基于深度学习的倾斜摄影建筑物表面损毁信息提取

建筑物受损信息是地震受灾程度评估的基础,针对传统建筑物表面信息识别人工成本高、效率低等问题,受深度学习提取建筑物影像的启发,提出利用无人机倾斜摄影模型与深度学习相结合的方法提取震后建筑物表面破损信息。以2019年长宁6.0级地震为例,选用双河镇震后倾斜摄影模型切片图为数据源,对比分析面向对象分类方法、VGG-16模型和DeeplabV3+模型对建筑物表面损毁信息的提取结果。分析结果表明,针对建筑物表面破损信息的提取,尤其是细小裂缝的提取,语义分割网络DeeplabV3+模型具有较强的优势(准确率96.93%、召回率96.85%、总体精度96.89%),可实现建筑物表面破损信息的有效提取,具有较强的应用价值。     

略图调制系统的设计与实现

介绍了应用Visual C 6.0开发工具开发的略图调制系统,内容主要包括：开发背景、系统总体构成、主要功能和应用实例,并对一些技术要点作了原理性阐述,最后简单展望了进一步的工作。     

中国北方四剖面CaCO3含量变化及其反映的古降水量

中国北方自西向东的4个剖面,各地层CaCO3的含量呈现逐渐减少的规律。科尔沁剖面和延长剖面、西宁盘子山剖面CaCO3含量曲线表现出的特征为：在每一个古土壤单元的下部一定深度内CaCO3曲线都表现为一个明显的峰;新疆塔勒德剖面的各古土壤的CaCO3含量表现为表层比中、下部要高。利用CaCO3淀积深度恢复出来的部分区域的古降水量显示,末次间冰期期间盘子山剖面的古降水量大为352mm,延长剖面和科尔沁剖面的古降水量大约为500-550mm;末次冰期的间冰阶阶段科尔沁剖面的古降水量大致为320～360mm;延长剖面古降水量大约为443mm;全新世以来盘子山剖面的古降水量大约为306mm,延长剖面和科尔沁剖面的古降水量大约为320-380mm。这些数据可为半定量恢复地质历史时期古环境状况提供依据。     

济南泉域岩溶地下水三维等参有限元数值模拟

基于济南泉群研究区域岩溶水贮存和运动特征,建立了三维有限元数学模型,并利用等参有限元方法求解,再根据2001年、2002年的水文资料,识别模型参数,进一步对济南泉域内岩溶地下水的开采方案进行评价,探讨了济南市“保泉”与“供水”的可行性,为科学管理济南市泉群区域内地下水资源提供依据。      

陷落柱的地震识别技术及其应用

通过地震资料解释的研究与实践,对华北石炭二叠系中陷落柱的地震识别技术及其应用情况,进行了较系统的分析和总结。这对今后正确分辨和圈定陷落柱具有一定指导意义。地震剖面、水平时间切片和沿层振幅切片、相干体技术、波阻抗反演技术等,在对陷落柱的发育部位及其空间关系的解释方面,各有其独特效果。山西某矿陷落柱相当发育,利用陷落柱的综合地震识别技术进行解释,效果明显。      

小波分析与C3相干体在小断层解释中的应用

利用小波分析的分频特性及C3相干数据体比常规振幅数据体具有更高的构造分辨能力的特点,将小波分析与C3相干体两种方法相结合,用来识别煤矿采区小尺度构造,特别是识别小断层.实际应用表明,利用小波分析与C3相干体可更加精细地揭示断层等构造信息,提高小断层识别能力.     

形变单位在地震文献中的正确使用

针对地震观测技术论文中有关形变单位"毫秒",即10~(-3″)混用为时间单位ms的情况,笔者从单位符号的意义及观测技术仪器的工作原理出发,最终明确,地震观测仪器中倾斜仪观测量单位应为10~(-3″),不能以时间单位ms代替,二者不能混为一谈。     

基于MPI并行算法的电阻率法多种装置数据的三维联合反演

由于地表电阻率法受到浅地表局部异常体的干扰,反演精度受到影响,井中装置数据资料参与反演虽然可以减小浅地表局部异常体的干扰,但是由于钻井位置的局限性,数据量得不到保障,也会导致反演精度降低.为此,本文开发了一套结合地表、地-井、井-地和井-井装置数据的三维联合反演算法.首先,利用有限差分法实现正演模拟,采用非线性共轭梯度法（NLCG）恢复电阻率结构;其次,调用Message Passing Interface（MPI）函数库加速正演模拟和灵敏度矩阵运算,当开辟12个进程时,反演程序获得最大加速比4.51;最后,通过合成数据和实测数据算例证明该反演算法的有效性,也证实了多种装置组合数据体反演结果明显优于单一地表装置数据体反演结果.     

THE APPLICATION OF IMAGE-BASED MODELING IN PALEOEARTHQUAKE TRENCH STUDY

The geological structure exposed by paleoearthquake trenches is the key material to the right cognition of fault activity and paleoearthquake. However, paleoearthquake trenching inevitably destroys active tectonic geomorphic evidence and trench exposures are usually difficult to reserve. The conventional process of recording the delicate geological information, manually constructing photomosaics by image-editing software, is time-consuming and produces undesirable artificial distortions. Herein, we explored the process of constructing trench orthophotomosaics and the 3D image model using the Image-based Modeling technology and applied it to the Liutiaohe trench across the Tianqiaogou-Huangyangchuan Fault, Gansu Province. Based on the 3D image modeling and orthophotomosaic, we firstly constructed the control points and scale bars on cleaned trench walls and collected photos of all sections of the trench with a digital camera in the field, and then reconstructed the 3D model of the trench through the Agisoft PhotoScan, an efficient image-based modeling software, and finally yielded the 3D image model of the trench and othophotomasaics of the trench exposures. The results show that the automated workflow can produce seamless, sub-millimeter-level high-resolution photomosaics more quickly, with precision in the centimeter range, and the 3D image model is of great help to identify strata and geological structures in trenches with much lower capital and labor costs and low expertise levels compared with LiDAR, meanwhile, the 3D archive benefits the share and communication and even allows future reinterpreting the site using new insights.     

A NEW FINDING OF SURFACE RUPTURE ZONES ASSOCIATED WITH THE 1936 LINGSHAN M6(3/4) EARTHQUAKE,GUANGXI, CHINA

On April 1, 1936, an M6(3/4) earthquake occurred on the Fangcheng-lingshan Fault. This event is the biggest historical earthquake on the coastal seismic zone, South China ever. But so far, no any findings about the surface rupture of this event have been reported. This paper is the first to find several intact surface rupture zones associated with the 1936 Lingshan seismic event, in the areas of Gaotang, Jiaogengping etc. on the northeast segment of the Fangcheng-Lingshan Fault. According to the field work, the surface rupture stretches to 10km and distributes along NE direction in front of Luoyang Mountain, represented by earthquake scarp, extensional fracture, dextrally faulted gully and river system etc. The characteristics of surface ruptures and faulted landforms indicate that the surface rupture is of normal-dextral strike slip faulting. The trenching on this fault exposed that at least three seismic events have been recorded, including two historical earthquake events and the latest one is the 1936 Lingshan M6(3/4) earthquake. These surface rupture zones are the key to the detection of seismogenic structure and the re-estimate of magnitude of this event. The new finding of these surface rupture zones would be particularly significant for the detection of the seismogenic structure of Lingshan M6(3/4) earthquake.