文章导读
总览 评价 朱永荣 , 苏杨 * ( 武汉理工大学信息工程学院,武汉 430070; ) 摘要: 关于人体姿态研究近些年来成为机器学习领域的热点话题。人体姿态的识别受到诸多外部条件的干扰,诸如气候,心情,肤色,体征差异等等。人体姿态的灵活性和多样性决定了要
朱永荣, 苏杨*
(
武汉理工大学信息工程学院,武汉 430070; )
摘要:
关于人体姿态研究近些年来成为机器学习领域的热点话题。人体姿态的识别受到诸多外部条件的干扰,诸如气候,心情,肤色,体征差异等等。人体姿态的灵活性和多样性决定了要对其进行研究就要花费时间进行大量的建模预估和计算能力。 目前来说,人体姿态的预估相对来说效果最好的算法就是基于图结构即PS(Picture Structure),这个算法的主要原理就是将人体不是视作一个整体,而是当作一个个组件相连接的模型。然而传统的PS算法缺点是对整体模型的表达可能会产生不清晰,鲁棒性较差。为了改进这个缺陷,本文提出了对相邻组件中的子组件相对于该相邻组件的空间相对位置来建模,也就是PS混合模型方法,然后进行实验得出该方法能提高预估的准确性。
关键词:
人体姿态;PS;混合模型
ZHU Yongrong, SU Yang*
(
WuHan university of technology,WuHan 430070; )
Abstract:
Research on human body posture in recent years become a hot topic in the field of machine learning。The human body gesture recognition by the interference of external conditions,such as climate, the mood, color of skin, and so on。The body posture of flexibility and diversity decided to study of its spends a lot of time to calculate the estimation to the modeling and capacity。 As matters stand,Body posture of the forecast effect is relatively the best algorithm is based on graph Structure or PS (Picture Structure),The main principle of this algorithm is not the human body as a whole, but as a connected component of the model。However, the disadvantage is that the traditional PS algorithm expression of whole model may produce is not clear, poor robustness。In order to improve the defect,in this paper to the adjacent components of child components relative to the space position of the adjacent components modeling,that is the PS hybrid model method,Then the experiment it is concluded that this method can improve the accuracy of the forecast.
Tag:
点此返回栏目查看更多>>>参考论文