基于GPU的全波形并行LM分解算法

波形分解是机载激光雷达全波形数据处理的重要基础工作,通过求解波形函数模型的参数,将波形数据利用具体的函数模型拟合出来,实现对全波形及其中各个子波形函数表达。LM(Levenberg-Marquardt)算法及其改进的算法是波形分解中对参数进行拟合求解的常用方法。针对LM算法在参数拟合计算的过程中存在大量迭代和矩阵运算,提出了基于线程块组和线程两级并行粒度的并行计算方案。将串行多次循环迭代求解参数改为单次并行计算取最佳值实现对参数的选择,将矩阵运算进行线程块的协同并行计算,实现了LM算法在通用计算图形处理器上的并行计算。实验证明,在规定阈值条件下,并行LM降低了算法的迭代次数,提高了波形分解LM算法的计算效率,为提高波形分解的处理效率提供了研究思路。     

Characteristics of vertical ozone distribution in the lower troposphere in the Yangtze River Delta at Lin’an in the spring of 2001

Deterioration of air quality as a result of rapid eco- nomic development in the Yangtze River Delta (YRD) has attracted much scientific and public attention[1―7]. To explore the air quality of a region, measurement of pollutants in locations reflecting the “background” atmospheric conditions of the region is essential. The Regional Baseline Air Quality Monitoring Station of the Chinese Academy of Meteorological Sciences (CAMS) at Lin’an is located in such an area in Zhe- jiang Pro…     

遥感图像震害特征组合增强模型及其算法实现

在高分辨率遥感图像中, 不同震害损毁程度的建筑物呈现不同的图像特征, 鉴于此本文提出一种利用遥感图像多特征分析建筑物损毁程度的检测方法. 以2015年尼泊尔M_S8.1地震为例, 结合震后高分一号卫星全色遥感图像和城市道路矢量数据提供的街区信息, 以建筑物街区为单元进行图像纹理提取和局部空间统计等多类别图像特征参数分析, 并构建多特征分类模型, 将震后建筑物街区划分为基本完好、 部分损毁和严重损毁等3个类别. 试验结果表明, 本文提取的参数能够有效地表征损毁建筑物的图像特征, 而且建筑物震害损毁检测精度较高. 该方法可用于建筑物震害损毁信息的快速提取, 为震后应急救援提供指导; 同时还可为我国自主研发高分卫星遥感数据在地震灾害信息提取中的应用提供技术参考与方法借鉴.     

高温磨粒条件下氟橡胶圈摩擦磨损机理研究

井下高温和磨粒是影响密封件密封失效的常见原因,了解高温条件下的弹性体在磨粒环境中的磨损机理对提高井下密封件寿命有着重要意义。对氟橡胶进行不同高温处理后,开展橡胶的摩擦磨损实验研究,分析在磨粒条件下橡胶的磨损情况和摩擦磨损机理。实验结果表明,随着橡胶加热温度的升高,摩擦系数曲线平衡所需的时间逐渐减少。未高温处理的橡胶表面的磨损形式以犁沟为主。100 ℃时橡胶的磨损表面受到高温的影响而发生改变,犁沟减少。200 ℃时氟橡胶圈表面以凹坑为主,并且磨损最严重。304不锈钢表面的磨损随着橡胶的加热温度的改变而变化,磨损面中心区域的磨损形式由犁沟向凹坑转变,而磨损面的内侧由凹坑向犁沟转变。     

基于数学形态学的测试卡图案识别方法

提出了一种基于数学形态学的双向匹配识别方法,通过对测试运动员的测试卡图案进行识别,试验结果表明该方法不仅能良好地去除测试卡图案上噪声的影响,而且能有效识别测试卡上图案,识别正确率高达95%以上。     

