MeanShift算法是一种无参概率密度估计法，算法利用像素特征点概率密度函数的梯度推导而得， MeanShift算法通过迭代运算收敛于概率密度函数的局部最大值，实现目标定位和跟踪，也能对可变形状目标实时跟踪，对目标的变形，旋转等运动也有较强的鲁棒性。MeanShift算法是一种自动迭代跟踪算法，由 MeanShift补偿向量不断沿着密度函数的梯度方向移动。在一定条件下，MeanShift算法能收敛到局部最优点，从而实现对运动体准确地定位。-MeanShift algorithm is a non-parametric probability density estimation method, the algorithm uses the pixel density gradient feature points obtained by derivation function, MeanShift iteration algorithm converges to a local maximum of the probability density function, location and tracking goals, but also to Object tracking variable shape, deformation, rotation and other moving targets but also robust. MeanShift iterative algorithm is an automated tracking algorithm, constantly moving　the MeanShift compensation vector along the direction of the gradient density function. Under certain conditions, MeanShift algorithm can converge to local optima, in order to achieve accurate positioning of the moving body.