In some image classifications the importance of classes varies, and it is desirable to weight allocation to selected classes. Often the desire is to weight
The use of remotely sensed data to estimate terrestrial properties usually involves the acquisition of ground data. Remotely sensed data are being applied to ever larger areas and the acquisition and use of ground data, being so expensive, requires optimization. This paper investigates a sampling strategy that has already been used to acquire ground data in support of National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (NOAA AVHRR) imagery of approximately 18 000 km2 of Cameroonian forest and attempts to validate both the strategy and the use of the ground data in regression modelling. Specifically, a geostatistical approach was used to quantify the variability in the scene, the precision of the ground data, the benefits of twostage sampling and the errors associated with regression modelling and prediction.
Methods for mapping the waterline at a subpixel scale from a soft image classification of remotely sensed data are evaluated. Unlike approaches based on hard classification, these methods allow the waterline to run through rather than between image pixels and so have the potential to derive accurate and realistic representations of the waterline from imagery with relatively large pixels. The most accurate predictions of waterline location were made from a geostatistical approach applied to the output of a soft classification (RMSE = 2.25 m) which satisfied the standards for mapping at 1 : 5000 scale from imagery with a 20 m spatial resolution.
Remote sensing is a major source of land cover information.Commonly, interest focuses on a single land cover class.Although a conventional multi-class classifier may be used to provide a map depicting the class of interest the analysis is not focused on that class and may be sub-optimal in terms of the accuracy of its classification.With a conventional classifier, considerable effort is directed on the classes that are not of interest.Here, it is suggested that a one-class classification approach could be appropriate when interest focuses on a specific class.This is illustrated with the classification of fenland, a habitat of considerable conservation value, from Landsat ETM+ imagery.A range of one-class classifiers are evaluated but attention focuses on the support vector data description (SVDD).The SVDD was used to classify fenland with an accuracy of 97.5% and 93.6% from the user's and producer's perspectives respectively.This classification was trained upon only the fenland class and was substantially more accurate in fen classification than a conventional multi-class maximum likelihood classification provided with the same amount of training data, which classified fen with an accuracy of 90.0% and 72.0% from the user's and producer's perspectives respectively.The results highlight the ability to classify a single class using only training data for that class.With a one-class classification the analysis focuses tightly on the class of interest, with resources and effort not directed on other classes, and there are opportunities to derive highly accurate classifications from small training sets.
Remote sensing has generally been used to study the role of tropical forests as a source of atmospheric carbon, primarily through land-use change, such as deforestation, and biomass burning. Regeneration of forest on previously cleared areas, however, is a significant carbon sink. The strength of this carbon sink is dependent on the age and composition of the regenerating forest. The ability to identify regenerating forest classes that may differ in terms of carbon sink strength was investigated with Landsat TM data of a test site near Manaus, Brazil. A number of forest age classes were defined from a time series of Landsat sensor data, and their separability in Landsat TM data was assessed by maximum likelihood classifications. A high level of class separability was observed with a weighted kappa coefficient of 0.8569 obtained for a classification of six forest regeneration classes. Of the classification errors observed most were found to be associated with the youngest forest age class. At the test site, however, two main successional pathways were followed and the differences between areas of forest of the same age but on different pathways was most apparent with the youngest forests. Splitting the regenerating forests by the successional pathway was found to increase classification accuracy, with a weighted kappa coefficient of 0.9315 observed for an 11 class classification. A range of tropical forest classes that vary in strength as a carbon sink could therefore be identified accurately from Landsat TM data. Although the broader generality of the results requires further investigation, this indicates the potential to use image classifications to scale-up point measurements of the carbon flux between regenerating forest classes and the atmosphere over large areas.
Freedom from restrictive assumptions that underlie many quantitative techniques make neural networks attractive for ecological investigations. The potential of the self organising feature map (SOFM) neural network for the classification, and to a lesser extent, ordination of vegetation data was investigated. The SOFM output was shown to correspond closely to classifications obtained from three alternative clustering algorithms, with similar samples located close together in the SOFM output space. Moreover, the classes were distributed spatially in the SOFM output by their relative similarity. This was evident with comparison against classifications derived at various levels of a hierarchical classification that revealed that the classes aggregated during each step of the hierarchical classification also tended to lie close together in the SOFM output space. As a consequence, the spatial distribution of classes in the SOFM output may represent the data in a manner similar to an ordination analysis. Some evidence for this inference is provided by comparison with the results of a standard ordination analysis.
