基于计算机视觉的无人值守井场安防视频监控系统研究

发布时间：2018-01-09 18:19

本文关键词：基于计算机视觉的无人值守井场安防视频监控系统研究　出处：《西南石油大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着油田信息化建设的大力开展,我国建立了大量的无人值守井场,然而多数无人值守井场地处偏远地区,传统的安防视频监控系统无法保障其生产安全以及财产安全。因此,加快无人值守井场安防视频监控系统的信息化建设和智能化改造显得尤为重要。本文在广泛调研视频监控系统和计算机视觉技术现有研究成果的基础上,结合无人值守井场的环境特点,给出了一种能自动监测无人值守井场入侵行为的安防视频监控系统方案,完成了安防视频监控系统的总体方案设计,其中包括摄像机选型、现场POE供电、现场视频信息管理与存储、远程视频传输方式、具有运动目标检测和行人检测功能的远程视频监控客户端软件,并搭建了相应的系统模拟实验平台。在运动目标检测方面,本文通过实验分析了帧差法和背景差法在摄像机固定情况下提取运动目标的特点,并结合两者的优点提出了改进背景差法,在全局光照变化的情况下通过对比实验表明,基于改进背景差法的运动目标检测效果好于基于混合高斯模型的背景差法。在行人检测方面,利用了主成分分析(Principal Component Analysis,PCA)方法对HOG行人描述特征进行降维后,并在OpenCv的支持下,以INRIA行人数据库和自行采集的井场实景作为样本集,通过支持向量机(Support VectorMachine,SVM)进行了行人检测分类器训练,实验表明该分类器具有较高的检测率。最后,利用EmguCv将改进背景差法和行人检测算法封装到了基于C#语言编写的PC式NVR客户端软件,同时该软件还具备视频信息管理、视频信息存储、视频信息分析、远程云台控制、远程参数配置等功能。通过对模拟实验平台进行的相关测试实验表明,本文研究的无人值守井场安防视频监控系统具有可行性,且对提高人工视频监控质量、防止人为蓄意入侵以及确保无人值守井场的生产安全和财产安全具有深远意义。
[Abstract]:With the construction of oilfield informationization vigorously, our country established a large number of unattended site, however most unattended well site in remote areas, security video monitoring system can not guarantee the traditional production safety and property safety. Therefore, to speed up the security video monitoring system for unmanned wellsite informationization construction and intelligent transformation is particularly important. In this paper, based on extensive research of video monitoring system and computer vision technology on existing research results, combined with the environmental features of unmanned wellsite, gives a kind of automatic monitoring of unattended site intrusion security video monitoring system, complete the overall design of video monitoring system, including camera selection, field POE power live video, information management and storage, remote video transmission, a moving target detection and pedestrian detection The function of remote video monitoring client software, and set up a corresponding system simulation platform. In the moving target detection, the characteristics of this experimental analysis by frame difference method and background extraction of moving objects in the camera under fixed difference method, which combines the advantages of improved background subtraction method, the global illumination change under the condition of the experiment shows that the moving target detection effect of improved background subtraction in mixed Gauss model based on the background subtraction. Based on pedestrian detection, using principal component analysis (Principal Component Analysis, PCA HOG) method to describe pedestrians feature dimensionality reduction, and under the support of OpenCv, the INRIA pedestrian database and self collected well as real samples, the support vector machine (Support VectorMachine, SVM) for pedestrian detection and classification for training, experiment The classifier has higher detection rate. Finally, the improved package background difference method and pedestrian detection algorithm to the C# language of the PC NVR client software based on EmguCv, at the same time, the software also has video information management, video storage, video analysis, remote PTZ control, remote parameter configuration and other functions through the relevant. To test the simulation platform show that the unmanned wellsite security video monitoring system in this paper is feasible, and to improve the quality of video surveillance, prevention is of far-reaching significance to ensure human invasion and deliberate unattended site production safety and property safety.

【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN948.6

【相似文献】