融合典型纹理特征的粒子滤波目标跟踪方法
吴刚;唐振民;杨静宇
【期刊名称】《计算机工程与应用》 【年(卷),期】2011(047)034
【摘要】Aiming at tracking windows' instability disturbed by noise on the process of image-object tracking, typical texture's characteristics are creatively integrated into the particle filter framework on the domain of real-time tracking.Three kinds of typical texture' s characteristics gray level co-occurrence matrix, breadth and direction texture, local binary model texture are tested on object tracking experiment.By doing contrastive experiment,the practical effect about texture's characteristics are measured and proved from view of precision and speed.As the choice about characteristics being tested in tracking,the results show the excellent quality of the gray level co-occurrence matrix texture which can timely resist analogous color's disturbance.%针对图像目标跟踪中,跟踪窗口易受噪声扰动所产生的跟踪不稳定问题.创造性地结合多种典型纹理特征和粒子滤波算法,将其应用于实时跟踪领域.选择3种典型的纹理特征,即灰度共生矩阵纹理特征、幅值与方向加权的梯度纹理特征、局部二进制模式纹理特征进行跟踪实验.通过对比实验,在精度与速度两方面测试纹理特征作为跟踪特征的实际效果.作为跟踪中特征的选择,实验结果说明灰度共生矩阵纹理信息在抗扰动与实时处理方面表现出良好的属性特征. 【总页数】5页(21-25)
融合典型纹理特征的粒子滤波目标跟踪方法
