低盐湖泊水体盐度光学遥感反演研究——以博斯腾湖为例

利用光学遥感反演盐度,可以充分利用遥感数据的空间代表性,以及目前高分率遥感数据的高时空精度。本文利用MERIS(Medium Resolution Imaging Spectrometer)300 m数据,以干旱区的博斯腾湖(博湖)为例,探讨了光学遥感数据反演低盐湖泊水体盐度的可行性。结果显示:在开都河入流影响的博湖西南角,存在光学遥感反演盐度利用的黄色物质(CDOM)与盐度的反比关系,但相关性不高,而且在博湖区域不同时间、不同区域CDOM与盐度的关系都不同。博湖盐度低于3 g·L-1,而遥感数据计算盐度的精度约为1.1 psu,因而用光学遥感数据计算博湖盐度的误差太大。博湖本身CDOM与盐度关系的时空异质性以及相关性不高,目前光学遥感反演精度有限,因此,在博湖用光学遥感数据反演整个湖区的盐度有困难。用光学数据反演水体盐度要求盐度足够高,盐度和CDOM存在梯度,并满足CDOM扩散守恒,因此用光学遥感反演低盐湖泊水体盐度较为困难。     

基于GF-2高分辨率遥感影像的水体提取方法研究

针对GF-2卫星影像数据的特点,利用单波段阈值法、多波段算法、归一化水体指数(NDWI)阈值法、单波段阈值法与阴影水体指数(SWI)相结合的决策树法对刘家峡地区的水体信息进行了提取,同时提出一种单波段阈值法与增强阴影水体指数(ESWI)相结合的决策树分类法,并对以上几种水体提取的效果进行比较分析,发现单波段阈值法与ESWI相结合的决策树分类法能够有效地消除绝大部分阴影的影响,且提取效果较SWI的效果要好,NDWI与多波段算法提取效果次之,单波段阈值法提取效果较差.      

基于局部体样条函数的海洋盐度场建模

海洋盐度是海洋水文要素的重要组成部分,是海洋战场环境研究的重点。在介绍海水盐度数据特点的基础上,采用局部体样条函数法对散乱的大规模盐度场标量数据进行建模,并利用遗传算法和非线性规划相结合的方法解决体样条函数法的区域划分问题。通过一组实测数据进行建模实验并将得到的数据场进行可视化,表明该方法适用于大规模散乱数据场的建模,可为海洋战场环境的数据分析和场景构建提供建模手段。     

A Robust Algorithm for Constructing Pit-Free Canopy Height Model

Due to the penetration ability of airborne light detection and ranging (lidar) into tree crowns, data pits commonly appear in lidar-derived canopy height models (CHMs). They have a seriously negative effect on the quality of tree detection and subsequent biophysical measurements. To construct a pit-free CHM, an algorithm based on robust locally weighted regression and robust z-score was presented to remove data pits. The significant advantage of the new algorithm is parameter-free, which makes it efficient and robust for practical applications. A numerical test and a real-world example were respectively employed to assess the performance of our method for CHM construction, and its results were compared with those of three classical methods including natural neighbor interpolation of the highest point method, mean and median filters. The numerical test demonstrates that our algorithm is more accurate than the other methods for generating pit-free CHMs under the presence of data pits. The real-world example shows that compared with the classical methods, our method has a better ability of data pit removal. Moreover, our method performs better than the other methods for deriving plot-level maximum tree height from CHMs. In a word, the new method shows high potential for pit-free CHM construction.     

Estimating soil salinity in Pingluo County of China using QuickBird data and soil reflectance spectra

