Object-based classification with features extracted by a semi-automatic feature extraction algorithm – SEaTH

Object-based image analysis (OBIA) uses object features (or attributes) that relateto the pixels contained by the image object to assist in image classification. These object features include spectral, shape, texture and context features. With hundreds of available features, the identification of those that can improve separability between classes is critical for OBIA. The Separability and Thresholds (SEaTH) algorithm calculates the SEaTH of object–classes for the given features. The SEaTH algorithm avoids time-consuming trial-and-error practice for seeking important features and thresholds. This article tests the SEaTH algorithm on Landsat-7 Enhanced Thematic Mapper (ETM+) imagery in a heterogeneous landscape with multiple land cover classes. The results suggest SEaTH is a strong alternative to other automated approaches, yielding an agreement of 79% with reference data. In comparison, an object-based nearest neighbour classifier yielded 66% agreement and a pixel-based maximum likelihood classifier yielded 69% agreement.     

A forward/backward principal component analysis of Landsat-7 ETM+ data to enhance the spectral signal of burnt surfaces

In this paper, principal component analysis (PCA), a dimensionality reduction method, has been applied successfully as an image enhancement technique to improve the spectral signal of burnt surfaces. Forward/backward PCA (F/B PCA) and image differencing, which the proposed method consists of, creates a new spectral space that preserves the original spectral patterns while enhancing particular structures of the original satellite data. Burnt surfaces constitute a spectrally enhanced feature after selective removal of spectral information from the original Landsat-7 Enhanced Thematic Mapper data.Improvement of the spectral separability of burnt surfaces is most evident in spectral channels ETM+4 and ETM+7, where burnt surfaces already compose distinct spectral objects, and channels ETM+2 and ETM+5. This improvement is reasonable since the third PC axis, which is not considered in the back-transformation, is composed mainly of the spectral information in these channels. Another benefit of the technique is a reduction of interband correlation in the satellite data.No clear differences between the standardized and non-standardized F/B PCA were identified to recommend the use of one over the other. Both methods show advances in certain aspects. Finally, an increase of the separability value between burnt areas and dry vegetated areas from 0.473 to 1.06 and 1.31 was obtained with the use of the standardized and non-standardized F/B PCA, respectively.     

Spatial interpolation and image-integrative geostatistical prediction of mosquito vectors for arboviral surveillance

Remote sensing can augment traditional methods of mosquito species surveillance for arboviruses. Abundance and patterns of mosquito vectors of West Nile virus in Chesapeake, Virginia, USA, were studied using light trap collection data and a Landsat-7 Enhanced Thematic Mapper+ digital image for spatial interpolation and geostatistical mapping of the abundance of 24 species of mosquitoes capable of transmitting West Nile virus to humans. We evaluated spatial interpolation techniques including inverse distance weighting, ordinary kriging, co-kriging geostatistics using combined Landsat-7 tasselled cap transform indices (brightness, greenness, and wetness) to characterize habitats and breeding conditions. Results highlight gaps in surveillance coverage, geostatistical improvement of vector patterns and abundance, and spatial patterns of error. Constraints and opportunities for adoption of remote sensing and spatial analysis for mosquito control are identified and discussed.     

Estimating landscape imperviousness index from satellite imagery

This letter presents a practical method for landscape imperviousness estimation through the synergistic use of Landsat Enhanced Thematic Mapper Plus (ETM+) and high-resolution imagery. A 1-m resolution color-infrared digital orthophoto was used to calibrate a stepwise multivariate statistical model for continuous landscape imperviousness estimation from medium-resolution ETM+ data. A variety of predictive variables were initially considered, but only brightness and greenness images were retained because they were account for most of the imperviousness variation measured from the calibration data. The performance of this method was assessed, both visually and statistically. Operationally, this method is promising because it does not involve any more sophisticated algorithms, such as classification tree or neural networks, but offers comparable mapping accuracy. Further improvements are also discussed.     

