【作者】 吴昊；

【导师】 邓喀中； 尹鹏程； 范洪冬；

【作者基本信息】 中国矿业大学 ， 测绘工程（专业学位）， 2023， 硕士

【摘要】 我国矿产资源丰富,各类矿产的储量和产量都位于世界前列,且矿业体系健全、品种齐全。因此,每年需要排放和存储大量尾矿,据统计,我国矿山产业每年排放的尾矿高达3亿t,累计尾矿存储量接近200亿t。尾矿大部分以尾矿库的形式储存。尾矿库是矿山三大工程其中之一,同时也是一个具有高势能的人造泥石流危险源。一旦发生溃坝事故将会严重破坏环境和威胁下游人民群众的生命财产安全。因此,对尾矿库本身及其周围地表进行形变监测尤为重要。目前,现有研究多利用水准仪、全站仪、GNSS(Global Navigation Satellite System)和传感器等方式获取尾矿库上部分点位形变,很难实现尾矿库的高效率、高精度和精细化的形变监测。而InSAR(Interferometric Synthetic Aperture Radar)技术具有全天时、全天候、穿透性强和大范围监测的能力,且时序InSAR技术得到了长足的发展。但是传统InSAR技术在非城镇存在监测点密度不足,监测效果较差的问题,而DS-InSAR技术通过同质点识别和相位优化提高了这些地区的相干性,增加了监测点密度,改善了监测效果。由于不同地物对雷达电磁波的反射不同,单极化只能反应部分地物的散射信息,而将不同极化信息进行融合有助于获取更多的地表信息。针对上述情况,本文对DS-InSAR(Distributed Scatterers InSAR)方法展开了深入研究,对比不同的同质点选取算法、和相位优化算法,同时以巴西Brumadinho尾矿库和云南卡房尾矿库为实验区域进行形变监测研究,并联合当地降水数据揭示了该区域动态形变规律。主要研究工作和成果如下:(1)分析了尾矿库进行大范围、高精度形变监测的必要性,介绍了InSAR技术的发展和国内外的研究现状。阐述了D-InSAR(Differential InSAR)技术的基本原理、方法和劣势,针对上述不足之处详细介绍了时序InSAR技术的优势,并指出将极化SAR数据用于形变监测的可行性、可靠性和优势,为后续的理论研究和应用奠定基础。(2)研究了一种基于FaSHPS(FastSHPSelection)算法和特征值分解方法的DS-InSAR方法。对比分析了不同的单极化同质点识别方法,实验结果表明,FaSHPS算法能够有效克服传统同质点识别方法效率不高的问题,减少了第II类误差,证明了FaSHPS算法在大范围同质点识别的高效性。在此基础上,根据同质点阈值选取单极化分布式目标,利用特征值分解方法进行相位优化,并对比分析了优化前后的干涉图,验证了特征值分解的有效性和可行性。(3)研究了融入极化信息的多极化DS-InSAR方法。利用蒙特卡罗模拟实验对比分析了HTCI(Hypothesis Test of Confidence Interval)相比其它同质点识别方法的优势。并基于HTCI算法研究了极化同质点识别方法。根据同质点阈值和振幅离差指数对像元进行分类,提高了永久散射体和极化分布式目标的分类精度。对PS(Persistent Scatterer)根据振幅离差指标进行优化;为抑制极化分布式目标噪声的影响,先对其进行MMSE(Minimum Mean Square Error)滤波,然后根据相干性指标进行优化,改善相位质量。(4)选取了巴西Brumadinho尾矿库和云南卡房尾矿库为研究区域,利用DS-InSAR和多极化DS-InSAR方法对不同时间段的两个尾矿库进行了形变监测。结果表明:本文的方法与其他方法之间具有较高的一致性,且在各监测时间段内,两种方法能够有效的增加尾矿库的监测点密度,获得了更为精细的尾矿库形变信息,并分析了尾矿库随时间的变化特征。将形变信息和当地降水数据相关联,表明尾矿库的形变受到当地降水的影响,且巴西Brumadinho尾矿库在溃坝前其形变有明显的加速趋势,而卡房尾矿库在研究时段内的形变量较小,形变较为稳定。本文研究方法可为其它尾矿库进行高效率、高精度和精细化的形变监测提供参考。该论文有图47幅,表8个,参考文献159篇。更多还原