Soil salinization is a worldwide environmental problem with severe economic and social consequences. In this paper, estimating the soil salinity of Pingluo County, China by a partial least squares regression (PLSR) predictive model was carried out using QuickBird data and soil reflectance spectra. At first, a relationship between the sensitive bands of soil salinity acquired from measured reflectance spectra and the spectral coverage of seven commonly used optical sensors was analyzed. Secondly, the potentiality of QuickBird data in estimating soil salinity by analyzing the correlations between the measured reflectance spectra and reflectance spectra derived from QuickBird data and analyzing the contributions of each band of QuickBird data to soil salinity estimation Finally, a PLSR predictive model of soil salinity was developed using reflectance spectra from QuickBird data and eight spectral indices derived from QuickBird data. The results indicated that the sensitive bands covered several bands of each optical sensor and these sensors can be used for soil salinity estimation. The result of estimation model showed that an accurate prediction of soil salinity can be made based on the PLSR method (R² = 0.992, RMSE = 0.195). The PLSR model's performance was better than that of the stepwise multiple regression (SMR) method. The results also indicated that using spectral indices such as intensity within spectral bands (Int1, Int2), soil salinity indices (SI1, SI2, SI3), the brightness index (BI), the normalized difference vegetation index (NDVI) and the ratio vegetation index (RVI) as independent model variables can help to increase the accuracy of soil salinity mapping. The NDVI and RVI can help to reduce the influences of vegetation cover and soil moisture on prediction accuracy. The method developed in this paper can be applied in other arid and semi-arid areas, such as western China.     

基于车载点云数据的树木提取与分析

本文基于高速公路高精度点云数据,首先通过点云数据的分类处理实现对树木点云数据的提取,将树木点云投影到水平面,采用DBSCAN密度聚类算法实现单根树木的提取;然后在数据密集区域存在树木树冠点云重叠的区域,本文结合树干几何特征提取树干的位置信息,计算所有点云到树干中心的欧氏距离,将所有点云归类到最近的树干进行粗分割;最后根据粗分割的树木轮廓特征确定树冠模型与树冠中心,提出了采用基于密度特征的格网竞争算法对重叠的区域进行精细分割。试验表明,本文采用的树木分割方法能够实现单棵树木精确提取。     

Sensitivity of narrowband vegetation indices to boreal forest LAI,reflectance seasonality and species composition

There is growing evidence that imaging spectroscopy could improve the accuracy of satellite-based retrievals of vegetation attributes, such as leaf area index (LAI) and biomass. In this study, we evaluated narrowband vegetation indices (VIs) for estimating overstory effective LAI (LAI_eff) in a southern boreal forest area for the period between the end of snowmelt and maximum LAI using three Hyperion images and concurrent field measurements. We compared the performance of narrowband VIs with two SPOT HRVIR images, which closely corresponded to the imaging dates of the Hyperion data, and with synthetic broadband VIs computed from Hyperion images. According to the results, narrowband VIs based on near infrared (NIR) bands, and NIR and shortwave infrared (SWIR) bands showed the strongest linear relationships with LAI_eff over its typical range of variation and for the studied period of the snow-free season. The relationships were not dependent on dominant tree species (coniferous vs. broadleaved), which is an advantage in heterogeneous boreal forest landscapes. The best VIs, particularly those based on NIR spectral bands close to the 1200 nm liquid water absorption feature, provided a clear improvement over the best broadband VIs.     

利用无人机数码影像进行密植型果园单木分割

单木树冠提取对果树健康状态、营养成分、产量预测具有重要意义。无人机获取的高分辨率遥感影像作为低成本、低风险的数据源,为准确估计棵数、描绘树木冠层轮廓提供了新的技术手段。以往关于单木冠层轮廓提取的研究大多集中在森林或稀疏果园,以局部最大值滤波结果作为基于标记分水岭算法的种子点,该方法在密植型果园的表现并不理想。提出了一种适用于密植型果园、以区域型种子块作为标记的分水岭算法,通过最大似然法提取果树冠层生成冠层数字表面模型,利用高斯滤波结合形态学开运算及自适应阈值分割方法生成区域型种子块,并执行基于种子块标记的分水岭算法,实现密植型果园单木分割。实例研究结果表明,总体棵数查全率为95.22%,查准率为99.09%,得到单木轮廓提取总体准确率为93.45%,总体欠分割误差为5.87%,总体过分割误差为0.90%。与局部最大值种子点提取结果对比,总体准确度提高18.66%,精细树冠轮廓提取精度提高17.75%,可为地形平缓地区密植型果园单棵果树树冠提取提供参考。     

