【摘要】 针对Kinect原始识别算法在观测条件不理想的情况下容易出现错误识别的问题，提出了一种检测关节信息可信度及修复关节点信息的方法：通过对比人体基本结构以及根据运动连续性原理可以鉴别出错误的关节点信息，并通过之前帧的相关信息对当前帧的信息进行修复。通过实验验证，该方法可有效检测出错误的关节点信息并进行修复，提高Kinect人体识别的准确率。更多还原

【Abstract】 Because of the powerful function of Kinect in human body recognition, the human recognition technology based on Kinect has become a hot topic in recent years.But when applying Kinect to other fields, it is impossible to ensure the ideal observation condition and the Kinect’s original algorithm may get inaccurate joint information casually if the observation condition is not ideal.To solve the problem, this paper proposes a method of judge the credibility and repair the joint information.By comparing the basic structure of human body and the principle of motion continuity, we can identify the wrong joint information and repair the information of the current frame through the relevant information of the previous frames.The experimental results show that the method can effectively detect and repair the inaccurate joint information and improve the accuracy of human body recognition.更多还原