【Abstract】 China is rich in mineral resources,with reserves and production of various minerals ranking among the top in the world,and a sound mining system with complete varieties.Therefore,it is necessary to discharge and store a large amount of tailings every year.According to statistics,China’s mining industry discharges up to 300 million tons of tailings annually,with a cumulative tailings storage capacity of nearly 20 billion tons.Most of the tailings are stored in the form of tailings ponds.Tailings pond is one of the three major projects of the mine,and it is also a hazard of man-made debris flow with high potential energy.Once a dam breach accident occurs,it will seriously damage the environment and threaten the safety of downstream people’s lives and property.Therefore,it is particularly important to monitor the deformation of the tailings pond itself and its surrounding surface.At present,the existing research mostly uses the level,total station,GNSS(Global Navigation Satellite System)and sensors to obtain the deformation of some points on the tailings pond,which is difficult to achieve efficient,high-precision and refined deformation monitoring of the tailings pond.And InSAR(Interferometric Synthetic Aperture Radar)technology has the ability to monitor all day,all weather,strong penetration,and large-scale,and temporal InSAR technology has made significant progress.However,the traditional InSAR technology has the problem of insufficient monitoring point density and poor monitoring effect in non-urban areas.The DS-InSAR(Distributed Scatterers InSAR)technology has improved the coherence of these areas,increased the density of monitoring points and improved the monitoring effect through homogeneous point identification and phase optimization.Due to the different reflections of radar electromagnetic waves by different ground objects,single polarization can only reflect the scattering information of some ground objects and fusing different polarization information will help to obtain more surface information.In response to the above situation,this article conducted in-depth research on the DSInSAR method.Comparing different homogeneous point selection algorithms and phase optimization algorithms.At the same time,deformation monitoring research was conducted on the Brumadinho tailings pond in Brazil and the Kafang tailings pond in Yunnan as experimental areas,and the dynamic deformation laws of the region were revealed in conjunction with local precipitation data.The main research work and achievements are as follows:(1)The necessity of large-scale and high-precision deformation monitoring of tailings ponds was analyzed,and the development of InSAR technology and research status at home and abroad were introduced.Elaborated on the basic principles,methods,and disadvantages of D-InSAR(Differential InSAR)technology.In response to the above shortcomings,the advantages of temporal InSAR technology were introduced in detail.Pointed out the feasibility,reliability,and advantages of using polarized SAR data for deformation monitoring,laying a foundation for subsequent theoretical research and application.(2)Research a DS-InSAR method based on the FaSHPS(FastSHPSelection)algorithm and eigenvalue decomposition method.A comparative analysis was conducted on different single polarization homogeneous point recognition methods.The experimental results showed that the FaSHPS algorithm can effectively overcome the problem of low efficiency in traditional homogeneous point recognition methods,reduce Class II errors,and demonstrate the efficiency of the FaSHPS algorithm in largescale homogeneous point recognition.On this basis,single polarization distributed targets were selected based on homogeneous point thresholds,and phase optimization was performed using eigenvalue decomposition method.The interferograms before and after optimization were compared and analyzed,verifying the effectiveness and feasibility of eigenvalue decomposition.(3)Research the multipolarized DS-InSAR method incorporating polarization information.The advantages of HTCI(Hypothesis Test of Confidence Interval)over other homogeneous point recognition methods were compared and analyzed using Monte Carlo simulation experiments.Based on the HTCI algorithm,a polarization homogeneous point recognition method was studied.The classification of pixels based on homogeneous point threshold and amplitude dispersion index improves the classification accuracy of persistent scatterers and polarized distributed targets.Optimize PS(Persistent Scatterer)based on amplitude dispersion index.In order to suppress the impact of polarized distributed target noise,MMSE(Minimum Mean Square Error)filtering is performed first,and then optimization is performed according to the coherence index to improve the phase quality.(4)The Brumadinho tailings pond in Brazil and the Kafang tailings pond in Yunnan were selected as the research areas,and deformation monitoring was conducted on two tailings ponds at different time periods using DS-InSAR and multipolar DSInSAR methods.The results show that the method proposed in this thesis has high consistency with other methods.Within each monitoring time period,the two methods can effectively increase the density of monitoring points in tailings ponds,obtain more detailed deformation information of tailings ponds,and analyze the variation characteristics of tailings ponds over time.The correlation between deformation information and local precipitation data indicates that the deformation of the tailings pond is influenced by local precipitation.the Brumadinho tailings pond in Brazil has a significant acceleration trend in deformation before the dam failure.However,the deformation of the Kafang tailings pond during the research period is relatively stable due to its small deformation variables.The research method in this article can provide reference for efficient,high-precision,and refined deformation monitoring of other tailings ponds.There are 47 pictures,8 tables and 159 references.更多还原