The potential of spectral mixture analysis to improve the estimation accuracy of tropical forest biomass

A main limitation of pixel-based vegetation indices or reflectance values for estimating above-ground biomass is that they do not consider the mixed spectral components on the earth's surface covered by a pixel. In this research, we decomposed mixed reflectance in each pixel before developing models to achieve higher accuracy in above-ground biomass estimation. Spectral mixture analysis was applied to decompose the mixed spectral components of Landsat-7 ETM+ imagery into fractional images. Afterwards, regression models were developed by integrating training data and fraction images. The results showed that the spectral mixture analysis improved the accuracy of biomass estimation of Dipterocarp forests. When applied to the independent validation data set, the model based on the vegetation fraction reduced 5–16% the root mean square error compared to the models using a single band 4 or 5, multiple bands 4, 5, 7 and all non-thermal bands of Landsat ETM+.     

Historical forest biomass dynamics modelled with Landsat spectral trajectories

Estimation of forest aboveground biomass (AGB) is informative of the role of forest ecosystems in local and global carbon budgets. There is a need to retrospectively estimate biomass in order to establish a historical baseline and enable reporting of change. In this research, we used temporal spectral trajectories to inform on forest successional development status in support of modelling and mapping of historic AGB for Mediterranean pines in central Spain. AGB generated with ground plot data from the Spanish National Forest Inventory (NFI), representing two collection periods (1990 and 2000), are linked with static and dynamic spectral data as captured by Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) sensors over a 25 year period (1984–2009). The importance of forest structural complexity on the relationship between AGB and spectral vegetation indices is revealed by the analysis of wavelet transforms. Two-dimensional (2D) wavelet transforms support the identification of spectral trajectory patterns of forest stands that in turn, are associated with traits of individual NFI plots, using a flexible algorithm sensitive to capturing time series similarity. Single-date spectral indices, temporal trajectories, and temporal derivatives associated with succession are used as input variables to non-parametric decision trees for modelling, estimation, and mapping of AGB and carbon sinks over the entire study area. Results indicate that patterns of change found in Normalized Difference Vegetation Index (NDVI) values are associated and relate well to classes of forest AGB. The Tasseled Cap Angle (TCA) index was found to be strongly related with forest density, although the related patterns of change had little relation with variability in historic AGB. By scaling biomass models through small (∼2.5 ha) spatial objects defined by spectral homogeneity, the AGB dynamics in the period 1990–2000 are mapped (70% accuracy when validated with plot values of change), revealing an increase of 18% in AGB irregularly distributed over 814 km² of pines. The accumulation of C calculated in AGB was on average 0.65 t ha⁻¹ y⁻¹, equivalent to a fixation of 2.38 t ha⁻¹ y⁻¹ of carbon dioxide.     

Towards decadal soil salinity mapping using Landsat time series data

Salinization is one of the major soil problems around the world. However, decadal variation in soil salinization has not yet been extensively reported. This study exploited thirty years (1985–2015) of Landsat sensor data, including Landsat-4/5 TM (Thematic Mapper), Landsat-7 ETM+ (Enhanced Thematic Mapper Plus) and Landsat-8 OLI (Operational Land Imager), for monitoring soil salinity of the Yellow River Delta, China. The data were initially corrected for atmospheric effects, and then matched the spectral bands of EO-1 (Earth Observing One) ALI (Advanced Land Imager). Subsequently, soil salinity maps were derived with a previously developed PLSR (Partial Least Square Regression) model. On intra-annual scale, the retrievals showed that soil salinity increased in February, stabilized in March, and decreased in April. On inter-annual scale, soil salinity decreased within 1985–2000 (−0.74 g kg⁻¹/10a, p < 0.001), and increased within 2000–2015 (0.79 g kg⁻¹/10a, p < 0.001). Our study presents a new perspective for use of multiple Landsat data in soil salinity retrieval, and further the understanding of soil salinization development over the Yellow River Delta.     

苏南大运河沿线城市热岛现象的卫星遥感分析

本文利用陆地卫星TM热红外波段的数据对苏南大运河沿线城市苏州、常州、镇江的热环境作了分析,了解各市热岛效应的强度,分布和成因,调查了热岛内部的精细结构。研究表明TM热红外图像能较好地揭示城市下垫面的微细热现象的景观结构,反映温度场的分布,可为城市热环境质量评价和热源调查提供丰富的信息。     

归一化光谱分解模型在城市化信息提取中的应用

以2001年7月11日LandsatETM7影像和2009年7月16日TM影像为数据源,基于V-I-S理论模型,采用归一化光谱分解模型提取了乌鲁木齐市区范围内2个时段的植被、土壤、不透水层3个连续地表参数分量。通过对不透水层不同阈值的划分,提取了2时段的乌鲁木齐市城市发展的空间信息,结果较为满意;通过空间叠加计算方式获取了8年来乌鲁木齐市城市化发展的空间信息和主要拓展方向。结果表明,乌鲁木齐城市化发展速度较快,特别是北扩趋势显著。     

