基于无人机影像与面向对象-深度学习的滨海湿地植物物种分类

明确滨海湿地植物物种类型及其分布状况是实现滨海湿地精细化生物多样性监测的基础,对于滨海湿地的保护管理与生态可持续发展均具有重要意义。本研究以无人机可见光遥感影像为基础数据源,在定量分析最优分割尺度与最优分类特征组合的基础上,应用面向对象-U-net深度学习方法对闽江河口湿地植物物种类型进行分类,并与K最近邻、决策树、随机森林和贝叶斯分类方法进行精度对比分析,以期为滨海湿地植物物种遥感精细分类与生物多样性保护管理提供方法借鉴与科学参考。研究结果表明,利用面向对象-U-net深度学习方法提取不同滨海湿地植物物种类型的分类精度可达95.67%,总体精度较其他分类方法提高6.67%–13.67%, Kappa系数提高0.12–0.31,且分类整体性好。此外,实现植物物种光谱特征、形状特征、纹理特征与高度特征的最优特征选择对于有效提高湿地植物物种信息分类精度具有重要作用,应用最优分割尺度实现影像分割可提高整体分类效率。     

基于不同决策树的面向对象林区遥感影像分类比较

面向地理对象影像分析技术(GEOBIA)是影像分辨率越来越高的背景下的产物.如何提高高分辨率影像分类精度和分类效率是影像处理的重要议题之一.本研究对QuickBird影像多尺度分割后的对象进行分类,分析了C5.0、C4.5、CART决策树算法在林区面向对象分类中的效率,并与kNN算法的分类精度进行比较.利用eCognition软件对遥感影像进行多尺度分割,分析得到最佳尺度为90和40.在90尺度下分离出植被和非植被后,在40尺度下提取不同类别植被的光谱、纹理、形状等共21个特征,并利用C5.0、C4.5、CART决策树算法分别对其进行知识挖掘,自动建立分类规则.最后利用建立的分类规则分别对植被区域进行分类,并比较分析其精度.结果表明: 基于决策树的分类精度均高于传统的kNN法.其中,C5.0方法的精度最高,其总体分类精度为90.0%,Kappa系数0.87.决策树算法能有效提高林区树种分类精度,且C5.0决策树的Boosting算法对该分类效果具有最明显的提升.     

基于中分辨率TM数据的湿地水生植被提取

利用湿地水生植被生长旺盛、光谱反射较强、光谱信息比较丰富的8月份中分辨率Landsat TM和ETM+多光谱遥感影像,采用面向对象的分类方法,进行野鸭湖湿地水生植被的提取。研究表明:在提取过程中,通过对原始影像进行主成分变换和穗帽变换,将主要信息与噪声分离,不仅减小了数据冗余和波段间的相关性,而且增大了影像上湿地水生植被与其他地物类型光谱和空间信息的差异性,并结合野外水生植被光谱特征分析,选择归一化植被指数NDVI与归一化水体指数NDWI辅助分类,构建特征波段或波段组合,然后,确定适当的隶属度函数和阈值范围,构建分类决策树,完成湿地水生植被的自动分类,提高了影像分割与面向对象分类的精度,取得了较为理想的湿地水生植被提取结果。2002年和2008年两景影像的总体分类精度分别达到86.5%和85.44%,表明中分辨率TM影像可以满足湿地水生植被提取的需要,又因为其具有较高的波谱分辨率、极为丰富的信息量、相对较低的价格、长时间序列,可以作为近20a湿地水生植被提取和动态变化监测的主要数据源。     

面向对象的优势树种类型信息提取技术

森林植被优势树种类型信息的提取是遥感影像分类中的难点.面向对象分类方法是用高空间分辨率遥感数据实现精确类型信息提取的新方法.本文以2013年Quickbird影像作为基础数据,选择福建省三明市将乐林场为研究区,采用面向对象多尺度分割方法提取耕地、灌草地、未成林造林地、马尾松、杉木和阔叶树等类型信息.分类特征融合植被的光谱、纹理和多种植被指数3类特征信息,建立类层次结构,对不同层次分别用隶属度函数和决策树分类规则,最终完成分类,并与只用纹理与光谱特征相结合的方法进行对比.结果表明:融合纹理、光谱、多种植被指数的面向对象的分类方法提取研究区优势树种类型信息的精度为91.3％,比只用纹理和光谱的方法精度提高了5.7％.     

