基于置信度策略选择的实时目标跟踪方法
发布时间：2018-11-16 08:01
【摘要】：目标跟踪一直是计算机视觉领域研究的热点和难点,受自然场景中复杂干扰因素影响,现有方法的速度和精度尚待改善。首先对基于颜色属性的目标跟踪算法进行改进,使之更为鲁棒且速度达到实时;接下来,针对被跟踪目标发生遮挡时,采用基于颜色属性的跟踪算法导致错误累积进而产生漂移甚至跟踪失败的问题,引入运算量较大但对遮挡有较强抵抗能力的稀疏协作表观模型。为了同时保证算法的速度和准确性,构建了一套基于跟踪结果置信度评价的策略选择机制,将两种算法有机整合。在多个公开数据集下的对比实验显示,与现有跟踪算法相比,该方法在跟踪效果和速度上具有较显著优势,并在目标存在严重遮挡、光照变化、运动模糊等情况时,均可以取得较好的跟踪效果。
[Abstract]:Target tracking has always been a hot and difficult point in the field of computer vision. The speed and accuracy of the existing methods need to be improved due to the influence of complex interference factors in the natural scene. Firstly, the target tracking algorithm based on color attribute is improved to make it more robust and real-time. Then, when the target is occluded, the color attribute based tracking algorithm leads to the error accumulation and the drift or even the tracking failure. This paper introduces a sparse cooperative surface model with large computation and strong resistance to occlusion. In order to ensure the speed and accuracy of the algorithm, a strategy selection mechanism based on the confidence evaluation of tracking results is constructed, and the two algorithms are integrated organically. Compared with the existing tracking algorithms, this method has significant advantages in tracking effect and speed, and when there is serious occlusion, light change, motion blur, etc. Good tracking effect can be obtained.
【作者单位】： 西南交通大学信息科学与技术学院;
【基金】：国家自然科学基金资助项目(61003143) 四川省科技支撑计划资助项目(2012FZ0004)
【分类号】：TP391.41
[Abstract]:Target tracking has always been a hot and difficult point in the field of computer vision. The speed and accuracy of the existing methods need to be improved due to the influence of complex interference factors in the natural scene. Firstly, the target tracking algorithm based on color attribute is improved to make it more robust and real-time. Then, when the target is occluded, the color attribute based tracking algorithm leads to the error accumulation and the drift or even the tracking failure. This paper introduces a sparse cooperative surface model with large computation and strong resistance to occlusion. In order to ensure the speed and accuracy of the algorithm, a strategy selection mechanism based on the confidence evaluation of tracking results is constructed, and the two algorithms are integrated organically. Compared with the existing tracking algorithms, this method has significant advantages in tracking effect and speed, and when there is serious occlusion, light change, motion blur, etc. Good tracking effect can be obtained.
【作者单位】： 西南交通大学信息科学与技术学院;
【基金】：国家自然科学基金资助项目(61003143) 四川省科技支撑计划资助项目(2012FZ0004)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 尤小泉;彭映杰;;一种基于指令预测的目标跟踪方法[J];电视技术;2010年02期
2 邵文坤;黄爱民;韦庆;;目标跟踪方法综述[J];影像技术;2006年01期
3 梁德群;阮文;;基于模型的线性组合目标跟踪方法[J];模式识别与人工智能;1995年04期
4 王思，
本文编号：2334931
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2334931.html
