高速公路监控视频异常检测技术研究

发布时间：2018-06-08 05:07

本文选题：高速公路 + 偏色检测　；参考：《重庆大学》2015年硕士论文

【摘要】：视频监控系统是高速公路管理系统中的重要组成部分,监控画面清晰正常是监控效果的重要指标。然而,实际的监控视频往往存在信号缺失、偏色、模糊、摄像头干扰等异常,影响视频监控系统的效果,且不利于后续高速公路异常事件的检测。现有的视频异常检测方法受复杂环境影响、实时性差,难以满足复杂的高速公路监控场景和实时性检测要求。为此,利用数字图像处理技术,实现高速公路监控视频异常的自动检测具有重要的学术意义和应用价值。针对上述问题,本文重点研究了视频图像偏色和摄像头干扰的检测技术,以及对不符合高速公路监控录像质量标准的视频自动筛选和判别技术。针对高速公路视频监控系统中经常出现的视频信号缺失和监控画面冻结的信号故障问题,本文分别给出了基于图像像素灰度值的方差、帧差法来实现这两种信号故障问题的检测。重点针对图像偏色问题,提出了一种偏色因子的计算方法,并给出了基于RGB三维颜色空间直角坐标系的偏色检测方法。通过在RGB颜色空间下,向量化处理每个颜色分量并计算其偏色因子实现对视频监控系统中偏色异常事件的检测。实验结果表明,该检测方法能够有效的检测出偏色的监控视频。针对高速公路摄像头经常发生平移、偏转、遮挡等干扰问题,本文提出了基于动态Harris角点模板匹配的高速公路摄像头干扰检测方法。利用角点在描述图像位置形状等方面的优势,基于掩膜提取技术,采用三张构造的灰度图对特定检测区域内的角点进行提取,然后根据角点数量和位置信息提出动态角点模板匹配算法计算匹配因子,通过匹配因子实现对摄像头的干扰检测。与Evan Ribnick检测方法相比,所提检测算法不仅耗时少,且具有较强的适应性和抗干扰能力。针对高速公路监控录像中常有的偏暗、偏白、模糊等不合格视频,本文提出了基于角点邻域像素标准差的视频图像清晰度评价方法来实现对高速公路视频图像质量的异常检测。该方法首先通过计算平均能量强度来排除偏暗或偏白的视频,再对视频图像结构特征进行分析,然后基于提出的视频图像清晰度评价方法对视频图像的清晰度进行评价,排除模糊的监控视频。实测高速公路视频验证了该方法的一致性和稳定性。最后,利用上述视频异常检测算法完成了高速公路监控视频异常检测系统的设计和实现,并编写了测试软件,对算法模块性能进行了测试,并对高速公路隧道、关键路段、收费广场场景下的监控视频进行了检测。实验结果表明,该系统能够较为正确地检测出高速公路监控异常视频,并能满足检测的实时性要求。
[Abstract]:Video surveillance system is an important part of highway management system. However, the actual surveillance video often has abnormal signal, color deviation, blur, camera interference and so on, which affects the effect of video surveillance system, and is not conducive to the subsequent detection of highway abnormal events. The existing video anomaly detection methods are affected by complex environment and have poor real-time performance, so it is difficult to meet the requirements of complex freeway monitoring scene and real-time detection. Therefore, it is of great academic significance and application value to use digital image processing technology to realize automatic detection of highway surveillance video anomalies. Aiming at the above problems, this paper focuses on the detection technology of video image color deviation and camera interference, as well as the automatic video screening and discrimination technology which does not meet the quality standard of highway surveillance video. In order to solve the problems of missing video signal and frozen image signal in freeway video surveillance system, the variance based on pixel gray value of image is given in this paper. Frame difference method is used to detect the two kinds of signal faults. Aiming at the problem of image color deviation, this paper puts forward a method to calculate the color deviation factor, and gives a method of color deviation detection based on RGB 3D color space right angle coordinate system. In the RGB color space, each color component is processed by vectorization and its coloration factor is calculated to detect abnormal events in video surveillance system. Experimental results show that the detection method can effectively detect the color deviation of surveillance video. Aiming at the interference problems of highway camera, such as translation, deflection, occlusion and so on, this paper presents an interference detection method for freeway camera based on dynamic Harris corner template matching. Taking advantage of corner points in describing image position and shape, based on mask extraction technology, three gray scale images are used to extract corner points in a specific detection region. Then according to the number of corner points and location information, a dynamic corner template matching algorithm is proposed to calculate the matching factor, and the interference detection of camera is realized by matching factor. Compared with Evan Ribnick detection method, the proposed detection algorithm not only takes less time, but also has strong adaptability and anti-jamming ability. Aiming at the substandard video such as dark, white and fuzzy in highway surveillance video, this paper proposes a method of video image definition evaluation based on corner neighborhood pixel standard deviation to realize the abnormal detection of highway video image quality. Firstly, the average energy intensity is calculated to eliminate the dark or white video, then the structure features of the video image are analyzed, and then the definition of the video image is evaluated based on the proposed method. Remove fuzzy surveillance video. The consistency and stability of the method are verified by the actual highway video. Finally, the design and implementation of the video anomaly detection system for expressway surveillance is completed by using the above video anomaly detection algorithm, and the test software is written, the performance of the algorithm module is tested, and the highway tunnel and key sections are tested. The surveillance video of the toll square scene was detected. The experimental results show that the system can detect the abnormal video of expressway monitoring correctly and can meet the real-time requirements of the detection.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN948.6;U495

【相似文献】