基于Sentinel–2影像特征优化的于桥水库水生植被提取

水生植被分布情况、结构和演变趋势对湿地生态环境变具有重要的指示意义和科学研究价值。基于Sentinel-2遥感数据,综合应用光谱信息、水体植被指数、最佳指数法(Optimal Index Factory,OIF)计算的纹理特征,结合随机森林分类法,构建特征优化后的随机森林水生植被提取模型,对于桥水库进行水生植被提取。结果显示:该方法能有效的提取出水生植被,总体精度为93.22%,Kappa系数为0.91。进一步与最大似然和支持向量机(SVM)方法进行对比分析,结果表明本算法的总体精度分别提高了19.96%、8.53%, Kappa系数分别提高了0.25、0.11。基于水生植被全年提取结果,分析了于桥水库的水生植被年内变化,发现于桥水库水生植被在五月份最繁盛,随后逐渐消减,直至十月份基本消亡。实验表明:特征优化后的随机森林分类法在Sentinel-2影像水生植被提取中具有较好的适用性。     

基于多源遥感数据的面向对象林分类型识别

林分类型的识别是森林资源监测的核心问题之一.为研究多源遥感数据协同的面向对象林分类型分类识别,采用Radarsat-2数据和QuickBird遥感影像协同进行面向对象分类.在面向对象分类过程中,采用3种分割方案:单独使用QuickBird遥感影像分割;单独使用Radarsat-2数据分割;Radarsat-2&QuickBird协同分割.3种分割方案均采用10种分割尺度(25～250,步长25),应用修正的欧式距离3指标评价不同分割方案的分割结果,确定最优分割方案及最优分割尺度.在最优分割结果的基础上,基于地形、高度、光谱及共同特征的不同特征组合,应用带有径向基(RBF)核函数的支持向量机(SVM)分类器进行杉木林、马尾松林、阔叶林3种林分类型识别.结果表明:与单独使用一种数据相比,Radarsat-2数据和QuickBird遥感影像协同方案在面向对象林分类型分类方面具有优势.Radarsat-2&QuickBird协同分割方案,以最优尺度参数100进行分割时,分割结果最好.在最优分割结果的基础上,应用两种数据源提取的全部特征进行面向对象林分类型识别的精度最高(总精度为86%,Kappa值为0.86).本研究结果不仅可为多源遥感数据结合进行林分类型识别提供参考和借鉴,而且对于森林资源调查和监测有现实意义.     

顾及无人机影像点云特征的绿地信息分类方法

采用无人机影像进行绿地信息分类时, 常利用影像光谱、纹理、形状等分类特征, 忽视了通过无人机影像生成点云构建的数字表面模型(Digital surface model, DSM)和数字高程模型(Digital elevation model, DEM)差异特征。基于此, 提出一种顾及无人机影像点云特征的绿地信息分类方法。方法首先基于摄影测量理论对研究区无人机影像进行空三计算, 并生成点云, 在此基础上构建DSM、DEM和数字正射影像(Digital Orthophoto Map, DOM); 然后, 利用DSM和DEM模型构建地物高度差异模型(normalized Digital Surface Model, nDSM); 最后, 利用可见光波段差异植被指数(Visible-band difference vegetation index, VDVI)对DOM进行植被与非植被分类, 并结合nDSM对植被进行分类。以昆明市呈贡区白龙潭公园为研究区进行绿地信息分类, Kappa系数精度达到0.862, 实验表明本文的方法对城市绿地调查具有实际意义。     

