基于优势演员-评论家算法的强化自动摘要模型

针对长文本自动摘要任务中抽取式模型摘要较为冗余,而生成式摘要模型时常有关键信息丢失、摘要不准确和生成内容重复等问题,提出一种面向长文本的基于优势演员-评论家算法的强化自动摘要模型（A2C-RLAS）。首先,用基于卷积神经网络（CNN）和循环神经网络（RNN）的混合神经网络的抽取器（extractor）来提取原文关键句;然后,用基于拷贝机制和注意力机制的重写器（rewriter）来精炼关键句;最后,使用强化学习的优势演员-评论家（A2C）算法训练整个网络,把重写摘要和参考摘要的语义相似性（BERTScore值）作为奖励（reward）来指导抽取过程,从而提高抽取器提取句子的质量。在CNN/Daily Mail数据集上的实验结果表明,与基于强化学习的抽取式摘要（Refresh）模型、基于循环神经网络的抽取式摘要序列模型（SummaRuNNer）和分布语义奖励（DSR）模型等模型相比,A2C-RLAS的最终摘要内容更加准确、语言更加流畅,冗余的内容有效减少,且A2C-RLAS的ROUGE和BERTScore指标均有提升。相较于Refresh模型和SummaRuNNer模型,A2C-RLAS模型的ROUGE-L值分别提高了6.3%和10.2%;相较于DSR模型,A2C-RLAS模型的F1值提高了30.5%。     

基于光谱归一化的变组分光谱混合分析(NMESMA)方法及其应用

混合像元问题在低、中分辨率遥感图像中尤为突出,混合像元的存在不仅会影响地物识别和图像分类精度,也是遥感科学向定量化发展的主要障碍之一。因此,遥感图像混合像元分解及其地表覆盖信息的定量提取是近年来研究的热点。针对城市土地覆盖信息的定量提取问题,利用中等分辨率遥感图像(Landsat TM),集成光谱归一化与变组分光谱混合分析(NMESMA)的方法,基于植被-非渗透表面-土壤(V\|I\|S)模型,定量提取研究区植被、土壤和非渗透表面3类土地覆盖的定量信息,并与固定组分的光谱混合分析(LSMA)分解结果进行对比分析。结果表明:基于光谱归一化的变组分光谱混合分析(NMESMA)方法获得的精度高于传统固定组分的光谱混合分析(LSMA)结果,可有效解决光谱异质性较高的城市区域的混合像元问题,为有效提取城市地表覆盖信息,研究城市生态环境变化和模拟分析,提供了有效的信息提取方法。     

利用夏玉米端元丰度估算夏玉米种植面积的研究

以河南省鹤壁市为研究区,以FY3/MERSI数据为主要遥感信息源,采用基于光谱匹配的自适应最佳端元组合混合像元分解方法提取夏玉米端元丰度值,构建像元中夏玉米端元丰度值和夏玉米种植面积百分比值之间多种方程形式的回归模型,综合模型建立时的相关系数、显著性水平和验证样点的均方根误差情况,选择立方模型对研究区夏玉米种植面积进行估算。经验证,遥感估算的夏玉米种植面积精度为97.1%,位置精度为82.5%。研究结果可为采用中低空间分辨率遥感数据在种植结构复杂地区准确估算大范围作物的种植面积提供科学基础。     

基于混合像元分解的薄云下光学遥感图像恢复方法

云遮挡是限制光学遥感卫星对地观测能力的主要因素之一。针对薄云下光学遥感图像的图像恢复问题,首先将云对光谱观测值的影响在线性混合像元模型中显性地加以表达,提出了针对云特性的改进型线性混合像元模型;其次给出了两种基于混合像元分解的图像恢复方法、直接消除法与丰度调整法;最后分别在两种混合像元分解算法与两种图像恢复方法,即VCA（顶点成分分析）算法/MDC-NMF(最小距离限制的非负矩阵分解)算法与直接消除法/丰度调整法的不同组合下,分别利用模拟数据和真实数据,对相关方法的图像恢复能力和图像恢复效果进行了定性和定量分析。实验结果表明,MDC-NMF算法与丰度调整法的组合处理能够获得最佳的图像恢复效果。     

