大名鼎鼎的方帅的博士学位论文---目前,计算机智能视频监控在理论和应用上都面临着很多难题,国内外大批学者投身于该领域的研究和探索,并且取得了大量的成果.本文是在这些成果的基础上,对计算机智能视频监控系统的关键技术进行研究.主要贡献可概括如下:首先,对目标检测技术进行了研究,并提出了一种基于背景建模的运动目标检测算法.利用统计的方法建立了基于颜色和颜色梯度的背景模型,并实时地对背景模型进行更新,最后将这两种背景模型综合考虑对目标进行了有效的检测.接着,研究了复杂背景下多目标跟踪问题,提出了基于蒙特卡罗粒子滤波器的复杂背景下多目标跟踪算法.然后,提出了目标分类算法.用有向无环图的多类支持向量机对目标进行分类,该分类器使用少量的样本,就可以得到较好的分类精度.最后,针对视频监控中的行为理解进行了探讨性的研究.建立一个行为理解的系统,这个系统由场景模型、行为模型和行为自动识别三部分组成.

-At present, the computer intelligence video frequency monitoring is facing many difficult problems in the theory and the application, the domestic and foreign large quantities of scholars join in this domain research and the exploration, and has made the massive progresses. This article is in these achievement foundation, conducts the research to the computer intelligence video frequency supervisory system s key technologies. The main contribution may summarize as follows: First, has conducted the research to the object detection technology, and proposed one kind based on the background modelling movement object detection algorithm. Has established using the statistical method based on the color and the color gradient background model, and real-time carries on the renewal to the background model, finally has carried on these two kind of background model overall evaluation to the goal the effective examination. Then, has studied under the complex background the multi-objective track que