基于环境星与MODIS时序数据的面向对象森林植被分类

林区地形复杂、植被分布无序,且森林植被光谱信息相近,因而森林二级类型边界的确定成为土地覆盖遥感分类的难点。选择吉林省东部山区为研究区,以环境星影像(HJ-1 CCD)和中等分辨率成像光谱仪(MODIS)时序数据为基础,采用面向对象的分类方法进行森林植被类型的提取。分类特征参数主要选取了HJ-1 CCD的光谱和纹理特征,以及MODIS时序数据的物候特征。研究区总体分类精度为91.5%,Kappa系数为0.88,森林二级类型的分类精度均较高,其中落叶阔叶林的制图精度达到了97.1%。所用的面向对象分类方法与未加入物候特征的面向对象分类方法相比,森林二级类型的分类精度得到大幅度提高。     

基于高分辨率与高光谱遥感影像的北亚热带马尾松及次生落叶树种的分类

利用遥感数据开展森林资源树种的分类对森林资源的监测、森林可持续经营及生物多样性研究都有重要意义。该文以江苏南部丘陵地区的北亚热带天然次生林为研究对象, 利用LiCHy (LiDAR、CCD、Hyperspectral)集成传感器同期获取的高分辨率和高光谱数据, 进行冠幅识别和多个层次的树种分类: 首先, 对高分辨率影像进行基于边缘检测的多尺度分割, 提取出单木冠幅; 其次, 对高光谱影像进行特征变量提取, 并对提取出的特征变量利用信息熵原理选取优化特征变量; 然后, 分别利用全部特征变量和经优化的重要特征变量对森林树种及森林类型进行预分类; 最后, 在预分类结果中加入单木冠幅信息对森林树种及森林类型进行重分类, 并分析分类结果的精度。研究表明: 1)利用全部特征变量进行4个典型树种分类时, 总体精度为64.6%, Kappa系数为0.493; 而针对森林类型的分类精度为81.1%, Kappa系数为0.584。2)利用选取的优化特征变量分类精度略低于利用全部特征变量的分类精度, 其中对4个典型树种分类时, 总体精度为62.9%, Kappa系数为0.459; 而针对森林类型的分类精度为77.7%, Kappa系数为0.525。通过集成传感器同期获取的高分辨率和高光谱数据可以有效地进行北亚热带森林的树种分类及森林类型的划分。     

基于多时相无人机遥感影像优化河口湿地景观分类

河口湿地具有丰富的生物多样性和高度异质化的景观格局。针对河口湿地景观的复杂性,采用传统的基于单幅遥感影像的分类方法并不能得到较好的分类结果。本研究采用多时相无人机遥感影像参与分类,以优化河口湿地景观自动分类结果。选择天目湖上游平桥河河口湿地为研究区,选取4个季节的无人机影像为基础数据源,采用面向对象与决策树相结合的分类方法,针对不同季节组合的影像进行分类。结果表明:采用多时相无人机影像能显著提升分类效果,且参与分类的时相越多,效果越好;单季影像中,春季是最适合进行景观分类的季节,分类总体精度为62.7%,Kappa系数为0.59;当4个季节获取的影像同时参与分类时,分类总体精度为91.7%,Kappa系数为0.90;参与分类的时相光谱特征差异越大,分类效果提升越明显。本研究可为河口湿地景观分类提供技术支持,并提出了一种利用可见光无人机遥感影像进行湿地景观分类的新思路。     

High‐resolution riparian vegetation mapping to prioritize conservation and restoration in an impaired desert river

In highly impaired watersheds, it is critical to identify both areas with desirable habitat as conservation zones and impaired areas with the highest likelihood of improvement as restoration zones. We present how detailed riparian vegetation mapping can be used to prioritize conservation and restoration sites within a riparian and instream habitat restoration program targeting 3 native fish species on the San Rafael River, a desert river in southeastern Utah, United States. We classified vegetation using a combination of object‐based image analysis (OBIA) on high‐resolution (0.5 m), multispectral, satellite imagery with oblique aerial photography and field‐based data collection. The OBIA approach is objective, repeatable, and applicable to large areas. The overall accuracy of the classification was 80% (Cohen's κ = 0.77). We used this high‐resolution vegetation classification alongside existing data on habitat condition and aquatic species' distributions to identify reaches' conservation value and restoration potential to guide management actions. Specifically, cottonwood (Populus fremontii) and tamarisk (Tamarix ramosissima) density layers helped to establish broad restoration and conservation reach classes. The high‐resolution vegetation mapping precisely identified individual cottonwood trees and tamarisk thickets, which were used to determine specific locations for restoration activities such as beaver dam analogue structures in cottonwood restoration areas, or strategic tamarisk removal in high‐density tamarisk sites. The site prioritization method presented here is effective for planning large‐scale river restoration and is transferable to other desert river systems elsewhere in the world.     

