智能安防视频监控算法研究与软件开发（设计成果+说明书+任务书+开通报告）.docVIP

摘要 分析了视频图像预处理的基本方法，在单视角目标检测算法的基础上，研究了多视角视频目标检测算法。以OpenCV开放源码为基础，设计了基于图像板卡的智能安防视频监控软件，并以实验室室内环境为背景进行了实验验证，实现了室内环境下对移动目标、火焰目标和人体目标的安防检测。实验结果表明，该软件系统具有较高的准确性。 本文的主要研究工作如下： 首先对运动目标检测算法：背景减除、帧间差分和光流法进行比较和学习。介绍了本文采用了基于局部更新的背景建模方法，进行背景减除，并以背景减除算法为基础实现移动目标侦测、人体检测和火焰检测。 其次，阐述了人体检测特征分类器的构造和训练过程。对特定环境下的人体特征建立分类器并进行机器训练，该算法取得较高的人体识别准确率。 第三，简要介绍了火焰的静态特征和动态特征；对火焰的R分量特征建立色彩模型，并结合火焰帧间特性以及角点特征作为火焰检测算法。在火焰检测时，着重采用基于多视角的检测算法，对各通道的火焰检测结果进行加权融合，得出最终检测结果。 关键词 多视角，视频监控，目标检测，人体检测，火焰检测 Abstract Analysis the methods of the video image preprocessing and the multi-angle video object detection algorithm based on the single perspective of the target detection algorithm. Design an intelligent security video monitoring software based on image acquisition card and OpenCV source. Realized the algorithm on moving targets detection, flame detection and human detection. Fartherly, verified by experiments in the laboratory indoor environment as the background. Experimental results show that the software system has a high accuracy. The main research work are as follows: First, compare and learn the moving target detection algorithm: background subtraction, frame difference and optical flow. Using a background modeling method based on local update to subtract the background. Secondly, described the structure and training process of the human body detection feature classification. The classification and machine training was under the certain circumstances. The algorithm achieved higher human recognition accuracy. Third, a brief introduction of static and dynamic characteristics of the flame was given. Create a color model of R component characteristics of the flame. In the flame detection, focusing on the detection algorithm based on multi-perspective algorithm, and weighted fusion channel flame detection results come to the final detection result. Keywords multi-angle, target detection, human detection, fire detection  目 录 1 引言 1 1.1 课题研究背景及意义 1 1.2 国内外研究与发展现状 2 2 多视角智能安防视频监控算法研究 5 2.1 单视角目标