煤矿井下视频监控异常行为识别算法的研究

发布时间：2018-06-07 02:02

本文选题：智能监控技术 + 煤矿灾害　；参考：《太原科技大学》2013年硕士论文

【摘要】：近年来，智能视频监控以及识别是目前计算机视觉领域中一个新兴的热点的前沿课题和研究方向。随着网络技术的发展，监控技术正沿着无人化、智能化、数字化方向不断前进，由此计算机领域学者专家适时提出新一代视频监控的概念。对于煤矿井下来说，由于事故频发，容易造成大量的人员伤亡，，采用智能监控系统能够在事故发生前能够及时地发现问题，及时地预警就能够减少伤亡，利用视频图像资料进行分析的结果直接关系到预警效果的好坏，以及是否能够准确预警。本文通过新一代智能监控技术应用于煤矿井下异常情况的识别，达到提前预警的目的。对于井下的视频图像实际情况，采用不同的处理方法，由于井下光线不足，就采取视频图像增强的处理，由于井下的特殊条件，采取最大内间方差法对阈值的选取，然后对图像进行分层处理，达到预期的效果结合前期的处理，以及获得的煤矿井下的视频资料，对片帮以及瓦斯突出等异常情况的特征有了了解，由于视频资料不能包括所有情况，对于井下的异常情况进行模拟实验，结合三帧帧差法对模拟的视频图像进行识别，无论是片帮还是瓦斯突出等异常出现都能够达到较高的识别率，达到了预期的实验目的。在分析其图像处理技术及架构后，针对实际监控条件与要求，模拟完成了井下检测和跟识别算法的设计，采用MATALB开发平台进行了图像提取和预处理、运动目标分割和识别。实验结果表明，该算法能够处理一般静态背景条件下的井下灾害识别。
[Abstract]:In recent years, intelligent video surveillance and recognition is an emerging hot topic and research direction in the field of computer vision. With the development of network technology, the monitoring technology is advancing along the direction of inhumanity, intelligence and digitization. Therefore, computer experts put forward the concept of a new generation of video surveillance in good time. For the underground coal mine, because of the frequent accidents, it is easy to cause a large number of casualties. The intelligent monitoring system can detect the problems in time before the accident, and timely early warning can reduce the casualties. The result of video image analysis is directly related to the effectiveness of early warning and whether it can be accurately forewarned. In this paper, a new generation of intelligent monitoring technology is applied to the recognition of underground anomalies in coal mines to achieve the purpose of early warning. For the actual situation of underground video image, different processing methods are adopted. Due to the shortage of underground light, video image enhancement is adopted. Due to the special condition of underground, the maximum internal variance method is adopted to select the threshold value. Then the image is stratified to achieve the desired effect. Combined with the previous processing and the video data obtained from the underground coal mine, the characteristics of the abnormal situation such as the slab and gas outburst are understood. Because the video data can not include all the cases, the simulation experiment is carried out for the abnormal situation in the underground. Combined with the method of three-frame difference to recognize the simulated video image, both the outliers and gas outbursts can achieve a high recognition rate and achieve the expected experimental purpose. After analyzing its image processing technology and structure, aiming at the actual monitoring conditions and requirements, the algorithm of downhole detection and identification is designed by simulation. The image extraction and preprocessing, moving object segmentation and recognition are carried out using MATALB development platform. The experimental results show that the algorithm can deal with downhole disaster identification under general static background conditions.
【学位授予单位】：太原科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TD76;TP391.41;TP277

【参考文献】