基于目标检测与跟踪算法的煤矿井下视频监控系统研究

发布时间：2018-05-07 16:02

本文选题：目标检测 + 目标跟踪　；参考：《辽宁工程技术大学》2014年硕士论文

【摘要】：我国是产煤大国,随着开采规模的不断扩大,受到井下特殊环境的影响,违规、误操作、冒顶等造成的事故越来越多。为保证安全,目前煤矿企业都配备了视频监控系统。但由于部分工人对井下安全意识不够强,以及设备重地、瓦斯等有害气体含量超标地区、巷道等危险地区的警示效果不明显,因此需要进一步通过视频监控系统加大监管力度。由于井下环境比较特殊,本文研究了针对性的检测与跟踪算法,能够从危险区域提取目标,方便之后的处理,并给出了视频监控系统总体设计思路与软件设计流程。主要工作如下：针对井下视频照度低的状况,提出了视频增强算法：通过对采集到的视频图像进行自适应中值滤波,以达到去除噪声的效果；然后对处理好的图像,对各个分量进行等比例变换,以达到对视频增强的目的。由于井下光照不均匀,有很多逆光情况,而且目标人员与背景相似度过高,导致了目标检测的困难,因此,本文提出了基于模糊理论的井下人员目标检测方法,通过定义目标区域的隶属函数,来实现对人员目标的检测,没有复杂的计算,实时性较高。人员目标跟踪方面,本文采用了基于Kalman滤波与Mean-Shift跟踪算法,避免了单纯的均值偏移方法的目标大面积遮挡问题与运动过快导致跟踪目标失败的问题,并可以减小搜索时间。实验证明,此算法可以有效的实现对目标的跟踪。系统总体架构方面,是由服务器与客户端还有嵌入式设备构成,服务器负责保存视频数据,客户端可以实时的从服务器端传输视频数据来观看,而嵌入式设备则负责将采集到的视频数据实时的传输到服务器端。软件方面采用Linux操作系统,三个部分的软件都采用C++来实现程序,编解码采用当前主流的H.264标准,界面支持则采用奇趣的Qt图像界面库,既方便了开发与维护,同时也节约了成本。
[Abstract]:China is a big coal-producing country. With the continuous expansion of mining scale, the accidents caused by special underground environment, illegal operation, misoperation and roof fall are more and more. In order to ensure safety, coal mining enterprises are equipped with video surveillance system. But because some workers are not strong enough to underground safety consciousness, and the equipment is heavy, the gas and so on harmful gas content exceeds the standard area, the roadway and so on dangerous area warning effect is not obvious, therefore needs to strengthen the supervision through the video surveillance system. Due to the special underground environment, this paper studies the targeted detection and tracking algorithm, which can extract the target from the dangerous area, which is convenient for processing. The overall design idea and software design flow of the video surveillance system are also given. The main work is as follows: aiming at the low illumination of downhole video, a video enhancement algorithm is proposed: through the adaptive median filter to the collected video image to achieve the effect of noise removal, and then to the processed image, In order to achieve the purpose of video enhancement, each component is transformed in equal proportion. Due to the uneven lighting in the underground, there are a lot of backlighting, and the similarity between the target and the background is too high, which leads to the difficulty of target detection. Therefore, a method of target detection based on fuzzy theory is proposed in this paper. By defining the membership function of the target area, we can detect the target without complicated calculation and high real-time performance. In the aspect of human target tracking, this paper uses Kalman filter and Mean-Shift tracking algorithm to avoid the problem of large area occlusion of the target and the problem that the fast motion leads to the target tracking failure, and the search time can be reduced. Experiments show that the algorithm can effectively achieve target tracking. The overall architecture of the system is composed of server, client and embedded device. The server is responsible for storing video data. The client can transmit video data from the server to watch in real time. The embedded device is responsible for the real-time transmission of the collected video data to the server. In software aspect, the Linux operating system is used, the three parts of the software are implemented by C, the coding and decoding adopt the current mainstream H.264 standard, and the interface support adopts the QT image interface library, which is convenient for development and maintenance. At the same time, it also saves the cost.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN948.6

【相似文献】