基于Delaunay三角网的等高线树生成方法

研究如何利用Delaunay三角网构建等高线树,提出一种新的等高线树生成方法。该方法充分利用Delaunay三角网在领域分析中的优势,通过两次利用Delaunay三角网来判明等高线的空间关系进而达到统一被图廓截断的等高线以生成等高线树的目的。本文将等高线作为约束边构建约束型Delaunay三角网,利用Delaunay三角网查找具有邻接关系的等高线,在此基础上结合邻近等高线的高程关系判明、识别,最终统一被截断的等高线;然后对统一后的等高线再次利用Delaunay三角网查找具有邻接关系的等高线对,利用等高线对的高程关系判断出其为父子关系或兄弟关系,据此将等高线插入到相应的位置,逐步生长成等高线树。同时给出了基于Delaunay三角网的等高线树生成方法的算法设计及试验结果。     

Estimating salinity stress in sugarcane fields with spaceborne hyperspectral vegetation indices

The presence of salt in the soil profile negatively affects the growth and development of vegetation. As a result, the spectral reflectance of vegetation canopies varies for different salinity levels. This research was conducted to (1) investigate the capability of satellite-based hyperspectral vegetation indices (VIs) for estimating soil salinity in agricultural fields, (2) evaluate the performance of 21 existing VIs and (3) develop new VIs based on a combination of wavelengths sensitive for multiple stresses and find the best one for estimating soil salinity. For this purpose a Hyperion image of September 2, 2010, and data on soil salinity at 108 locations in sugarcane (Saccharum officina L.) fields were used. Results show that soil salinity could well be estimated by some of these VIs. Indices related to chlorophyll absorption bands or based on a combination of chlorophyll and water absorption bands had the highest correlation with soil salinity. In contrast, indices that are only based on water absorption bands had low to medium correlations, while indices that use only visible bands did not perform well. From the investigated indices the optimized soil-adjusted vegetation index (OSAVI) had the strongest relationship (R² = 0.69) with soil salinity for the training data, but it did not perform well in the validation phase. The validation procedure showed that the new salinity and water stress indices (SWSI) implemented in this study (SWSI-1, SWSI-2, SWSI-3) and the Vogelmann red edge index yielded the best results for estimating soil salinity for independent fields with root mean square errors of 1.14, 1.15, 1.17 and 1.15 dS/m, respectively. Our results show that soil salinity could be estimated by satellite-based hyperspectral VIs, but validation of obtained models for independent data is essential for selecting the best model.     

Mapping the distributions of C3 and C4 grasses in the mixed-grass prairies of southwest Oklahoma using the Random Forest classification algorithm

The objective of this paper is to demonstrate a new method to map the distributions of C₃ and C₄ grasses at 30 m resolution and over a 25-year period of time (1988–2013) by combining the Random Forest (RF) classification algorithm and patch stable areas identified using the spatial pattern analysis software FRAGSTATS. Predictor variables for RF classifications consisted of ten spectral variables, four soil edaphic variables and three topographic variables. We provided a confidence score in terms of obtaining pure land cover at each pixel location by retrieving the classification tree votes. Classification accuracy assessments and predictor variable importance evaluations were conducted based on a repeated stratified sampling approach. Results show that patch stable areas obtained from larger patches are more appropriate to be used as sample data pools to train and validate RF classifiers for historical land cover mapping purposes and it is more reasonable to use patch stable areas as sample pools to map land cover in a year closer to the present rather than years further back in time. The percentage of obtained high confidence prediction pixels across the study area ranges from 71.18% in 1988 to 73.48% in 2013. The repeated stratified sampling approach is necessary in terms of reducing the positive bias in the estimated classification accuracy caused by the possible selections of training and validation pixels from the same patch stable areas. The RF classification algorithm was able to identify the important environmental factors affecting the distributions of C₃ and C₄ grasses in our study area such as elevation, soil pH, soil organic matter and soil texture.     