基于Landsat 的城市热特征研究——以兰州市为例

运用Landsat5和Landsat7获得的遥感数据评估兰州市区的热特征,为了量化城市土地利用和覆盖密度,在对植被-非渗透面-土壤(V-I-S) 模型修正的基础上,运用线性光谱混合分析技术(LSMA),分解了不同种类的地表组成并对非渗透面密度进行了分类。结果表明,研究区有很高的热效应,这种热效应与城市开发密度高相关,并通过植被覆盖信息、非渗透面的空间分布及与其相关联的热特征,可以有效量化城市土地利用、开发密度和热格局。     

基于混合像元分解的高光谱影像柑橘识别方法

为及时准确地监测柑橘种植信息,以江西省会昌县作为研究区,采用EO-1 Hyperion高光谱影像作为数据源,构建了基于混合像元分解的高光谱影像柑橘识别方法。首先,针对EO-1 Hyperion高光谱影像提供了242个波段,光谱范围广的特点,在波段选择、大气校正等预处理的基础上,提取研究区典型地物端元光谱曲线;然后,利用全约束线性光谱混合模型进行混合像元分解,提取出柑橘端元的丰度值,并通过对照高分遥感影像,构建柑橘端元丰度与柑橘实际种植的对应的关系。结果表明:由于典型地物端元提取中不可避免的误差及柑橘冠层覆盖度的差异,柑橘种植的准确识别与其柑橘端元丰度阈值存在对应关系。在经过反复试验的条件下,研究区柑橘端元丰度阈值设定在0.30~0.45范围之内,总精度达到90%以上,能够满足柑橘种植识别要求。     

混合分层抽样与协同过滤的旅游景点推荐模型研究

采用问卷调查与自动抓取相结合的方式,采集用户信息、用户评分等旅游数据,对数据做分层抽样,生成包含用户旅游喜好信息的“智慧旅游”数据集。围绕该数据集,预处理用户评分并执行基于用户聚类的协同过滤算法,以计算目标用户与聚类中心的相似性。结合分层抽样模型生成的旅游喜好信息,输出混合推荐列表。实验结果表明：相比基线,混合分层抽样与协同过滤的推荐模型对评分预测的均方根误差（Root mean square error, RMSE）和平均绝对误差（Mean absolute error,MAE）分别降低11.5%~64.9%和18.8%~47.7%。混合推荐的准确率和召回率相比基线也有较大程度提升,旅游景点推荐效果良好。     

应用ANN/HMM混合模型预测蛋白质二级结构

针对3状态隐马尔可夫模型（hidden Markov model,HMM）预测蛋白质二级结构准确率不高的问题,提出15状态HMM,通过改进的算法与BP神经网络相结合进行二级结构预测。研究对象为CB513数据集中筛选出的492条蛋白质序列,将其随机均分7组。应用混合模型进行预测,对准确率进行7交叉验证,Q3准确率达7721%,SOV值为7252%。结果表明,混合模型既能充分考虑相邻氨基酸残基间的相互影响,也能在一定程度上照顾二级结构的远程相关性,因此带来了较好的预测准确率。     

内陆水体悬浮物波段比值优化模型及藻类丰度对模型精度影响研究

内陆水体中浮游植物的存在对悬浮物（TSM）遥感反演模型精度具有一定的影响,藻类丰度会导致水体遥感反射率降低。实验基于中国、澳大利亚和美国内陆水体的372个采样点（4个数据集）水质分析和光谱实测数据,构建内陆水体遥感反射率与TSM的相关关系,建立最优波段比模型（OBR）,并分析了藻类颗粒物存在对该模型精度的影响。由于水质的不均一性,不同区域的水质参数敏感波段存在差异,因此各数据集用于建模的最优波段比值不同。结果表明,OBR模型精度较高,误差较小,中国水体模型验证均具有较好效果（石头口门水库：R²=0.87,RMSE=14.1 mg/L;查干湖：R²=0.82,RMSE=23.6 mg/L）,澳大利亚水体模型验证效果最佳,R²值高达0.95（RMSE=4.2 mg/L）,美国水体模型精度较低（R²=0.78,RMSE=3.7 mg/L）。研究发现,模型精度受水体叶绿素（Chla）浓度和Chla/TSM比率影响,当水体以TSM浓度较高的非藻类颗粒物为主时（如中国石头口门水库和南澳洲地区水体数据集）,最优波段比值模型表现更好;而当水体以浮游植物为主时,水体中的浮游植物的丰度会使光谱信号复杂化,从而限制或降低TSM浓度遥感算法的精度（如美国印第安纳州中部水库数据集）。     

