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

项目名称： 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

项目编号： No.U1261105

项目类型： 联合基金项目

立项/批准年度： 2013

项目学科： 冶金与矿业学科

项目作者： 程德强

作者单位： 中国矿业大学

项目金额： 55万元

中文摘要： 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。

中文关键词： 矿井；电子稳像；图像处理；自适应滤波；运动矢量估计

英文摘要： 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 i

英文关键词： coal mine；electronic image stabilization；image processing；self-adaptive filtering；motion vector estimation