A soft-classification-based chlorophyll-a estimation method using MERIS data in the highly turbid and eutrophic Taihu Lake

Soft-classification-based methods for estimating chlorophyll-a concentration (C_chla) by satellite remote sensing have shown great potential in turbid coastal and inland waters. However, one of the most important water color sensors, the MEdium Resolution Imaging Spectrometer (MERIS), has not been applied to the study of turbid or eutrophic lakes. In this study, we developed a new soft-classification-based C_chla estimation method using MERIS data for the highly turbid and eutrophic Taihu Lake. We first developed a decision tree to classify Taihu Lake into three optical water types (OWTs) using MERIS reflectance data, which were quasi-synchronous (±3 h) with in situ measured C_chla data from 91 sample stations. Secondly, we used MERIS reflectance and in situ measured C_chla data in each OWT to calibrate the optimal C_chla estimation model for each OWT. We then developed a soft-classification-based C_chla estimation method, which blends the C_chla estimation results in each OWT by a weighted average, where the weight for each MERIS spectra in each OWT is the reciprocal value of the spectral angle distance between the MERIS spectra and the centroid spectra of the OWT. Finally, the soft-classification based C_chla estimation algorithm was validated and compared with no-classification and hard-classification-based methods by the leave-one-out cross-validation (LOOCV) method. The soft-classification-based method exhibited the best performance, with a correlation coefficient (R²), average relative error (ARE), and root-mean-square error (RMSE) of 0.81, 33.8%, and 7.0 μg/L, respectively. Furthermore, the soft-classification-based method displayed smooth values at the edges of OWT boundaries, which resolved the main problem with the hard-classification-based method. The seasonal and annual variations of C_chla were computed in Taihu Lake from 2003 to 2011, and agreed with the results of previous studies, further indicating the stability of the algorithm. We therefore propose that the soft-classification-based method can be effectively used in Taihu Lake, and that it has the potential for use in other optically-similar turbid and eutrophic lakes, and using spectrally-similar satellite sensors.     

CUBE曲面滤波参数联合优选关键技术及应用

CUBE(combined uncertainty and bathymetry estimator)算法是国际上主流的多波束测深异常值自动探测与处理算法,在国内外被广泛应用,但对其核心算法和参数知之甚少,不利于该项技术的国产化。本文详细阐述了CUBE算法的基本原理、数学模型、关键参数和处理步骤,进而建立了CUBE曲面滤波参数联合优选方法。通过选取典型地形区、参数试验、对比分析等步骤完成参数的联合优选,并用台湾浅滩实测数据进行了验证。结果表明,优化后的参数可有效提升多波束数据自动处理的精度和效率。本文成果可应用于国产多波束测深处理软件的深化研发以及多波束实测数据处理。     

基于无人机影像密集匹配点云的传统村落地面点提取及DEM生成——以湘西德夯村为例

目前,针对利用无人机技术在山地起伏大、山体植被密集区域,难以获取地面点及DEM等问题,本文提出了一种结合布料模拟算法和改进的局部最大值算法,利用树顶点、树高等植被信息,提取地面点,进而生成整个区域的DEM的方法。以中国传统村落德夯村为例,利用植被系数和高程信息将点云分割为植被密集区和非植被密集区两个部分。在非植被密集区,通过布料模拟算法和改进的局部最大值算法分别提取地面点和树顶点,计算平均树高;在植被密集区,通过该区域的树顶点推算得到植被密集区的近似地面点,最终将两部分的地面点云进行TIN插值得到该地区的DEM。试验结果表明,利用此方法生成的DEM均方根误差,在非植被密集区达0.037 m,植被密集区可达1.606 m,整体平均误差达1.492 m,总体精度较好,基本可以满足村落尺度空间分析的需求。     

机器学习算法在森林地上生物量估算中的应用

森林地上生物量是森林生产力的重要评价指标,对其进行高效监测对维持全球碳平衡和保护生态系统具有重要意义。本文首先基于冠层高度模型数据,通过分水岭分割算法得到单木冠幅边界;然后在单木冠幅范围内提取23个LiDAR变量,结合佩诺布斯科特试验森林的87组实测数据,利用随机森林和支持向量机建立森林地上生物量估算模型;最后对样地模型估算的结果进行了比较,讨论了预测结果及其精度。结果表明:本文选用的随机森林模型和支持向量机模型在估算森林地上生物量的应用中获得了较高的精度;并且,随机森林模型在基于机载雷达数据估测森林地上生物量中的估算精度更高,模型泛化能力更强,制图精度也更好,具有更好的适用性。     

