公共自行车视频监控中异常事件检测技术研究

发布时间：2018-03-14 06:04

本文选题：公共自行车　切入点：视频监控　出处：《江南大学》2017年硕士论文　论文类型：学位论文

【摘要】：公共自行车作为一种低碳环保的交通工具,有效解决了最后一公里难题,在城市公共交通系统中发挥着不可替代的作用。但是目前社会上正在运营的是一车一桩式的有桩公共自行车系统,不仅土地资源浪费严重,而且建设和维护的成本较高。因此,无桩公共自行车是未来发展的一个趋势。但是由于没有固定桩位的限制,自行车的安全得不到保障,在停车区域内容易发生一些异常事件,例如在未得到许可的情况下公共自行车被人移动、目标在停车区域内徘徊等,而传统的视频监控系统一般不能起到异常事件报警作用。为此,根据无锡某公共自行车公司的开发需求,为实现公共自行车停车区域内异常事件的检测,开展了公共自行车视频监控中异常事件检测技术研究,并设计了嵌入式应用验证平台对研究结果进行了测试和分析,为公司基于机器视觉的公共自行车系统开发提供了技术支撑。针对视频监控中公共自行车无故被人移动或被盗,以及目标在停车区域内徘徊这两个异常事件检测技术开展了研究,提出了这两种异常事件的检测算法。最后设计了嵌入式应用验证平台,并将这两种算法移植到嵌入式应用验证平台上进行了测试和分析。论文的主要工作如下:(1)为了检测出公共自行车无故被人移走的异常事件,针对户外环境复杂动态程度较高,背景中存在大量噪声的场景,提出了一种改进的视觉背景提取的运动目标检测算法。针对ViBe(visual background extractor,视觉背景提取)算法在初始化背景模型时容易产生鬼影以及对复杂动态场景适应能力差的问题,通过前景点计数的方法抑制鬼影的产生并根据场景的复杂动态程度自适应调整阈值提高算法的适应能力。实验表明,与ViBe算法相比改进算法目标检测的准确率提高了30%以上。(2)为了有效检测出监控区域内发生的异常徘徊事件,在本文提出的运动目标检测算法的基础上,提出了一种基于运动目标轨迹分析的徘徊检测算法。通过提取目标的运动轨迹,并计算轨迹离散点的运动方向熵和运动轨迹的主方向角,结合行人的运动距离来判定行为是否是徘徊行为。与其他算法相比,提出的算法无需建立样本库,简单、有效,能满足实际应用要求。(3)为了对研究结果进行验证,设计了嵌入式应用验证平台,主要包括硬件平台的设计、嵌入式Linux操作系统的搭建以及相关软件的设计。针对实际应用场景,基于该平台对上文的研究结果进行了验证,验证结果表明,在复杂动态场景下能够实现公共自行车视频监控中异常事件检测,满足实际应用的需求。
[Abstract]:As a low-carbon and environmentally friendly vehicle, public bicycle has effectively solved the last kilometer problem. It plays an irreplaceable role in the urban public transportation system. But at present, what is being operated in the society is a public bicycle system with piles, which is not only a serious waste of land resources, but also a high cost of construction and maintenance. Pile-free public bicycle is a trend of development in the future. However, due to the lack of fixed pile position, the safety of bicycle is not guaranteed, so it is easy to occur some abnormal events in the parking area. For example, the public bicycle is moved without permission, and the target hovers in the parking area, while the traditional video surveillance system does not usually serve as an alarm for abnormal events. According to the development demand of a public bicycle company in Wuxi, in order to detect abnormal events in public bicycle parking area, the research on detection technology of abnormal events in public bicycle video surveillance is carried out. The embedded application verification platform is designed to test and analyze the research results, which provides technical support for the development of public bicycle system based on machine vision. And two abnormal event detection techniques are studied, and the detection algorithms of these two abnormal events are proposed. Finally, an embedded application verification platform is designed. The two algorithms are transplanted to the embedded application verification platform for test and analysis. The main work of this paper is as follows: 1) in order to detect the abnormal events of the public bicycle being removed by people without reason, the outdoor environment is more complex and dynamic. Where there's a lot of noise in the background, An improved algorithm for moving target detection based on visual background extraction (ViBe(visual background extractor) is proposed. When initializing the background model, the algorithm is prone to produce ghost images and poor adaptability to complex dynamic scenes. The method of spot counting is used to suppress the ghost and adjust the threshold to improve the adaptive ability of the algorithm according to the complex dynamic degree of the scene. Compared with the ViBe algorithm, the accuracy of the improved algorithm is improved by more than 30%.) in order to detect the abnormal hovering events in the monitoring area effectively, the proposed algorithm is based on the moving target detection algorithm proposed in this paper. A hovering detection algorithm based on trajectory analysis of moving object is proposed. The entropy of motion direction and the main direction angle of trajectory are calculated by extracting the moving trajectory of the target, and calculating the entropy of the motion direction of the discrete point of the trajectory, and the main direction angle of the trajectory. Compared with other algorithms, the proposed algorithm does not need to establish a sample base, simple, effective, can meet the practical application requirements. The embedded application verification platform is designed, including the design of hardware platform, the construction of embedded Linux operating system and the design of related software. The verification results show that the detection of abnormal events in public bicycle video surveillance can be realized in the complex dynamic scene and meet the practical requirements.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN948.6

【相似文献】