基于渗透算法和改进型OPTA的裂纹检测算法

针对低压电流互感器表面裂纹的提取与判定,提出基于渗透算法和改进型OPTA（One-Pass Thinning Algorithm）的互感器表面裂纹检测算法。首先获取互感器表面的灰度图像;其次根据裂纹区域像素值、亮度变化,通过设定种子像素点、亮度阈值,利用渗透算法渗透得到二值图;再次从裂纹连通性入手,利用改进型OPTA提取ROI（Region of Interest）的骨架,骨架由单像素点组成;最后利用裂纹具有分叉性的特点,像素点的邻域点个数超过2的即可判定为裂纹。实验表明,渗透算法能够有效地从图像中提取出ROI,并保持了ROI的线性特征,改进型OPTA使ROI完全细化为单像素图像,提出的邻域点判别法检测效率在97%以上,相较于所提其他检测方法有明显提高。     

多时相影像福州市不透水面对城市地表温度的空间分布研究

随着城市化进程的加快,城市热力场也随之发生变化,从而影响着城市区域环境、社会经济以及社会环境。由于NDVI具有季相变化的不稳定性,本研究采用两个时相TM/ETM+影像分析福州市及其周边地区不透水面对热力场的时空分布变化状况。为了获取精确的城市不透水面信息,本实验采用NDVI二元法结合2000年同区域的IKONOS影像提取不透水面信息。通过定量分析不透水面百分比、NDVI与地表温度的关系,得出不透水面百分比与城市地表温度呈线性相关,其相关系数在0.7左右;尤其30%以上的不透水面对地表热环境的空间分布影响最为突出,因此,相对于不稳定的NDVI而言,不透水面信息能更好地反映城市热环境的空间分布状况。     

Normalized spectral mixture analysis for monitoring urban composition using ETM+ imagery

With rapid urban growth in recent years, understanding urban biophysical composition and dynamics becomes an important research topic. Remote sensing technologies introduce a potentially scientific basis for examining urban composition and monitoring its changes over time. The vegetation-impervious surface-soil (V-I-S) model, in particular, provides a foundation for describing urban/suburban environments and a basis for further urban analyses including urban growth modeling, environmental impact analysis, and socioeconomic factor estimation. This paper develops a normalized spectral mixture analysis (NSMA) method to examine urban composition in Columbus Ohio using Landsat ETM+ data. In particular, a brightness normalization method is applied to reduce brightness variation. Through this normalization, brightness variability within each V-I-S component is reduced or eliminated, thus allowing a single endmember representing each component. Further, with the normalized image, three endmembers, vegetation, impervious surface, and soil, are chosen to model heterogeneous urban composition using a constrained spectral mixture analysis (SMA) model. The accuracy of impervious surface estimation is assessed and compared with two other existing models. Results indicate that the proposed model is a better alternative to existing models, with a root mean square error (RMSE) of 10.1% for impervious surface estimation in the study area.     

福州城区不透水面的光谱混合分析与识别制图

作为Ridd V-I-S模型中的一个重要组成部分,城市不透水面在监测城市扩展和解释人类活动对生态环境的影响起着非常重要的作用。利用图像处理技术,可以迅速地从遥感图像中提取城市不透水面信息。本文以福州城区为例,利用最小噪音分量变换法研究Landsat ETM 影像中城市不透水面信息的提取。通过选取最小噪音分量变换后的前3个分量和线性光谱混合模型,测算得到了高反照率、低反照率、植被及土壤4个模拟城市不同土地覆盖类型的终端地类分量。通过综合低反照率和高反照率两个终端地类,最后得到了不透水面分量。结果表明,城市不透水面的增加对城市生态环境有负面影响。     

Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery

This paper compares the normalized difference vegetation index (NDVI) and percent impervious surface as indicators of surface urban heat island effects in Landsat imagery by investigating the relationships between the land surface temperature (LST), percent impervious surface area (%ISA), and the NDVI. Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) data were used to estimate the LST from four different seasons for the Twin Cities, Minnesota, metropolitan area. A map of percent impervious surface with a standard error of 7.95% was generated using a normalized spectral mixture analysis of July 2002 Landsat TM imagery. Our analysis indicates there is a strong linear relationship between LST and percent impervious surface for all seasons, whereas the relationship between LST and NDVI is much less strong and varies by season. This result suggests percent impervious surface provides a complementary metric to the traditionally applied NDVI for analyzing LST quantitatively over the seasons for surface urban heat island studies using thermal infrared remote sensing in an urbanized environment.     

