The conditional random field was introduced to classify the actions,complete action recognition and understanding.Therefore,the proposed algorithm has a good reference value in human-machine interaction,smart home,video surveillance and other fields.