Vicarious Cross-Calibration of the China Environment Satellite Using Nearly Simultaneously Observations of Landsat-7 ETM+ Sensor

To understand the absolute radiometric calibration accuracy of the HJ-A CCD-1 sensors, image from these sensors were compared to nearly simultaneously image from Landsat-7 ETM+ sensors. Although the HJ-A CCD-1 sensor has almost the same wavelength of each central band and band width as Landsat-7 ETM+ sensor, there is slightly difference in spectral response function (SRF). The impacts of SRF difference effects would produce ~2 % uncertainty in predicting reflectance of HJ-A CCD-1 sensor using Landsat-7 ETM+ sensor. The reflectance observed by satellite at top-of-atmosphere generally depends on its’ geometric conditions. The results reveal that the impacts of geometrical conditions would impact on the vicarious cross-calibration accuracy, which should be removed. The performances of cross-calibration are calibrated and validated by four image pairs collected from Yellow River Delta, China, and Qingdao City, China, at four independent times. The results indicate that the HJ-A CCD-1 sensors can be cross calibrated to the Landsat-7 ETM+ sensors to within an accuracy of 3.99 % (denoted by Relative Root Mean Square Error) of each other in all bands except band 4, which has a 6.33 % difference.     

Geostatistical interpolation of SLC-off Landsat ETM+ images

The scan-line corrector (SLC) for the Enhanced Thematic Mapper Plus (ETM+) sensor, on board the Landsat 7 satellite, failed permanently in 2003. The consequence of the SLC failure (or SLC-off) is that about 20% of the pixels in an ETM+ image are not scanned. We aim to develop a geostatistical method that estimates the missing values. Our rationale is to collect three cloud-free images for a particular Landsat scene, taken within a few weeks of each other: the middle image is the target whose un-scanned locations we wish to estimate; the earlier and later images are used as secondary information. We visit each un-scanned location in the target image and, for each reflectance band in turn, predict the missing value with cokriging (resorting to kriging when there is not enough local secondary information to justify cokriging). For three Landsat scenes in different bio-regions of Queensland, Australia, we compared the performance of geostatistical interpolation with image compositing. Geostatistics was a generally superior estimator. In contrast to compositing, geostatistics was able to estimate accurately values at all un-scanned locations, and was able to quantify the variance associated with each prediction. SLC-off images interpolated with geostatistics were visually sensible, although changes in land-use from pixel to pixel affected adversely the accuracy of prediction. The primary disadvantage of geostatistics was its relatively slow computing speed. We recommend the geostatistical method over compositing, but, if speed takes priority over statistical rigour, a hybrid technique–whereby composites are corrected to the local means and variances of the bands in the target image, and any un-estimable locations are interpolated geostatistically–is an adequate compromise.     

Synoptic Monitoring of Gross Primary Productivity of Maize Using Landsat Data

There is a growing interest in monitoring the gross primary productivity (GPP) of crops due mostly to their carbon sequestration potential. Both within- and between-field variability are important components of crop GPP monitoring, particularly for the estimation of carbon budgets. In this letter, we present a new technique for daytime GPP estimation in maize based on the close and consistent relationship between GPP and crop chlorophyll content, and entirely on remotely sensed data. A recently proposed chlorophyll index (CI), which involves green and near-infrared spectral bands, was used to retrieve daytime GPP from Landsat Enhanced Thematic Mapper Plus (ETM+) data. Because of its high spatial resolution (i.e., 30 30 m/pixel), this satellite system is particularly appropriate for detecting not only between- but also within-field GPP variability during the growing season. The CI obtained using atmospherically corrected Landsat ETM+ data was found to be linearly related with daytime maize GPP: root mean squared error of less than 1.58 in a GPP range of 1.88 to 23.1 ; therefore, it constitutes an accurate surrogate measure for GPP estimation. For comparison purposes, other vegetation indices were also tested. These results open new possibilities for analyzing the spatiotemporal variation of the GPP of crops using the extensive archive of Landsat imagery acquired since the early 1980s.     

