矿井重大危险源视觉预警中移动载体摄像的自适应稳像关键技术研究

Mobile visualing and remote early-alarming for major hazard sources can effectively improve the intelligent level of the safety production of coal mines. Yet, the techniques for image adaptive stabilization of mobile carriers are not solved during the course of visual early warning. Based on the characteristics of underground images and movement states of carriers, the techniques for self-adaptive image stabilizing of underground mobile cameras are researched. The Researches include the following contents, image adaptive enhancement methods for underground images with low illumination and uniform light are first developed. Then, global motion estimation methods based on fast projection filters are proposed as well as adaptive mean filter algorithms to stabilize the video sequences under translational jitter. Finally, for stabilizing the video with complex motion, global motion estimation algorithms based on extracted features of images are studies including a fast filtering algorithm. With these studies, we will propose an image enhancement algorithm with the using of fuzzy theory, and establish a real-time image stabilizing method to filter high frequency jitter from moving backgrounds of images. Also, an image stabilizing method for complex motion video jitters is proposed. The final purpose of this project is to establish the basic principles of self-adaptive image stabilization for visual early-warning of major hazard sources. This project is a novel challenge under interdisciplinary efforts such as mine safety and monitoring, image processing, pattern recognition and multimedia information processing, and meets the requirements of national mining industries. The corresponding researching achievements can not only enrich the theory of visual early-warning of major hazard sources, but also provide academic foundation for underground devices to improve the intelligent levels. So, the research of this project has not only theory significances but also practical utility values.

矿井重大危险源移动可视化和远程主动智能预警可有效提高矿井安全生产智能化水平，但是视觉预警中移动载体摄像的自适应稳像技术还未有效解决。本项目根据载体运动方式，研究矿井移动载体摄像的自适应稳像关键技术。内容包括：矿井低照度、不均匀光照下的图像自适应增强方法；面向平移抖动的分区快速投影全局运动估计和自适应均值滤波方法；面向复杂运动视频抖动的全局运动估计和快速滤波稳像方法。通过研究，拟构建基于模糊理论的井下视频增强方法；提出移动背景高频抖动下的快速实时稳像方法；提出复杂运动下的视频抖动快速稳像方法，最终构建矿井重大危险源视觉预警中移动载体摄像的自适应稳像理论。本项目是安全检测与监控、图像处理、模式识别和多媒体信息处理的有机交叉，紧密联系国家煤炭工业的规划需求和技术发展现状，预期研究成果不但可进一步丰富危险源视觉预警理论，也可为对提高矿井装备的智能化改造提供依据，因此，具有重要理论和实际应用价值。

矿井移动载体可视化和远程主动智能预警可有效克服当前矿井被动式移动机械和设备的检测监控问题，移动载体基于视觉实时分析周边运动物体，并识别预警，从而实现移动设备重大危险源的远程主动预警。但是，在矿井移动载体摄像中获得的视频信息，由于移动载体的行进造成的不确定抖动使得拍摄的视频产生不稳定现象，极大地影响了现场情景的呈现质量和远程预警效果。矿井移动载体摄像的自适应稳像和防抖动是矿井移动可视化内容识别和远程主动预警中急需解决的关键基础问题。问题的解决对对矿井检测监控自动化装备的智能化提升和安全生产具有重大意义。.本项目结合矿井重大危险源视觉预警的技术现状，主要研究内容包括：.1）为了能精确地检测出图像的运动矢量，提高图像序列的帧间运动估计和稳像精度，研究矿井低照度、不均匀光照下的图像快速自适应增强方法，建立了面向矿井巷道特殊照明环境的图像增强和预处理方法。.2）旨在提高移动设备主动可视化预警能力，研究了面向平移抖动的分区快速投影全局运动估计和自适应均值滤波方法。实验结果表明，本方法不仅能够克服局部运动对全局运动矢量提取的不利影响，在去除抖动的同时保留摄像系统可能的扫描分量，最终构建了移动背景高频抖动下的快速实时稳像方法；.3）针对面向复杂运动视频抖动的全局运动估计和快速滤波稳像，根据人眼对于复杂运动中图像特征十分敏感的特点，提出了基于图像特征的复杂运动视频全局运动估计和快速滤波稳像方法，构建了复杂运动下的视频抖动快速稳像方法。.通过项目研究，最终形成了面向矿井重大危险源视觉预警的移动载体自适应稳像理论。研究成果不但进一步丰富和深化了电子稳像和矿井可视化监控理论，也为矿井智能化自动化检测监控设备研制提供了依据。因此，具有重要理论意义和实际应用价值。