东昆仑五龙沟金矿集中区遥感地质 信息提取与找矿预测

东昆仑五龙沟地区是青海省最具资源潜力的金矿集中区之一,遥感地质信息的识别与提取时找矿预测具有重要的指导意义.作者利用ETM+遥感影像,通过彩色变换、波段比值、主成分分析、二次主成分分析及其组合运算等图像处理技术,提取了东昆仑五龙沟地区与金矿化有关的地质信息.改进了蚀变信息的提取方法.所提取的线性构造反映了该区的构造格架,提取的蚀变信息与已知矿化点基本吻合,这些成矿信息的复合部位为金矿成矿的有利环境.最后结合区域成矿地质特征进行了成矿远景区的划分.     

包裹体岩片卸洗新方法—索氏抽提法

有机溶剂在恒温常压下封闭系统内,以浸泡淋洗的原理,建立了包裹体岩片卸洗新方法——索氏抽提法,克服了以往卸洗方法手工操作不便、难洗干净、易破碎、不能大批量处理等缺点。方法具有卸洗效率高、省心省力、操作方便、环保等优点,大大地提高了包裹体岩片卸洗的质量和效率。     

深源花岗岩中的碳氢化合物

利用石油地球化学的溶剂抽提和气相色谱技术研究了北京延庆碓臼峪花岗岩（深源花岗岩）样品中的可溶有机质，同时用高温处理过的石英砂进行空白对比实验。结果表明，所研究的两个花岗岩样品中均存在少量的高碳数碳氢化合物，且具有相同的成分特征。这为自然界存在深源的长链碳氢化合物提供了新的证据，也为解释花岗岩浆演化晚期水热流体中短链羧酸的成因提供了线索。     

基于Mask R-CNN的遥感影像土地分割与轮廓提取

传统的模式识别难以对土地遥感影像一次性精确统计分析.在精确分割出土地种类的前提下,本文提出了一种基于骨干网络为ResNet-101-RPN的Mask R-CNN的遥感影像土地分割与轮廓提取方法.该方法包括以下步骤：数据获取、图像去雾、遥感影像土地统计分析、土地分割和轮廓获取.在一个具有挑战性的卫星地图瓦片数据集上对所提出的方法进行训练和测试.实验结果表明,该方法以0.907的均值平均精度（mAP）和31.33像素的均值平均距离误差（mADE）获得了令人满意的不同种类土地分割和轮廓提取结果.     

暴雨过程的卫星云图纹理特征研究

针对卫星云图上云团所具有的半流体特性，应用纹理特征分析法对其进行定量研究。首先用灰度共生矩阵统计法提取100个样本云团的纹理特征参数，然后应用Mahalanobis距离作相似性测度进行聚类分析，提取出4类云以及无云区的具有旋转不变性的纹理特征参数。在此基础上分析了2003年7月4—5日江淮流域发生特大暴雨地区云团的纹理特征参数的变化规律与降水过程之间的对应关系。从实际的降水资料和纹理特征参数变化的回归分析结果来看，纹理特征参数的变化规律与降水量之间有较好的对应关系。这些结果不仅对中短期天气过程的诊断分析而且对数值预报都具有一定的实用意义。     

Extraction and analysis of gully head of Loess Plateau in China based on digital elevation model

In China′s Loess Plateau area, gully head is the most active zone of a drainage system in gully areas. The differentiation of loess gully head follows geospatial patterns and reflects the process of the loess landform development and evolution of its drainage system to some extent. In this study, the geomorphic meaning, basic characteristics, morphological structure and the basic types of loess gully heads were systematically analysed. Then, the loess gully head′s conceptual model was established, and an extraction method based on Digital Elevation Model(DEM) for loess gully head features and elements was proposed. Through analysing the achieved statistics of loess gully head features, loess gully heads have apparently similar and different characteristics depending on the different loess landforms where they are found. The loess head characteristics reflect their growth period and evolution tendency to a certain degree, and they indirectly represent evolutionary mechanisms. In addition, the loess gully developmental stages and the evolutionary processes can be deduced by using loess gully head characteristics. This study is of great significance for development and improvement of the theoretical system for describing loess gully landforms.     

基于地磁水平分量序列的磁暴自动识别

应用小波包分析技术提取了地磁水平分量序列的能量特征,用Fisher方法进行识别,研究了磁暴与非磁暴的识别率在不同时间长度t下的变化规律.当t=1h时,对磁暴与非磁暴的平均识别率达到83.8%;在确定为磁暴的前提下,进一步将磁暴按强度分为强磁暴(磁情指数K≥8)和非强磁暴(K≤7),平均识别率为72.9%.在对地磁水平分量进行连续实时监测的情况下,用这种方法可实现对磁暴的自动识别.     

过去2000a中国东部冬半年温度变幅与周期

The amplitude and rhythm of temperature changes at inter-decadal and inter-centennial timescale were studied, based on the winter-half year temperature change series reconstructed from historical phenological events in eastern China for the past 2000 years, together with the temperature change simulation from ECHO-G model for the past 1000 years, and the quasi-periods of temperature fluctuation were discussed by using wavelet analysis. The results indicate: 1) the maximal amplitude of winter half-year temperature change in eastern China at decadal and centennial scale, was above 2℃ and 0.5-1.0℃ respectively. The reconstructed result indicates that the amplitude of warming during the 20th century was identical with the maximum amplitude before the 20th century in eastern China, but the simulated result suggests that the amplitude of warming has exceeded the maximum amplitude. 2) The rhythms of temperature change at centennial to millennial scale in eastern China were about 100-year, 250-year, 400-year, 600-year and 1000-year. The 20th century, the 1st-3rd century and the 9th-13th century were warm peaks at inter-centennial scale as well as at millennial scale. It is implicated that the warming during the 20th century should be attributed to not only anthropogenic effect, but also natural climate variation.     

一种基于数值逼近的KPCA改进算法

核方法广泛应用于模式识别等领域,但其存在着特征抽取效率和样本集的大小成反比的瓶颈问题.因此提出一种基于数值逼近的方法确定虚拟样本矢量,以此代替训练样本,提高KPCA(Kernel Principle Component Analysis)特征抽取效率.在确定虚拟样本矢量时,只需将样本矢量的初值设定为随机变量,算法实现简单、高效.在基准数据集上的实验结果显示该算法优于同类算法.     

新疆阿祖拜矿田伟晶岩型矿床遥感找矿综合信息研究

对利用ASTER、RapidEye、QuickBird及ASAR等多种类型遥感影像在伟晶岩型矿床遥感找矿中的综合应用进行了探讨。云英岩化和钠长石化是阿祖拜矿田两类主要的蚀变类型。在分析其光谱特征的基础上,确定了利用ASTER数据对这两类蚀变进行信息提取的方法。利用ASTER热红外数据进行了岩浆岩类型的划分,为不同岩性接触带的准确定位提供了依据。伟晶岩在可见光-近红外光谱段的高反射率是伟晶岩的自动识别及遥感解译重要的物理基础。以QuickBird的伟晶岩解译结果为检验标准,得到RapidEye的伟晶岩自动提取精度约86%。构造交汇部位、岩性接触带以及伟晶岩富集区与蚀变异常的叠加部分是该区成矿预测的重要依据。基于蚀变异常、伟晶岩富集区、地质构造等遥感找矿综合信息,圈定了4个远景成矿靶区,为进一步开展地质详查工作提供参考。关键词:伟晶岩;蚀变异常;遥感;信息提取;成矿预测