基于混合高斯模型与camshift跟踪的智能视频监控技术研究

发布时间：2018-03-11 07:09

本文选题：智能监控　切入点：运动检测　出处：《中南大学》2014年硕士论文　论文类型：学位论文

【摘要】：近年来,随着计算机和图像处理技术的进步,智能视频监控技术发展迅速,对维护公共安全起到了巨大作用。运动目标检测和运动目标跟踪作为智能视频监控技术的核心一直是计算机视觉领域的研究热点。本文对运动目标检测与跟踪技术展开研究,并设计实现一个智能视频监控系统。本文首先介绍视频监控系统的相关背景和研究现状,阐述进行系统设计与实现的关键技术及其相关理论基础。随后对智能视频监控系统中的运动目标检测和跟踪技术进行了详细研究,针对传统算法的不足,结合实际运用场景提出有效的改进方案。最后对系统总体框架和详细模块设计进行介绍,并对系统运行结果进行详细分析。本文针对传统混合高斯算法使用场景受限、实时性差等不足,使用分块匹配以及跳帧处理的策略降低算法的复杂度,采用基于混合高斯算法和帧差法的双模型检测方法提高运动检测的可靠性。系统改进camshift算法提高运动目标跟踪效果。针对算法中存在的背景点干扰严重、只能手动选取目标以及算法迭代次数多、容易跟踪失败等问题,提出基于线性预测的camshift跟踪方法,通过进行帧差操作提取运动区域以及采用运动检测结果作为自动跟踪目标,减少了噪声和人为干预,提高了算法的智能性和准确性。
[Abstract]:In recent years, with the development of computer and image processing technology, intelligent video surveillance technology has developed rapidly. Moving target detection and moving target tracking, as the core of intelligent video surveillance technology, has been the research hotspot in the field of computer vision. An intelligent video surveillance system is designed and implemented. Firstly, this paper introduces the background and research status of video surveillance system. This paper expounds the key technology and related theoretical basis of system design and implementation. Then, the moving target detection and tracking technology in intelligent video surveillance system is studied in detail, aiming at the shortcomings of traditional algorithms. Finally, the overall framework and detailed module design of the system are introduced, and the results of the system operation are analyzed in detail. Aiming at the limitations of the traditional mixed Gao Si algorithm, such as limited scene and poor real-time performance, the complexity of the algorithm is reduced by using the strategy of block matching and frame hopping processing. The dual model detection method based on mixed Gao Si algorithm and frame difference method is used to improve the reliability of motion detection. The camshift algorithm is improved to improve the tracking effect of moving target. We can only select the target manually and the algorithm has many iterations, so it is easy to track the problems such as failure. The camshift tracking method based on linear prediction is proposed. The motion region is extracted by the frame difference operation and the motion detection result is used as the automatic tracking target. The noise and human intervention are reduced, and the intelligence and accuracy of the algorithm are improved.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN948.6

【共引文献】