Scaling of impervious surface area and vegetation as indicators to urban land surface temperature using satellite data

Vegetation and impervious surface as indicators of urban land surface temperature (LST) across a spatial resolution from 30 to 960 m were investigated in this study. Enhanced thematic mapper plus (ETM+) data were used to retrieve LST in Nanjing, China. A land cover map was generated using a decision tree method from IKONOS imagery. Taking the normalized difference vegetation index (NDVI) and percent vegetation area (V) to present vegetated cover, and the normalized difference building index (NDBI) and percent impervious surface area (I) to present impervious surface, the correlation coefficients and linear regression models between the LST and the indicators were simulated. Comparison results indicated that vegetation had stronger correlation with the LST than the impervious surface at 30 and 60 m, a similar magnitude of correlation at 120 and 240 m, and a much lower correlation at 480 and 960 m. In total, the impervious surface area was a slightly better indicator to the LST than the vegetation because all of the correlation coefficients were relatively high (>0.5000) across the spatial resolution from 30 to 960 m. The indicators of LST, V and I are slightly better than the NDVI and NDBI, respectively, based on the correlation coefficients between the LST and the four indices. The strongest correlation of the LST and vegetation at the resolution of 120 m, and the strongest correlation between the LST and impervious surface at 120, 480 and 960 m, denoted the operational scales of LST variations.     

Characterizing fractional vegetation cover and land surface temperature based on sub-pixel fractional impervious surfaces from Landsat TM/ETM+

Estimating the distribution of impervious surfaces and vegetation is important for analysing urban landscapes and their thermal environment. The application of a crisp classification of land-cover types to analyse urban landscape patterns and land surface temperature (LST) in detail presents a challenge, mainly due to the complex characteristics of urban landscapes. In this article, sub-pixel percentage impervious surface areas (ISAs) and fractional vegetation cover (FVC) were extracted from bitemporal Thematic Mapper/Enhanced Thematic Mapper Plus (TM/ETM+) data by linear spectral mixture analysis (LSMA). Their accuracy was assessed with proportional area estimates of the impervious surface and vegetation extracted from high-resolution data. A range approach was used to classify percentage ISA into different categories by setting thresholds of fractional values and these were compared for their LST patterns. For each ISA category, FVC, LST, and percentage ISA were used to quantify the urban thermal characteristics of different developed areas in the city of Fuzhou, China. Urban LST scenarios in different seasons and ISA categories were simulated to analyse the seasonal variations and the impact of urban landscape pattern changes on the thermal environment. The results show that FVC and LST based on percentage ISA can be used to quantitatively analyse the process of urban expansion and its impacts on the spatial–temporal distribution patterns of the urban thermal environment. This analysis can support urban planning by providing knowledge on the climate adaptation potential of specific urban spatial patterns.     

Impacts of landscape structure on surface urban heat islands: A case study of Shanghai, China

Urbanization is taking place at an unprecedented rate around the world, particularly in China in the past few decades. One of the key impacts of rapid urbanization on the environment is the effect of urban heat island (UHI). Understanding the effects of landscape pattern on UHI is crucial for improving the ecology and sustainability of cities. This study investigated how landscape composition and configuration would affect UHI in the Shanghai metropolitan region of China, based on the analysis of land surface temperature (LST) in relation to normalized difference vegetation index (NDVI), vegetation fraction (Fv), and percent impervious surface area (ISA). Two Landsat ETM+ images acquired on March 13 and July 2, 2001 were used to estimate LST, Fv, and percent ISA. Landscape metrics were calculated from a high spatial resolution (2.5 × 2.5 m) land-cover/land-use map. Our results have showed that, although there are significant variations in LST at a given fraction of vegetation or impervious surface on a per-pixel basis, NDVI, Fv, and percent ISA are all good predictors of LST on the regional scale. There is a strong negative linear relationship between LST and positive NDVI over the region. Similar but stronger negative linear relationship exists between LST and Fv. Urban vegetation could mitigate the surface UHI better in summer than in early spring. A strong positive relationship exists between mean LST and percent ISA. The residential land is the biggest contributor to UHI, followed by industrial land. Although industrial land has the highest LST, it has limited contribution to the overall surface UHI due to its small spatial extend in Shanghai. Among the residential land-uses, areas with low- to-middle-rise buildings and low vegetation cover have much high temperatures than areas with high-rise buildings or areas with high vegetation cover. A strong correlation between the mean LST and landscape metrics indicates that urban landscape configuration also influences the surface UHI. These findings are helpful for understanding urban ecology as well as land use planning to minimize the potential environmental impacts of urbanization.     

