运动目标跟踪算法及应用研究

发布时间：2019-01-10 16:07

【摘要】：首先本文具体介绍了运动目标识别与跟踪方法在国内外的发展与研究的状况,并且对相关的算法做了简要的说明。然后针对目标识别与跟踪方法,本文分别详细介绍和说明了粒子滤波跟踪算法与Camshift跟踪算法,并对两种跟踪算法分别做了相应的对比跟踪实验来说明各自的特性及其优缺点。对两种跟踪算法所适用的跟踪环境进行了分析,当目标与背景颜色差异比较大时,粒子滤波跟踪算法和Camshift算法能够有效的跟踪目标,但是当目标与背景颜色差异较小或者目标处于复杂背景区域时目标跟踪就会产生偏差,甚至无法准确的跟踪目标。为提高复杂背景下上述两种跟踪算法的稳定性与准确性,在两种基本算法的基础上,分别提出了各自的改进算法来改善其在复杂背景下的跟踪性能。提出基于显著性直方图模型的粒子滤波跟踪方法。通过对比目标与背景区中像素色调的分布,确定出不同色调等级的显著性权值,从而建立起目标的显著性直方图模型。显著性直方图模型可抑制背景中与目标具有相似色调的区域对目标识别的干扰,突出目标显著色调在目标识别中的作用,从而提高目标识别的准确性。提出了基于边缘抑制的Camshift跟踪算法。利用上一帧目标的位置和大小通过权值函数,在反向投影图中降低目标边缘的亮度权值,被抑制的边缘可以有效的区分目标和背景,削弱质心向背景方向迭代的趋势,提高目标识别的准确性。仿真实验结果表明,本文提出的两种算法都能改善目标跟踪的准确性和稳定性,且计算量增加不多,能够满足电视跟踪系统实时性的要求。最后将本文提出的两种改进跟踪算法应用到智能小车的跟踪中,实验表明本文提出的改进算法在实际应用中可以获得较好的跟踪效果。
[Abstract]:Firstly, this paper introduces the development and research status of moving target recognition and tracking methods at home and abroad, and gives a brief description of the relevant algorithms. Then, the particle filter tracking algorithm and the Camshift tracking algorithm are introduced and explained in detail, and the two tracking algorithms are compared with each other in order to explain their characteristics and advantages and disadvantages. This paper analyzes the tracking environment of the two tracking algorithms. When the color difference between the target and the background is large, the particle filter tracking algorithm and the Camshift algorithm can effectively track the target. However, when the color difference between the target and the background is small or the target is in the complex background area, the target tracking will produce deviation, and even can not track the target accurately. In order to improve the stability and accuracy of the above two tracking algorithms in complex background, based on the two basic algorithms, their respective improved algorithms are proposed to improve their tracking performance in complex background. A particle filter tracking method based on significant histogram model is proposed. By comparing the distribution of pixel hue in the target and background region, the significance weights of different hue levels are determined, and the significance histogram model of the target is established. The significant histogram model can suppress the interference of the region with similar hue to the target recognition in the background and highlight the role of the significant hue of the target in target recognition so as to improve the accuracy of target recognition. A Camshift tracking algorithm based on edge suppression is proposed. By using the position and size of the object in the previous frame through the weight function, the brightness weight of the edge of the object is reduced in the reverse projection, and the suppressed edge can effectively distinguish the object from the background, and weaken the tendency of centroid iterating towards the background. Improve the accuracy of target recognition. The simulation results show that the two algorithms proposed in this paper can improve the accuracy and stability of target tracking, and the amount of computation is not much increased, which can meet the real-time requirements of TV tracking system. Finally, the two improved tracking algorithms proposed in this paper are applied to the tracking of intelligent cars. The experimental results show that the proposed improved tracking algorithm can obtain better tracking results in practical applications.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】