摘要
研究视频图像中准确实现运动目标跟踪问题,要求在视频图像中找到目标确切位置,并反馈给跟踪系统。针对传统基于特征匹配的跟踪,当被跟踪物体所处环境中存在颜色、形状接近的物体时,会出现像素特征误匹配,造成运动目标跟踪错误率较高。提出基于Markov Chain Monte Carlo数据关联的运动目标跟综方法。通过建立像素概率模型,将运动目标跟踪问题公式化,运用MCMC方法对后验概率进行采样估计,避免了模型匹配像素点的不确定性。实验证明,运动目标跟踪方法实现了在与自身相似背景下的准确跟踪,有效降低了跟踪错误率,取得了满意的效果。
Researching the target tracking of the video images.In traditional feature matching based tracking method,when the tracked object is similar in colors and shapes to other objects in the environment,,the pixel features mismatch can appear,which causes higher target tracking error rates.This paper puts forward a moving object tracking method based on Markov Chain Monte Carlo data relating.By establishing pixel probability model,moving target tracking is formulazed,and by using the method of MCMC to estimate the posteriori probability,the uncertainty of model matching pixel is avoided.Experiments have proved that this moving object tracking method can realize accurate tracking under the background similar to the moving object,reduce the tracking error rate,and achieve satisfactory results.
出处
《计算机仿真》
CSCD
北大核心
2011年第4期299-303,共5页
Computer Simulation
关键词
概率模型
运动目标跟踪
概率估计
Probabilistic model
Moving object tracking
Probability estimates