尼泊尔Ms 8.1级地震活动构造及次生地质灾害研究

2015年尼泊尔Mw 7.8大地震诱发了大量的山体滑坡,对尼泊尔境内与周边地区造成了严重的影响。选取离震中较近的辛杜帕尔乔克地区作为研究区,基于L波段ALOS-2和C波段Sentinel-1A两种合成孔径雷达数据,采用堆叠合成孔径雷达干涉测量（interferometric synthetic aperture radar,InSAR）技术开展震后滑坡的探测与识别,结合光学影像圈定出滑坡隐患点14处。在此基础上,联合升降轨数据和多维小基线集（multidimensional small baseline subset,MSBAS)-InSAR技术获取了典型滑坡的二维时间序列形变特征,结果表明,该典型滑坡的主要形变发生在水平东西向,最大形变速率为-69 mm/a。同时,通过对该典型滑坡时间序列中的趋势项与周期项形变信号进行分析,发现地震对于滑坡运动具有明显的加速作用,且降雨量的增加使得滑坡位移在每年的8月-11月呈现出周期性变化,可为震后滑坡监测研究提供参考。

     

The Thickness of Seismogenic Layer Under the Areas Surrounding the Ordos Block, Northern China

—?A seismic approach was used to calculate the thickness distribution of the seismogenic layer under the block Ordos in northern China. This block was chosen because of its stability, the complicated tectonics around its boundaries and the completeness of the data set. Several strong earthquakes occurred in this area in the 20^th century, with the largest one (Haiyuan, M= 8.4, 1920) in the southwest of the Ordos. Most of the large faults around the Ordos are strike-slip ones. The breaking point (i.e., the saturation of earthquake magnitudes M ^c ) of the self-similarity from small to large events based on the Gutenberg-Richter relation is calculated. Under the assumption of the L model which expresses that the offset is proportional to the length of the fault (Scholz, 1982, 1990, 1994), the thickness (W _c ) of the seismogenic layer is directly related to M ^c . To display smoothly the change of the thickness of the seismogenic layer and to ensure the reliability of the results, a moving cell of 150 events, with a moving step of 0.1 degree, was adopted. To take advantage of the existing data set, the “mean value method,” which is based on the latest complete data information to make the existing data set a `complete' one, was used to extrapolate it so that the data covers the whole time period. The results show that the average thickness of the seismogenic layer in the southwest, in the northwest and in the east of the Ordos is around 30, 9 and 17?km, respectively. The thickness generally decreases from the southwest to the northeast. This kind of spatial variation is in agreement with the maximum depth distribution of large amounts of microearthquakes, with the correlation coefficient to be about 0.88, and the Moho depth.     

Experimental study on bed-load sediment transport under irregular wave and current combined flow

Using an irregularly oscillating tray and flume,a series of experiments are completed to evaluate bed-load sediment transport rate under irregular wave-current coexistent fild.Testing conditions include three interaction angles 0°,45°,90° and two kinds of median sizes (0.38 and 1.10 mm).The results of transport rate show that the net sediment transport rate can be expressed approximately as the function of the maximum bottom shear stress of waves,mean shear stress of current and the grain size.     

Fisher算法及加权马尔可夫链模型的耕地需求量预测方法——以山东省聊城市为例

根据聊城市1949-2007年的耕地数据,以耕地变化率来描述耕地数量变化的速率,借助Fisher最优分割法将耕地变化率状态分为5个等级,利用规范化后的各阶自相关系数为权重,运用加权马尔可夫链模型对聊城市耕地需求量进行预测。结果表明,预测值与实际值的相对误差为0.01%,预测结果非常理想,同时也验证了加权后的马尔可夫预测法更具科学性和实用性。     

Estimation of forest LAI by inverting canopy reflectance models and multi-angle imagery

Computer simulation models have seldom been applied for estimating the structural and biophysical variables of forest canopy. In this study, an approach for the estimation of leaf area index (LAI) using the information contained in hyperspectral, multi-angle images and the inversion of a computer simulation model are explored. For this purpose, L-systems combined with forest growth model ZELIG were applied to render 3-D forest architectural scenarios. The Radiosity-graphics combined model (RGM) was used to estimate forest LAI from the Compact High-Resolution Imaging Spectrometer/Project for On-Board Autonomy (CHRIS/PROBA) data. LAI inversion was performed using the look-up table (LUT) method. The estimated LAI was evaluated against in situ LAI measurement and compared against the LAI predictions from CHRIS data obtained using the Li-Strahler geometric-optical canopy reflectance model (GOMS). The results indicated that the method used in this study can be efficient strategy to estimate LAI by RGM model inversion.