CAJ | 학술논문

为了提高背景混淆下的跟踪效果,提出了基于谱直方图和Kalman滤波的目标跟踪方法.通过将图像与滤波器组进行卷积计算获取图像的频率特征,统计获得图像的谱直方图.谱直方图将图像转移到频域进行统计,减少了颜色以及噪声等因素的干扰.为了提高跟踪效率,引入Kalman滤波运动估计,进行局部搜索运动目标,并详细探讨了Kalman参数的设置.实验结果表明,在背景混淆的场景下,相对于颜色直方图来说,该算法能够取得较好的跟踪效果.
위료제고배경혼효하적근종효과,제출료기우보직방도화Kalman려파적목표근종방법.통과장도상여려파기조진행권적계산획취도상적빈솔특정,통계획득도상적보직방도.보직방도장도상전이도빈역진행통계,감소료안색이급조성등인소적간우.위료제고근종효솔,인입Kalman려파운동고계,진행국부수색운동목표,병상세탐토료Kalman삼수적설치.실험결과표명,재배경혼효적장경하,상대우안색직방도래설,해산법능구취득교호적근종효과.
To improve the tracking performance with scene complicated backgrounds, a algorithm of object tracking based on spectral histogram and Kalman filter is proposed. To obtain the image' s frequency characteristic, firstly convoluted with image, and then computed its spectral histogram. Spectral histogram is computed on the frequency domain, so the color and noise interference is reduced. To save the cost of time, Kalman filter is adopted to predict object' s location, in order to search object on local area. Finally, th eparametric setting of Kalman is discussed. The results indicated this algorithm do better than color histogram on the scene with complicated backgrounds.