The accuracy of a classification is fundamental to its interpretation, use and ultimately decision making. Unfortunately, the apparent accuracy assessed can differ greatly from the true accuracy. Mis-estimation of classification accuracy metrics and associated mis-interpretations are often due to variations in prevalence and the use of an imperfect reference standard. The fundamental issues underlying the problems associated with variations in prevalence and reference standard quality are revisited here for binary classifications with particular attention focused on the use of the Matthews correlation coefficient (MCC). A key attribute claimed of the MCC is that a high value can only be attained when the classification performed well on both classes in a binary classification. However, it is shown here that the apparent magnitude of a set of popular accuracy metrics used in fields such as computer science medicine and environmental science (Recall, Precision, Specificity, Negative Predictive Value, J, F1, likelihood ratios and MCC) and one key attribute (prevalence) were all influenced greatly by variations in prevalence and use of an imperfect reference standard. Simulations using realistic values for data quality in applications such as remote sensing showed each metric varied over the range of possible prevalence and at differing levels of reference standard quality. The direction and magnitude of accuracy metric mis-estimation were a function of prevalence and the size and nature of the imperfections in the reference standard. It was evident that the apparent MCC could be substantially under- or over-estimated. Additionally, a high apparent MCC arose from an unquestionably poor classification. As with some other metrics of accuracy, the utility of the MCC may be overstated and apparent values need to be interpreted with caution. Apparent accuracy and prevalence values can be mis-leading and calls for the issues to be recognised and addressed should be heeded.
SVM classification has great potential in remote sensing. The nature of SVM classification also provides opportunities for accurate classification from relatively small training sets, especially if interest is focused on a single class. Five approaches to reducing training set size from that suggested by conventional heuristics are discussed: intelligent selection of the most informative training samples, selective class exclusion, acceptance of imprecise descriptions for spectrally distinct classes, the adoption of a one-class classifier and a focus on boundary regions. All five approaches were able to reduce the training set size required considerably below that suggested by conventional widely used heuristics without significant impact on the accuracy with which the class of interest was classified. For example, reductions in training set size of ~90% from that suggested by a conventional heuristic are reported with the accuracy of the class of interest remaining nearly constant at ~95% and ~97% from the user's and producer's perspectives respectively.
The remote sensing of biophysical properties has generally relied on the use of data acquired in red and near-infrared channels only, often combined in a vegetation index such as the Normalized Difference Vegetation Index (NDVI). This is wasteful of information acquired in other channels and may prevent the accurate estimation of biophysical properties. The use of vegetation indices based on red and near-infrared radiation to estimate biophysical properties of tropical forests has met with little success and this may be due to the asymptotic nature of the relation between the indices and biophysical properties, the variable sensitivity of vegetation indices to vegetation biophysical properties in different environments, the low radiation reflected in red and near-infrared channels, and severe attenuation by atmospheric water and aerosols. For tropical forests the only feasible way to estimate biophysical properties at regional to global scales is through the use of the National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) which operates in three channels in addition to the red and near-infrared. The potential of the data acquired in all five AVHRR channels for the estimation of tropical forest biophysical properties was investigated. Radiometrically calibrated AVHRR imagery of Ghana were related to ground data on tree density and mean basal area. Radiation measured in the middle-and thermal-infrared channels were more strongly correlated with the forest biophysical properties than radiation acquired in red and near-infrared channels. Moreover, vegetation indices containing data acquired in the middle- and thermal-infrared were also more strongly correlated with biophysical properties than the widely used NDVI. Correlation coefficients of 0·71 and 0·23 were derived between the NDVI and tree density and basal area respectively, while the corresponding correlation coefficients for indices based on data acquired in AVHRR channels 3–5 were up to −0·88 and −0·52 respectively. Since current and proposed sensors operate in visible to thermal-infrared channels users should consider the potential of data acquired all channels and not just a subset.
