This paper presented a method to locate rigid object and estimate its pose in monocular image sequences accurately.The particle filter,also known as sequential Monte Carlo method,is an efficient algorithm dealing with Bayesian inference,especially motion track.In this paper,we combined the affine parameters and their velocities as the measurement vector of the particle filter.And the weights of the particles were calculated through the shape similarity between the object's contour model and the edge in each frame.And the parameters of the geometric transformation can be regarded as the location and pose of the target.Experiments show that our method can locate object accurately with an average relative error of 7%.