【作者】 杨静；

【导师】 仝卫国；

【作者基本信息】 华北电力大学 ， 控制理论与控制工程， 2015， 硕士

【摘要】 随着网络技术的高速发展以及多媒体技术在生产生活中的广泛应用,随之出现大量的数字视频产品,由于视频数据的非结构化、数据冗余多以及数据量巨大等特点,使得视频检索技术成为网络技术中的主要研究热点。在电力行业中,随着直升机巡线成为电力线路巡检的主要方式,巡检视频的处理主要以人工方式为主,工作量大,效率低,漏检率高,本文提出了基于视频检索的航拍视频的处理方法有效的解决了这一问题。视频信息检索主要是将自己所需的视频图像片段从海量的视频数据中提取出来,它以多学科理论为基础,涉及模式识别、图像处理、计算机图形学、人工智能等技术。本文重点研究了压缩域视频中的镜头边界检测与关键帧提取技术,提出了基于傅里叶描绘子的视频检索算法,利用压缩视频的编码数据提取I帧的DCT系数,量化出I帧的特征差值曲线,利用傅里叶描绘子对特征差值曲线进行重构,完成镜头边界的检测;通过提取出每个镜头的关键帧形成新的视频,利用新视频进行特征匹配得到所需的视频图像。利用该方法对一段直升机巡线的实拍图像进行处理,提取分镜头的关键帧形成的新视频,数据量是原视频的十分之一,大大减小了巡检视频处理的工作量,在此基础上,本文提出了基于Haar-like特征的绝缘子特征提取算法,通过Adaboost算法对新视频进行特征匹配提取出所需的绝缘子视频。上述实验结果表明了本文提出算法的有效性和可行性,可以提高电力线路巡检视频处理的高效性和准确性。更多还原

【Abstract】 With the rapid development of network technology, as well as the promotion of the multimedia, continuously produce various video data, along with more and more established video database, however, the utilization of a lot of digital video is very low under the existing conditions, because the video data is unstructured, data redundancy and data amount. In the power industry, with the helicopter patrol has become the main way of power line inspection, inspection of video processing is mainly in manual, however, manual is heavy workload, low efficiency, high undetected, this paper presents a video-based video retrieval aerial approach to solve this problem effectively.Video information retrieval is from a large amount of video data to find the desired video clip, it is based on the multi-disciplinary theory, mainly related to pattern recognition, image processing, computer graphics, artificial intelligence. This paper focuses on shot segmentation and key frame extraction in the compressed domain video. Paper proposed a key frame extraction algorithm based on compressed domain, by this algorithm, coding data can be extracted and processed to form the characteristic curve directly, then using Fourier descriptors to redraw the curve for shot segmentation and key frame extraction. Research work is as follows:Using this method,processing on a shot image of helicopter patrol,key frame extracted from every storyboard to form a new video. The amount of data is one tenth of the original video, which greatly reduces the workload of inspection. On this basis, this paper puts forward extraction algorithm based on the characteristics of insulator through the Haar-like, ever use the Adaboost algorithm to extract required video of insulator from feature matching. The above results show the effectiveness and feasibility of the algorithm proposed in this paper, can improve the efficiency and accuracy of video processing for power line inspection.更多还原