Remote sensing imagery in vegetation mapping: a review

Aims Mapping vegetation through remotely sensed images involves various considerations, processes and techniques. Increasing availability of remotely sensed images due to the rapid advancement of remote sensing technology expands the horizon of our choices of imagery sources. Various sources of imagery are known for their differences in spectral, spatial, radioactive and temporal characteristics and thus are suitable for different purposes of vegetation mapping. Generally, it needs to develop a vegetation classification at first for classifying and mapping vegetation cover from remote sensed images either at a community level or species level. Then, correlations of the vegetation types (communities or species) within this classification system with discernible spectral characteristics of remote sensed imagery have to be identified. These spectral classes of the imagery are finally translated into the vegetation types in the image interpretation process, which is also called image processing. This paper presents an overview of how to use remote sensing imagery to classify and map vegetation cover.Methods Specifically, this paper focuses on the comparisons of popular remote sensing sensors, commonly adopted image processing methods and prevailing classification accuracy assessments.Important findings The basic concepts, available imagery sources and classification techniques of remote sensing imagery related to vegetation mapping were introduced, analyzed and compared. The advantages and limitations of using remote sensing imagery for vegetation cover mapping were provided to iterate the importance of thorough understanding of the related concepts and careful design of the technical procedures, which can be utilized to study vegetation cover from remote sensed images.     

基于面向对象的QuickBird遥感影像林隙分割与分类

传统的实地调查和人工解译方法已经不能满足区域尺度的林隙获取,高空间分辨率遥感影像的出现为区域尺度的林隙获取提供了可能.本研究采用QuickBird高空间分辨率光学遥感影像,结合面向对象分类技术对福建省三明市将乐县将乐国有林场进行林隙分割与分类.在面向对象分类过程中,采用10种尺度(10~100,步长为10)对QuickBird遥感影像进行分割,应用参考对象相交面积(RA_or)和分割对象相交面积(RA_os)进行分割结果评价.对每个尺度分割结果应用16个光谱特征,采用向量机分类器(SVM)进行林隙、非林隙和其他类型分类.结果表明:通过RA_or和RA_os等值法获得最优分割尺度参数为40.不同尺度参数之间的分类总精度最高相差22%.在最优尺度下,应用SVM分类器对林隙、非林隙和其他类型分类的总精度高达88%(Kappa=0.82).采用高空间分辨率遥感数据并结合面向对象的方法,可以代替传统的实地调查和人工解译对区域尺度的林隙进行识别分类.     

基于百度街景图像的行人视角城市街道植被绿化格局分析

城市街道绿化植被作为城市景观的重要组成部分, 其分布格局对城市景观美学发展及行人身心健康有显著影响, 立足行人视角准确监测街道绿植分布信息对城市规划与管理有明确的辅助作用。该文针对已有研究多采用沿天底方向垂直向下观测的遥感影像监测地表植被而对行人视角的绿色植被分布格局研究涉及不多的现状, 基于免费获取的百度街景图像, 选取绿植覆被典型的泰安市区为案例区, 结合网络信息抓取与空间地理信息处理技术, 分析百度街景图像提取侧视绿植信息的可行性, 统计并对比其计算结果与遥感影像提取结果的关系, 以期为城市规划与管理提供辅助参考信息。网络抓取案例区273个样点共3 276幅百度街景图像, 利用计算机监督分类提取图像中的绿植区域; 基于空间分析模型分析街道绿色植被的分布格局; 利用SPSS软件趋势拟合模块分析百度街景图像与遥感影像提取的植被信息的相关性。主要结果为: 百度街景图像可作为主数据源提取城市街道的侧视绿植分布情况; 案例区不同区域植被分布指数区别较大, 空间格局差异明显; 百度街道植被分布指数与基于遥感图像提取的10、20、50 m缓冲距离范围内植被覆盖面积呈显著正相关关系, 但两者的变化趋势并非完全一致。百度街道植被分布结果可作为遥感监测结果的辅助信息更好地指导城市绿色景观规划与精准管理。     

