改进的运动目标检测与跟踪算法在嵌入式平台上的研究

发布时间：2018-09-03 09:00

【摘要】：近年来,随着计算机视觉技术、嵌入式技术和通信技术的迅猛发展,智能视频监控技术在社会中诸多领域得到了越来越广泛的应用。运动目标检测与跟踪是智能视频监控技术的核心技术之一,同时也是实现智能视频监控应用的基础,目前正朝向结合嵌入式技术和无线通信网络技术不断进步,对它无论是在科学研究上还是实际应用上都具有良好的前景,因而对在嵌入式平台上进行智能视频监控系统的研究与开发已成为了广大研究人员的研究热点。本文主要针对静止场景下的运动目标检测与跟踪技术进行深入的研究,并对传统算法存在的问题设计出改进的运动目标检测与跟踪算法。本文的主要研究工作如下：在运动目标检测方面,首先对目前常用的三种运动目标检测方法：光流法、背景差分法和帧间差分法进行简要分析,其中重点讨论帧间差分法和背景差分法的优缺点,之后分别对两种算法进行改进,并引入边缘检测算子,对其的选取进行了客观评价,设计了一种基于五帧差分和背景边缘检测差相结合的运动目标检测算法。最后通过实验结果证明,本文算法不仅能够快速地提取完整、准确的运动目标轮廓还可以消除阴影和空洞现象,为后续的运动目标跟踪奠定基础。在运动目标跟踪方面,首先介绍了不同的运动目标跟踪算法,重点研究了传统均值漂移运动跟踪算法的理论和跟踪原理,其次针对传统均值漂移跟踪算法所存在的两点不足：跟踪快速运动目标和出现严重遮挡时容易造成目标丢失等问题,通过确定目标形心和监测巴氏系数的变化,设计了基于卡尔曼滤波和均值漂移跟踪算法相结合的运动目标跟踪算法来解决上述不足。最后通过实验证明,该算法跟踪的准确性和稳定性相对于传统均值漂移跟踪算法跟踪效果有了显著提高。在系统功能实现方面,首先在开发板中进行了嵌入式操作系统、计算机开源视觉库OpenCV和图形界面库Qt的移植,之后对算法实现进行了模块化的软件设计,并通过相关程序的编写在嵌入式平台上实现了运动目标的检测与跟踪。实验结果证明,本文中改进的运动目标检测与跟踪算法,提高了算法在复杂环境下的适应能力,可以实现对运动目标的快速、准确的检测和稳定的跟踪,具有广泛的应用前景。
[Abstract]:In recent years, with the rapid development of computer vision technology, embedded technology and communication technology, intelligent video surveillance technology has been more and more widely used in many fields of society. Moving target detection and tracking is one of the core technologies of intelligent video surveillance technology, and it is also the basis of intelligent video surveillance application. At present, moving target detection and tracking is moving towards the integration of embedded technology and wireless communication network technology. The research and development of intelligent video surveillance system based on embedded platform has become a hot topic for many researchers because it has a good prospect in both scientific research and practical application. In this paper, the technology of moving target detection and tracking in static scene is studied deeply, and an improved moving target detection and tracking algorithm is designed for the problems existing in traditional algorithms. The main research work of this paper is as follows: in the aspect of moving target detection, firstly, three commonly used moving target detection methods: optical flow method, background difference method and inter-frame difference method, are briefly analyzed. The advantages and disadvantages of the inter-frame difference method and the background difference method are discussed, and then the two algorithms are improved, and the edge detection operator is introduced to evaluate their selection objectively. A moving target detection algorithm based on five frame difference and background edge detection difference is designed. Finally, the experimental results show that the proposed algorithm can not only quickly extract the complete and accurate contour of moving objects, but also eliminate the shadow and void phenomena, and lay the foundation for the follow-up tracking of moving targets. In the aspect of moving target tracking, different moving target tracking algorithms are introduced, and the theory and principle of the traditional mean drift motion tracking algorithm are studied. Secondly, aiming at the two shortcomings of the traditional mean shift tracking algorithm: tracking fast moving target and losing the target easily when there is serious occlusion, the change of target centroid is determined and the change of pasteurian coefficient is monitored. A moving target tracking algorithm based on Kalman filter and mean drift tracking algorithm is designed to solve the above problems. Finally, it is proved by experiments that the accuracy and stability of the algorithm is significantly improved than that of the traditional mean shift tracking algorithm. In the aspect of system function realization, the embedded operating system, computer open source vision library OpenCV and graphical interface library Qt are transplanted in the development board first, and then the modular software design of the algorithm implementation is carried out. The detection and tracking of moving targets are realized on the embedded platform through the programming of related programs. The experimental results show that the improved algorithm of moving target detection and tracking in this paper improves the adaptability of the algorithm in complex environment, and can achieve fast, accurate detection and stable tracking of moving targets, and has a wide application prospect.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41;TP368.1

【相似文献】