复杂场景下运动目标检测与跟踪技术的研究

发布时间：2019-01-10 20:50

【摘要】：随着计算机技术的迅速发展,智能视频监控技术在人们的日常生活中有着非常广泛地应用,同时给人们的日常生活带来了极大的便利,运动目标的检测与跟踪是智能视频监控技术中的关键性问题。运动目标的检测与跟踪是计算机视觉领域一个非常热门的课题,在人工智能、模式识别、图像处理、医学成像等多个领域有着非常广泛地应用,更是很多视觉应用系统的重要组成部分,如智能视频监控、人机交互、智能驾驶等。复杂场景下目标本身地变化以及背景环境地变化也是该技术面临的一项挑战。因此,对于复杂场景下运动目标的检测与跟踪的研究有着非常重要的研究价值和现实意义。本文的主要工作如下:(1)针对混合高斯模型在背景建模时采用固定的更新率,在复杂场景下对运动目标不能进行准确地检测,提出了一种改进的混合高斯算法,该算法在混合高斯模型的基础上结合改进的帧间差分法,将背景划分为不同区域,同时对不同区域选取适合的更新率,使得背景模型能够更好的适应复杂场景中的外界干扰。实验结果表明,本文提出的改进的混合高斯算法能够适应复杂场景下光照变化、树木遮挡等外界环境的干扰,提高了检测的准确性。(2)针对粒子滤波在重采样过程中舍弃低权值的粒子,所造成的粒子退化问题,对粒子滤波进行了改进,提出了一种基于萤火虫算法的粒子滤波算法。该算法在重要性采样过程中,使用萤火虫算法对粒子进行迭代寻优,使得粒子更接近后验概率分布,在重采样过程中利用萤火虫算法思想,使粒子模拟萤火虫向高似然区域移动,并更新全局最优值,提高粒子的有效性及多样性。实验结果表明,在背景发生变化、不规则的运动、以及树木的遮挡等多因素的影响下,本文提出的算法能够使用更少的粒子数目,进行高效、精确的目标跟踪。(3)本文在VS2010环境下,采用MFC界面类库和计算机视觉类库OpenCV,设计并实现了具有视频输入、图像预处理、运动目标检测、形态学处理、目标跟踪以及目标轨迹绘制等功能的视频监控系统。其中目标检测以及目标跟踪模块分别使用本文所提出的检测和跟踪方法。
[Abstract]:With the rapid development of computer technology, intelligent video surveillance technology has been widely used in people's daily life, at the same time, it has brought great convenience to people's daily life. Detection and tracking of moving targets is a key problem in intelligent video surveillance technology. Moving target detection and tracking is a very hot topic in the field of computer vision. It is widely used in artificial intelligence, pattern recognition, image processing, medical imaging and other fields. It is also an important component of many visual application systems, such as intelligent video surveillance, human-computer interaction, intelligent driving and so on. It is also a challenge to change the target itself and the background environment in complex scenarios. Therefore, the research of moving target detection and tracking in complex scene has very important research value and practical significance. The main work of this paper is as follows: (1) aiming at the fixed updating rate of the mixed Gao Si model in the background modeling, the moving target can not be detected accurately in the complex scene, so an improved mixed Gao Si algorithm is proposed. Based on the mixed Gao Si model and the improved inter-frame difference method, the background is divided into different regions, and the suitable updating rate is selected for the different regions, so that the background model can better adapt to the external interference in the complex scene. The experimental results show that the improved mixed Gao Si algorithm proposed in this paper can adapt to the external environment interference, such as illumination variation, tree occlusion and so on. The accuracy of detection is improved. (2) aiming at the problem of particle degradation caused by particle filter abandoning low weight particles in the process of resampling, the particle filter is improved, and a particle filter algorithm based on firefly algorithm is proposed. In the process of importance sampling, the algorithm uses the firefly algorithm to iteratively optimize the particle, so that the particle is closer to the posteriori probability distribution, and the idea of the firefly algorithm is used in the process of resampling. The particle simulates the firefly to move to the high likelihood region, and updates the global optimum value, enhances the particle validity and the diversity. The experimental results show that under the influence of many factors, such as background change, irregular motion and tree occlusion, the proposed algorithm can use fewer particles and be efficient. (3) in the environment of VS2010, this paper designs and implements video input, image preprocessing, moving target detection, morphological processing by using MFC interface class library and computer vision class library OpenCV,. Video surveillance system for target tracking and target trajectory rendering. Target detection and target tracking module respectively use the detection and tracking methods proposed in this paper.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】