Relationship of wheat grain yield with spectral indices

An experiment was conducted during 1996–97 and 1997–98 to study spectral indices and their relationships with grain yield of wheat. Variations of ratio vegetation index (RVI), normalized differences vegetation index (NDVI). difference vegetation index (DVI), transformed vegetation index (TVI), perpendicular vegetation index (PVI) and greenness vegetation index (GVI) have been studied at anthesis stage under different moisture and nitrogen levels. Spectral indices were correlated with crop parameters and it was found that GVI was the best index for yield estimation (r = 0.91 ).     

光谱特征扩展的时间序列Landsat数据地表覆盖分类

时间序列遥感影像常用于地表覆盖监测及其变化监测。然而,利用时序遥感数据—尤其是中分辨率遥感数据监测地表覆盖变化,其方法基本是先对多期影像分别进行监督分类然后对比分类结果。由于这种方法需要对每期遥感影像单独选择分类训练样本,而对于历史影像,常常难以获得可靠的样本数据。本文基于遥感数据定量化处理,尝试利用光谱特征扩展方法对时间序列Landsat数据进行分类:首先,结合一种新的大气校正方法和相对辐射归一化方法,对时间序列Landsat数据进行定量化处理,以消除各期影像之间的辐射差异,获得地表反射率数据。然后,论文选择一期易于获得分类训练样本的反射率数据作为"参考影像",并结合样本数据提取不同地表覆盖类型的光谱特征。最后,将"参考影像"中提取的地物光谱特征,扩展到所有时间序列反射率数据进行分类。论文利用青藏高原玛多地区的5景Landsat数据对本文的方法进行了验证,结果显示:基于光谱特征扩展的分类方法,可有效对定量化处理后的Landsat数据进行分类,分类总体精度为88.35%—94.25%,分类结果和传统的单景监督分类结果具有较好的一致性。此外,研究也发现,"参考影像"和待分类图像获取时间的季相差异会影响其分类的精度。     

Spatiotemporal reflectance blending in a wetland environment

To understand the mechanism of wetland cover change with both moderate spatial resolution and high temporal frequency, this research evaluates the applicability of a spatiotemporal reflectance blending model in the Poyang Lake area, China, using 9 time-series Landsat-5 Thematic Mapper images and 18 time-series Terra Moderate Resolution Imaging Spectroradiometer images acquired between July 2004 and November 2005. The customized blending model was developed based on the enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM). Reflectance of the moderate-resolution image pixels on the target dates can be predicted more accurately by the proposed customized model than the original ESTARFM. Water level on the input image acquisition dates strongly affected the accuracy of the blended reflectance. It was found that either of the image sets used as prior or posterior inputs are required when the difference of water level between the prior or posterior date and target date at Poyang Hydrological Station is <2.68 m to achieve blending accuracy with a mean average absolute difference of 4% between the observed and blended reflectance in all spectral bands.     

Mapping Oolitic Iron Ore Deposits in the Ashumaysi Formation,Western Saudi Arabia,Using Different Remote Sensing Techniques Applied to Landsat 7 Enhanced Thematic Mapper Plus Data (ETM+)

Mapping the surficial extent of oolitic iron ore deposits hosted in the Oligo–Miocene sedimentary rocks of the Ashumaysi Formation, western Saudi Arabia, was carried out using Landsat 7 Enhanced Thematic Mapper Plus (ETM+) data. Ore samples were collected from four various locations in the study area, and were studied in the laboratory using the GER 3700 Spectroradiometer (0.4–2.5 µm) and X-ray diffraction (XRD). Principal component analysis (PCA), minimum noise fraction (MNF), and minimum distance classification were used and assessed to map mineralization zones in the study area. Good correspondences were observed between the results obtained from the above mentioned techniques, spectral reflectance analyses, and XRD. The confusion matrix results revealed that mapping of iron ores using MNF is better and more accurate than using PCA. Good matching was also observed between the spectral reflectance curves of the collected samples and the corresponding pixels from Landsat 7 ETM+. The results demonstrated the usefulness of the image processing and interpretation of Landsat 7 ETM+ data for the detection and delineation iron ore deposits in arid and semi-arid areas.     

