提出一种跟踪单眼图像序列中的行人 ,并恢复其运动参数的新方法 .在跟踪中采用了基于SPM (ScaledPrismatModel)扩展的二维纸板人模型取代三维人体模型 ,以获取更快的计算速度 .作者使用EM算法在概率框架下进行运动估计 ,同时 ,算法也考虑了混合的运动模型和运动约束 ,以减小解的搜索空间 .试验结果证明了该方法的有效性 A new method to track pedestrians in single-eye image sequences and restore their motion parameters is proposed, in which a two-dimensional cardboard human model based on SPM (Scaled Prism Matodel) expansion is used instead of three-dimensional human model to obtain faster calculation speed. The author uses the EM algorithm to estimate the motion in the framework of probability, meanwhile, the algorithm also considers the mixed motion model and the motion constraint to reduce the search space of the solution.The experimental results prove the effectiveness of the method