智能视频监控中的运动目标检测相关技术研究

发布时间：2018-08-24 11:32

【摘要】：智能视频监控技术的研究属于近些年来在计算机视觉领域新兴的方向。它主要的研究目标是通过计算机视觉技术、图像视频处理技术和人工智能技术,对监控视频的内容进行描述、分析和理解,同时根据分析处理所得的结果对监控系统进行控制,进而使得视频监控系统能够满足人们对于智能化的要求水平。它的主要研究内容包括:监控视频中运动物体的检测、跟踪、识别和行为分析等。本文主要的研究内容为智能视频监控中的运动目标检测提取方法。针对传统的运动目标检测诸多方法中经常出现的易受光照变化、复杂背景、阴影等因素影响的问题,提出了一种由混合高斯模型、边缘检测法与连续帧间差分法三种算法相结合的运动目标检测算法。该算法通过混合高斯模型在时间域上进行背景的建模与更新,在空间域上利用由边缘检测算法、连续帧间差分法以及混合高斯模型相结合的检测算法得出初始的运动目标轮廓,并且经过后续的运算处理,得到完善的所需运动物体。该算法不仅能够很好的适应所处场景中的背景干扰与渐变的光照条件,而且能够克服传统算法中对于目标检测不准确、边缘检测不完整、容易产生空洞和重影等问题的发生。实验结果显示该运算方法复杂度相对适中,具有比较好的实时性和鲁棒性,对运动物体检测的精确度较高。运动目标检测是智能视频监控中的一个重要环节,而运动目标的阴影检测又是运动物体检测的一个重要步骤。对于目标阴影检测的正确与否将直接影响到对目标物体的检测结果。通过对各种阴影检测方法的学习与研究,我们发现仅仅通过一种特征进行处理并不能准确的检测出阴影。因此,本文提出了一种混合颜色信息、光学不变性以及纹理特征的目标阴影检测方法,通过综合分析三种信息检测的结果,从而实现对阴影的有效确定。该算法能够有效地结合各种方法的优势,在实验中取得了较好的效果和运行效率。
[Abstract]:The research of intelligent video surveillance technology is a new direction in the field of computer vision in recent years. Its main research goal is to describe, analyze and understand the content of surveillance video through computer vision technology, image and video processing technology and artificial intelligence technology, and control the monitoring system according to the result of analysis and processing. So that the video surveillance system can meet the requirements of people for the level of intelligence. Its main research contents include: detection, tracking, recognition and behavior analysis of moving objects in surveillance video. The main research content of this paper is the method of moving target detection in intelligent video surveillance. A mixed Gao Si model is proposed to solve the problems which are often affected by the changes of illumination, complex background, shadow and so on in the traditional methods of moving target detection. A moving target detection algorithm based on edge detection and continuous frame difference. The algorithm uses the hybrid Gao Si model to model and update the background in the time domain, and uses the edge detection algorithm, the continuous inter-frame difference method and the mixed Gao Si model to obtain the initial moving target contour in the spatial domain, which is composed of the edge detection algorithm, the continuous inter-frame difference method and the mixed Gao Si model. And after the subsequent processing, we can get the perfect moving object. This algorithm can not only adapt to the background interference and the gradual illumination condition in the scene, but also overcome the problems of inaccurate target detection and incomplete edge detection in traditional algorithms. The experimental results show that the algorithm is relatively moderate in complexity, real-time and robust, and has high accuracy for moving object detection. Moving target detection is an important part of intelligent video surveillance, and shadow detection of moving object is an important step in moving object detection. Whether the target shadow detection is correct or not will directly affect the target object detection results. Through the study of various shadow detection methods, we find that only one feature processing can not accurately detect shadow. Therefore, in this paper, a method of shadow detection based on mixed color information, optical invariance and texture features is proposed. By synthetically analyzing the results of three kinds of information detection, the shadow can be effectively determined. The algorithm can effectively combine the advantages of various methods and achieve good results and operational efficiency in the experiment.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TN948.6

【参考文献】