Comparison Between Four Methods for Data Fusion of ETM＋ Multispectral and Pan Images

Four data fusion methods, principle component transform (PCT), brovey transform (BT), smoothing filter-based intensity modulation(SFIM), and hue, saturation, intensity (HSI), are used to merge Landsat--7 ETMq- multispectral bands with ETM panchromatic band. Each of them improves the spatial resolution effectively but distorts the original spectral signatures to some extent. SFIM model can produce optimal fusion data with respect to preservation of spectral integrity. However, it results the most blurred and noisy image if the coregistration between the multispectral and pan images is not accurate enough. The spectral integrity for all methods is preserved better if the original multispectral images are within the spectral range of ETM pan image.     

基于植被指数模型的淡水湖泊湿地景观信息提取

高邮湖湿地是江苏省重要湿地之一,对生态、环境控制、调节气候和保护生物多样性具有重要意义。采用2007年的LandsatTM影像作为遥感信息源,选择影像的光谱特征和比值植被指数(RVI)、差值植被指数(DVI)、归一化植被指数(NDVI)、归一化差异绿度指数(NDGI)、土壤调节植被指数(SAVI)和最佳土壤调节植被指数(OSAVI)6种植被指数做了光谱特征分析,从而确定出最佳指数模型,并基于决策树方法,实现研究区景观信息的遥感分类。研究结果表明,决策树分类法易于综合多种特征进行遥感影像分类,植被指数参与到决策树分类中能够提高分类的总体精度,其总体精度达到79.58%,Kappa系数为0.721 0,分类结果理想且人工参与灵活。     

A Comparative Study of Built-up Index Approaches for Automated Extraction of Built-up Regions From Remote Sensing Data

Accurate built-up information is imperative for loss estimation and disaster management after the occurrence of catastrophic events such as earthquake, tornado, tsunami and flood. These catastrophic events leave behind a trail of mass destruction with property and human losses amounting to millions. Once a natural disaster hits a region, built-up information is required within a short span of time for disaster management. Nowadays, earth observation satellite imagery serves as a promising source to extract the land use / land cover classes. However, the automatic extraction of urban built-up from remote sensing data is a known challenge in the remote sensing community. The normalized difference built-up index (NDBI) algorithm has been recognized as an effective algorithm for automatic built-up identification from medium spatial and spectral resolution satellite images. Few researchers have modified this algorithm and proposed new quantitative expressions for the built-up index. In this paper, three built-up index based, unsupervised built-up extraction algorithms have been reviewed and compared. An automated kernel-based probabilistic thresholding algorithm is used to assort the built-up index values, obtained from modified built-up index algorithms, into built-up and non built-up regions for enhancing the efficiency of the built-up detection process. Qualitative assessment of these algorithms involves computation of several parameters including recently developed parameters like allocation disagreement and quantity disagreement, and classical parameters such as error of omission, error of commission and overall accuracy. This paper presents a case study where the algorithms have been implemented on Landsat-5 Thematic Mapper (TM) image of the city of Delhi and its surrounding areas for detection of built-up regions automatically.     

Forest cover classification using Landsat ETM+ data and time series MODIS NDVI data

Forest cover plays a key role in climate change by influencing the carbon stocks, the hydrological cycle and the energy balance. Forest cover information can be determined from fine-resolution data, such as Landsat Enhanced Thematic Mapper Plus (ETM+). However, forest cover classification with fine-resolution data usually uses only one temporal data because successive data acquirement is difficult. It may achieve mis-classification result without involving vegetation growth information, because different vegetation types may have the similar spectral features in the fine-resolution data. To overcome these issues, a forest cover classification method using Landsat ETM+ data appending with time series Moderate-resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data was proposed. The objective was to investigate the potential of temporal features extracted from coarse-resolution time series vegetation index data on improving the forest cover classification accuracy using fine-resolution remote sensing data. This method firstly fused Landsat ETM+ NDVI and MODIS NDVI data to obtain time series fine-resolution NDVI data, and then the temporal features were extracted from the fused NDVI data. Finally, temporal features combined with Landsat ETM+ spectral data was used to improve forest cover classification accuracy using supervised classifier. The study in North China region confirmed that time series NDVI features had significant effects on improving forest cover classification accuracy of fine resolution remote sensing data. The NDVI features extracted from time series fused NDVI data could improve the overall classification accuracy approximately 5% from 88.99% to 93.88% compared to only using single Landsat ETM+ data.