基于概率模型的行人四肢自遮挡的检测

针对人体运动跟踪领域中的自遮挡现象, 提出了一种基于概率模型的行人四肢自遮挡的检测算法. 首先, 该方法定义了基于马尔可夫模型的自遮挡状态概率模型; 其次, 利用椭圆肤色模型, 定义了层次肤色模型; 最后, 通过上述模型, 算法将行人四肢自遮挡状态的识别转换为计算自遮挡状态转换概率的过程. 实验表明, 该方法具有较高的准确性.

Estimating the Pedestrian 3D Motion Indoor via Hybrid Tracking Model

Focusing on the problem of self-occlusion in the field of human motion tracking, the paper deals with an algorithm for detecting the pedestrian limbs self-occlusion probability model. First, the algorithm defines the self-occlusion state probability model based on a Markov model. Second, the paper proposes a hierarchical skin model using ellipse skin model. Last, the algorithm changes the detection of pedestrian limbs self-occlusion to the calculation of the self-occlusion state transition probability. The result of experiment shows that the algorithm has a higher accuracy.