The accuracy of a supervised classification is dependent to a large extent on the training data used. The aim in training is often to capture a large training set to fully describe the classes spectrally, commonly with the requirements of a conventional statistical classifier in mind. However, it is not always necessary to provide a complete description of the classes, especially if using a support vector machine (SVM) as the classifier. An SVM seeks to fit an optimal hyperplane between the classes and uses only some of the training samples that lie at the edge of the class distributions in feature space (support vectors). This should allow the definition of the most informative training samples prior to the analysis. An approach to identify informative training samples was demonstrated for the classification of agricultural classes in south‐western part of Punjab state, India. A small, intelligently selected, training dataset was acquired in the field with the aid of ancillary information. This dataset contained the data from training sites that were predicted before the classification to be amongst the most informative for an SVM classification. The intelligent training collection scheme yielded a classification of comparable accuracy, ∼91%, to one derived using a larger training set acquired by a conventional approach. Moreover, from inspection of the training sets it was apparent that the intelligently defined training set contained a greater proportion of support vectors (0.70), useful training sites, than that acquired by the conventional approach (0.41). By focusing on the most informative training samples, the intelligent scheme required less investment in training than the conventional approach and its adoption would have reduced the total financial outlay in classification production and evaluation by ∼26%. Additionally, the analysis highlighted the possibility to further reduce the training set size without any significant negative impact on classification accuracy.
Lakes are important to the healthy functioning of the urban ecosystem. The urban lakes in Wuhan, China, which is known as 'city of hundreds of lakes', are facing substantial threats mainly due to rapid urbanization. This paper focused on detecting the spatial and temporal change of urban lakes in Wuhan, using a long time series of Landsat and HJ-1A remotely sensed data from 1987 to 2016. The permanent disappearance and seasonal fluctuation of 28 main urban lakes were analysed, and their relationships with climatic change and human activities were discussed. The results show that most lakes in Wuhan had shrunk over the past 30 years resulting in a permanent change from water to land. The shrinkage was also most apparent in the central region of the city. Seasonal fluctuations of lake area were evident for most lakes but the relative important driving variable of lake area change varied between sub-periods of time for different lakes. The explanatory power of impervious surface to five-year permanent water change is 91.75%, suggesting that urbanization – as increasing impervious surface – had led to the shrinkage of urban lakes in Wuhan. In all, 128.28 km2 five-year permanent water disappeared from 1987 to 2016.
Super-resolution mapping (SRM) is a technique to estimate a fine spatial resolution land cover map from coarse spatial resolution fractional proportion images. SRM is often based explicitly on the use of a spatial pattern model that represents the land cover mosaic at the fine spatial resolution. Recently developed deep learning methods have considerable potential as an alternative approach for SRM, based on learning the spatial pattern of land cover from existing fine resolution data such as land cover maps. This letter proposes a deep learning-based SRM algorithm (DeepSRM). A deep convolutional neural network was first trained to estimate a fine resolution indicator image for each class from the coarse resolution fractional image, and all indicator maps were then combined to create the final fine resolution land cover map based on the maximal value strategy. The results of an experiment undertaken with simulated images show that DeepSRM was superior to conventional hard classification and a suite of popular SRM algorithms, yielding the most accurate land cover representation. Consequently, methods such as DeepSRM may help exploit the potential of remote sensing as a source of accurate land cover information.
No abstract is provided for this article.
The coastal zone is under considerable pressure from development and is subject to change. Consequently, shoreline monitoring has grown in importance. Remotely sensed imagery from satellite sensors has been used as an alternative to conventional methods, such as those based on the interpretation of aerial photography and ground-based surveying, for monitoring shoreline position. However, the accuracy of shoreline mapping from satellite sensor imagery has been limited because of the relatively coarse spatial resolution (>10 m) of the sensors commonly used. Because of major practical and financial constraints, very fine spatial resolution (<5 m) data are often impractical for mapping large stretches of shoreline, so refinement of image analysis methods are needed to extract the desired subpixel-scale information from relatively coarse spatial resolution imagery. In this paper, the potential to map the shoreline at a subpixel scale from a soft classification of relatively coarse spatial resolution satellite sensor imagery was evaluated. Unlike conventional approaches, the methods used allowed the shoreline to be mapped within image pixels and have the potential to yield an accurate and realistic prediction of shoreline location. The approach involved the use of a soft image classification to estimate the subpixel-scale thematic composition of image pixels, which were then located geographically through postclassification analysis. Specifically, a contouring and geostatistical method based on a two-point histogram was used to position geographically the shoreline within image pixels. The approach was applied to differently shaped shoreline extracts in imagery at two spatial resolutions. The most accurate prediction of the shoreline position from images with 16- and 32-m spatial resolutions were typically for a simple linear stretch of coast for which the smallest root mean square error values were ≤1.20 m. The shoreline predictions satisfied the map accuracy standards specified for large-scale maps.
