基于视频跟踪系统的卡尔曼滤波器方法研究

Research of a Visual Tracking System Using the Kalman Filter

针对运动物体速度以及形状容易发生变化,导致目标跟踪失败等问题,本文提出了一种基于卡尔曼滤波模型,同时考虑跟踪状态的马尔科夫性设计而成新的跟踪算法。算法首先建立目标状态和观测值的转移变化矩阵模型,然后依据马尔科夫性简化传统卡尔曼滤波算法模型,对目标方位和速度进行预判断。在此基础上,结合传统模板匹配和更新机制,在预测范围内搜索目标,并依据目标变化等因素更新模板的选择,从而保证在快速搜索目标的同时动态地调整模板,确保跟踪目标在发生形变或者加速等状态下能够实现稳定跟踪。实验结果验证了本文算法的有效性和实用性。

As the speed of moving objects, and shape are easy to change, leading to the failure of target tracking problems,the paper proposes a model based on the Kalman filter, taking into account the state of the Markov track design from the new tracking algorithm. The algorithm first creates a matrix model of the target states and observation changes in transfer, then simplifies the model of the traditional Kalman filter algorithm based on the Markov chain, and pre-determines the target position and speed. On this basis, the paper combines with the traditional template matching and updating mechanisms, within the search target in the forecast, and updates based on the objective factors such as changes in the choice of templates, thus ensuring quick search target, while in the dynamic adjustment of the templates, we make sure to track the target in the event of deformation or acceleration which can be achieved under steady state tracking. The experimental results demonstrate the effectiveness and practicality of the algorithm.