视频识别技术在变电站中的应用研究

发布时间：2018-04-28 14:30

本文选题：变电站 + 视频识别　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：随着视频监控的编解码技术和智能分析技术的发展,视频监控系统作为电网运行、检修的辅助监控手段,其业务正在向智能化、高清化和网络化方向发展。变电站内数量庞大的数字仪表,单纯凭借作业人员的主观判断和巡查方式进行仪表读数,不仅精确度不能保证,还极大的浪费了人力物力。另外变电站的安全问题也需要重点关注,如果有人员误入或为了人为破坏目的进入重要区域,为了避免造成经济损失和消除安全隐患,需要及时发现并制止。针对上述问题论文工作对智能视频监控系统中,数字仪表识别和重要区域人员入侵检测进行深入研究。首先,分析并研究了国内外智能视频监控、数字仪表识别和运动目标检测的研究现状,以及目前视频识别的主要技术。其次,通过研究目前成熟的数字仪表识别方法,基于穿线识别法和交点特征提取,提出了一种改进的数字识别方法,该方法在字符识别之前对分割得到的数字区域,从区域外接矩形的纵坐标、面积和高度差三个方面设置若干限制条件,有效防止将噪声区域误判为数字区域,然后采用直线和字符的交点数来构造数字0到9的识别特征,在确保准确率和抗干扰性的前提下,显著降低了计算量。最后,论文工作研究了目前主流的运动目标检测方法,例如光流法、背景减除法和帧间差分法等,在此基础上提出了一种改进的二帧差分法和背景减除法相结合的运动目标检测方法,该方法在二帧差分法的基础上,考虑光照变化的影响,通过把背景减除法和二帧差分法的差值图做与运算避免了图像序列差分法引起的“双影”现象,又可以避免背景减除法将光线突变误判为运动目标的情况,提高了识别准确率,并能够满足智能监控系统的实时性要求。
[Abstract]:With the development of coding and decoding technology and intelligent analysis technology of video surveillance, video surveillance system is developing to intelligent, high-definition and network as the auxiliary monitoring means of power grid operation and maintenance. A large number of digital instruments in the substation simply rely on the subjective judgment and inspection of the operator to read the instrument. It is not only the accuracy can not be guaranteed but also a great waste of manpower and material resources. In addition, the safety problems of substation also need to be paid attention to. In order to avoid the economic loss and eliminate the hidden danger of safety, it is necessary to detect and stop in time if there are personnel entering the important area by mistake or for the purpose of human destruction. In order to solve the above problems, the paper deeply studies the digital instrument recognition and the intrusion detection in important area in the intelligent video surveillance system. Firstly, the research status of intelligent video surveillance, digital instrument recognition and moving target detection at home and abroad is analyzed and studied, as well as the main technology of video recognition. Secondly, by studying the current mature digital instrument recognition methods, a modified digital recognition method is proposed based on the piercing recognition method and the intersection feature extraction method, which is used to segment the digital regions before character recognition. In order to effectively prevent the noise region from being misjudged as a digital region, the recognition features of the numbers 0 to 9 are constructed by using the number of points at the intersection of the straight line and the character to set up some limiting conditions from the three aspects of the vertical coordinate, area and height difference of the rectangle connected to the region, so as to effectively prevent the noise region from being misjudged into the digital region. On the premise of ensuring accuracy and anti-interference, the computational complexity is reduced significantly. Finally, the main methods of moving target detection, such as optical flow method, background subtraction method and inter-frame difference method, are studied. On this basis, an improved two-frame differential method combined with background subtraction method is proposed for moving target detection. Based on the two-frame difference method, the influence of illumination variation is considered. By doing and calculating the difference graph of background subtraction method and two-frame difference method, the "double shadow" phenomenon caused by the image sequence difference method is avoided, and the case that the background subtraction method misjudges the sudden change of light into moving target can be avoided, and the recognition accuracy is improved. And can meet the real-time requirements of intelligent monitoring system.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TM63

【参考文献】