基于对象特征的山东省丘陵地区多时相遥感土地覆被自动分类

对于基于像元的土地覆被分类来说,植被的分类是难点。使用多时相面向对象分类方法可以较好的解决这个问题。以山东省烟台市丘陵地区为研究区,采用Landsat TM(Landsat Thematic Mapper remotely sensed imagery)、DEM(Digital Elevation Model)、坡度、坡位、坡向等多种数据,利用基于对象特征的多时相分类方法对研究区进行土地覆盖自动分类。首先对影像进行多尺度分割并检验分割结果选取合适的分割尺度,然后分析对象的光谱、纹理、形状特征。根据各类地物的光谱特征、地理相关性、形状、空间分布等特征,明确类别之间的差异。建立决策树使用隶属度函数进行模糊分类,借助支持向量机提高分类精度。研究结果表明,通过使用多时相影像采用面向对象分类方法,相对于传统的基于像素的分类可以明显提高分类精度,尤其是解决了乔灌草的区分问题。     

Computerized classification of Mediterranean vegetation using panchromatic aerial photographs

Abstract. Historical aerial photographs are an important source for data on medium- to long-term (10 - 50 yr) vegetation changes. Older photographs are panchromatic, and manual interpretation has traditionally been used to derive vegetation data from such photographs. We present a method for computerized analysis of panchromatic aerial photographs, which enables one to create high resolution, accurate vegetation maps. Our approach is exemplified using two aerial photographs (from 1964 and 1992) of a test area on Mt. Meron, Israel. Spatial resolution (pixel size) of the geo-rectified photos was 0.30 m and spatial accuracy (RMS error) ca. 1 m. An illumination adjustment prior to classification was found to be essential in reducing misclassification error rates. Two classification approaches were employed: a standard maximum-likelihood supervised classifier, and a modification of a supervised classification, which takes into account spectral properties of individual pixels as well as their neighbourhood characteristics. Accuracy of the maximum likelihood classification was 81 % in the 1992 image and 54 % in the 1964 image. The neighbour classifier increased accuracy to 89 % and 82 % respectively. The overall results suggest that computerized analysis of sequences of panchromatic aerial photographs may serve as a valuable tool for the quantification of medium-term vegetation changes.     

Characterization of moorland vegetation and the prediction of bird abundance using remote sensing

Aims To characterize and identify upland vegetation composition and height from a satellite image, and assess whether the resulting vegetation maps are accurate enough for predictions of bird abundance. Location South‐east Scotland, UK. Methods Fine‐taxa vegetation data collected using point samples were used for a supervised classification of a Landsat 7 image, while linear regression was used to model vegetation height over the same image. Generalized linear models describing bird abundance were developed using field‐collected bird and vegetation data. The satellite‐derived vegetation data were substituted into these models and efficacy was examined. Results The accuracy of the classification was tested over both the training and a set of test plots, and showed that more common vegetation types could be predicted accurately. Attempts to estimate the heights of both dwarf shrub and graminoid vegetation from satellite data produced significant, but weak, correlations between observed and predicted height. When these outputs were used in bird abundance–habitat models, bird abundance predicted using satellite‐derived vegetation data was very similar to that obtained when the field‐collected data were used for one bird species, but poor estimates of vegetation height produced from the satellite data resulted in a poor abundance prediction for another. Conclusions This pilot study suggests that it is possible to identify moorland vegetation to a fine‐taxa level using point samples, and that it may be possible to derive information on vegetation height, although more appropriate field‐collected data are needed to examine this further. While remote sensing may have limitations compared with relatively fine‐scale fieldwork, when used at relatively large scales and in conjunction with robust bird abundance–habitat association models, it may facilitate the mapping of moorland bird abundance across large areas.