决策树结合混合像元分解的中国竹林遥感信息提取

竹林是中国亚热带地区特殊而重要的森林资源,现有方法难以实现全国范围竹林时空分布信息快速准确提取。针对此问题,本研究利用2003年、2008年、2014年MODIS NDVI、反射率产品数据和省域Landsat分类数据,提出了基于决策树结合混合像元分解的全国竹林信息提取方法。首先,通过最大似然法获取中国林地分布信息;然后,在林地信息的基础上,构建决策树模型提取中国竹林分布信息;最后,采用线性最小二乘法混合像元分解得到中国竹林丰度图,并计算竹林面积。研究结果表明:(1)最大似然法提取的3个时期中国林地信息的生产者与用户精度均在90%以上,Kappa系数均值为0.93,为竹林信息提取奠定了基础。(2)C5.0算法构建的决策树模型能够很好的提取中国竹林时空分布信息,3个时期竹林分类精度均在80%左右。(3)在混合像元分解的基础上,统计得到的全国各省竹林估算面积与清查面积具有较高的相关性,R~2分别为0.98、0.97和0.95,RMSE范围为3.92万—9.58万ha,说明估算得到全国竹林面积与实际情况较为吻合。本研究所提出基于MODIS遥感数据运用C5.0算法决策树结合混合像元分解的方法,实现了全国竹林时空分布信息的准确提取,为全国竹林资源信息动态监测及管理提供了技术手段和数据支撑。     

基于Aster数据的新疆渭-库绿洲及其周边地区土壤盐分的定量反演

以新疆渭干河-库车河地区为研究区域,在野外调查采样的基础上,对土样进行实验室光谱测量并重采样与Aster波段相匹配,利用偏最小二乘回归建模方法建立了土壤盐渍化定量反演模型,其精度满足大区域的土壤盐渍化监测要求,表明该建模方法具有较好的普适性和稳定性。用10景Aster图像数据实现了该区域的土壤盐渍化定量反演与制图,反演的盐分分布与实地调查较为一致,为大面积区域性土壤盐渍化的遥感定量调查与监测提供了较为有效的技术方法。     

基于遗传算法的道路网综合模型

提出了一种基于遗传算法的道路网综合模型。实验表明，该模型在保持道路网的空间分布性和实用性方面具有庭好的效果。     

多平台点云数据的单木参数提取精度分析

以浙江省海宁市4种代表行道树(广玉兰、无患子、悬铃木、香樟树)为研究对象,结合无人机(UAV)影像和三维激光扫描数据,利用ContextCapture、LiDAR360软件完成点云拼接、滤波、降噪和编辑,通过迭代最近点算法实现点云精细匹配,完成多平台点云数据融合,进而得到数字表面模型与数字高程模型,并制作冠层高度模型;采用分水岭分割算法对不同行道树树种的冠层高度模型进行单木分割,并综合局部最大值法实现单木树高、冠幅的参数提取。结果表明,本文方法进行行道树单木分割的精度高,树高、冠幅参数提取值的效果好,满足行道树几何参数调查要求。     

结合面向对象和深度特征的高分影像树种分类

针对传统手工提取特征方法需要专业领域知识,提取高质量特征困难的问题,将深度迁移学习技术引入到高分影像树种分类中,提出一种结合面向对象和深度特征的高分影像树种分类方法。为了获取树种的精确边界,该方法首先利用多尺度分割技术分割整幅遥感影像,并选择训练样本作为深度卷积神经网络的输入。为了避免样本数量少导致过拟合问题,采用迁移学习方法,使用ImageNet上训练的VGG16模型参数初始化深度卷积神经网络,并利用全局平局池化压缩参数,在网络最后添加1024个节点的全连接层和7个节点的Softmax分类器,利用反向传播和Adam优化算法训练网络。最后分类整幅遥感影像,生成树种专题地图。以安徽省滁州市的皇甫山国家森林公园为研究区,QuickBird高分影像作为数据源,采用本文方法进行树种分类。试验结果表明,本文方法树种分类总体精度和Kappa系数分别为78.98%和0.685 0,在保证树种精度的同时实现了端到端的树种分类。