Land Surface Temperature Data Fusion of Landsat ETM and MODISby Combining TsHARP and STITFM Model

Land Surface Temperature (LST) is an important parameter that describes energy balance of substance and energy exchange between the surface and the atmosphere,and LST has widely used in the fields of urban heat island effect,soil moisture and surface radiative flux.Currently,no satellite sensor can deliver thermal infrared data at both high temporal resolution and spatial resolution,which strongly limits the wide application of thermal infrared data.Based on the MODIS land surface temperature product and Landsat ETM+image,a temporal and spatial fusion method is proposed by combining the TsHARP (Thermal sHARPening) model with the STITFM (Spatio\|Temporal Integrated Temperature Fusion Model) algorithm,defined as CTsSTITFM model in this study.The TsHARP method is used to downscale the 1 km MODIS land surface temperature image to LST data at spatial resolution of 250 m.Then the accuracy is verified by the retrieval LST from Landsat ETM+ image at the same time.Land surface temperature image at 30 m spatial scale is predicted by fusing Landsat ETM+ and downscaling MODIS data using STITFM model.The fusion LST image is validated by the estimated LST from Landsat ETM+ data for the same predicted.The results show that the proposed method has a better precision comparing to the STITFM algorithm.Under the default parameter setting,the predicted LST values using CTsSTITFM fusion method have a root mean square error (RMSE) less than 1.33 K.By adjusting the window size of CTsSTITFM fusion method,the fusion results in the selected areas show some regularity with the increasing of the window.In general,a reasonable window size set may slightly improve the effects of LST fusion.The CTsSTITFM fusion method can solve the problem of mixed pixels caused by coarse\|scale MODIS surface temperature images to some degree.     

Estimating impervious surface distribution by spectral mixture analysis

Estimating the distribution of impervious surface, a major component of the vegetation-impervious surface-soil (V-I-S) model, is important in monitoring urban areas and understanding human activities. Besides its applications in physical geography, such as run-off models and urban change studies, maps showing impervious surface distribution are essential for estimating socio-economic factors, such as population density and social conditions. In this paper, impervious surface distribution, together with vegetation and soil cover, is estimated through a fully constrained linear spectral mixture model using Landsat Enhanced Thematic Mapper Plus (ETM+) data within the metropolitan area of Columbus, OH in the United States. Four endmembers, low albedo, high albedo, vegetation, and soil were selected to model heterogeneous urban land cover. Impervious surface fraction was estimated by analyzing low and high albedo endmembers. The estimation accuracy for impervious surface was assessed using Digital Orthophoto Quarterquadrangle (DOQQ) images. The overall root mean square (RMS) error was 10.6%, which is comparable to the digitizing errors of DOQQ images. Results indicate that impervious surface distribution can be derived from remotely sensed imagery with promising accuracy.     

基于多端元混合光谱模型与Landsat影像的北京不透水层动态研究

中国正在经历快速地城市化过程,及时又准确地掌握城市化过程对我国社会经济发展具有重要的实际意义。以Landsat-TM和ETM+为主要数据源,通过多端元光谱混合分析法(MESMA)提取北京建成区不透水层的时空演变信息。在Ridd的V-I-S(植被—不透水层—土壤)概念模型框架下,基于最小噪音变换(MNF)将TM或ETM+的6个光谱波段转换成MNF空间,并定义4种端元光谱分别代表植被、高反射率地表、低反射率地表和土壤,同时构建北京建成区端元光谱数据库。然后在MATLAB软件包中实现MESMA模型程序,依次提取北京市6个时段的不透水层信息。研究结果表明:MESMA方法能够提高植被、土壤和不透水层提取精度,相对误差分别为14.6%、17.3%和11.9%。研究结论充分说明MESMA方法应用到一个时间序列的中分辨率多光谱遥感影像是非常有效的。MESMA光谱分解方法能高效实现北京城市动态变化和城市扩张的监测。