CAJ | 학술논문

在视频车辆跟踪算法中针对传统粒子滤波的非线性、非高斯性可能导致跟踪过程的不准确性，提出一种基于Mean-Shift的卡尔曼（Kalman）粒子滤波算法。该算法利用建立基于目标颜色直方图特征模型对视频车辆目标进行建模，并将其与Kalman滤波相结合进行更新；通过采用Mean Shift算法将Kalman滤波器引用到粒子滤波器当中，通过预测迭代，从而达到对车辆的运行轨迹的修正。将先验信息预测与粒子滤波相结合在保持跟踪系统整体上的非线性、非高斯性，兼顾了卡尔曼滤波局部的线性高斯特性。实验结果表明，该方法与传统粒子滤波方法相比，具有较好的实时性和较高的准确率，能够准确稳定地对目标车辆进行跟踪。
재시빈차량근종산법중침대전통입자려파적비선성、비고사성가능도치근종과정적불준학성，제출일충기우Mean-Shift적잡이만（Kalman）입자려파산법。해산법이용건립기우목표안색직방도특정모형대시빈차량목표진행건모，병장기여Kalman려파상결합진행경신；통과채용Mean Shift산법장Kalman려파기인용도입자려파기당중，통과예측질대，종이체도대차량적운행궤적적수정。장선험신식예측여입자려파상결합재보지근종계통정체상적비선성、비고사성，겸고료잡이만려파국부적선성고사특성。실험결과표명，해방법여전통입자려파방법상비，구유교호적실시성화교고적준학솔，능구준학은정지대목표차량진행근종。
In video vehicle tracking algorithm,the nonlinear,non-Gaussian property in traditional particle filter may lead to inaccuracy intracking process.This paper puts forward a mean-shift-based Kalman particle filter algorithm to solve this problem.The algorithm models thevideo vehicle target by making use of building the target colour-based histogram feature model,and combines it with the Kalman filter forupdating;it also applies the Kalman filter to particle filter by using mean-shift algorithm,and achieves the correction of vehicles moving trackthrough prediction iteration.The method combines the priori information and the particle filtering on the overall nonlinear and non-Gaussianproperties of the tracking system,and takes into account the local linear Gaussian feature of Kalman filter as well.Experimental result showsthat this method has better real-time property and higher accuracy rate than the traditional particle filter methods,and is able to track thetarget vehicles accurately and stably.