Ke-Sheng Cheng

Locating landslides using multi-temporal satellite images

Landuse/landcover change detection using remotely sensed images has been widely investigated. Most applications of this type involve either image differencing or image classification using multi-temporal images. If multi-temporal images are to be used for quantitative analysis based on their radiometric information, as in the case of change detection or landuse classification, geometric rectification and radiometric correction must be performed priori to subsequent image analyses. In particular, geometric rectification has significant effect on the accuracy of landuse change detection in areas of rugged terrain. Remote sensing image rectification is commonly done by applying a polynomial trend mapping (PTM) model to image coordinates and map coordinates of ground-control-points. A major drawback of the PTM model is that it does not capture the random characteristics of terrain elevation. In this study an ordinary kriging approach is applied for image-to-image registration. The approach considers residuals of the PTM model as anisotropic random fields and employs the ordinary kriging method for spatial interpolation of the residual random fields. Band-ratioing technique was also employed for relative radiometric normalization. From the grey-level histograms of pre- and post-event band-ratio images, we determined the percentage of landuse changes in the study area. Image differencing was then performed using the pre- and post-event band-ratio image pair. Finally, a grey-level threshold of the band-ratio difference image is determined as the value whose exceeding probability equals the areal percentage of landuse change. DTM data of the study area were also used to further restrict landslide areas to steep slope areas.

Combining Pixel Swapping and Contouring Methods to Enhance Super-Resolution Mapping

Combining super-resolution techniques can increase the accuracy with which the shape of objects may be characterised from imagery. This is illustrated with two approaches to combining the contouring and pixel swapping methods of super-resolution mapping for binary classification applications. In both approaches, the output of the pixel swapping method is softened to allow a contour of equal class membership to be fitted to it to represent the inter-class boundary. The accuracy of super-resolution mapping with the individual and combined techniques is explored, including an assessment of the effect of variation in the number of neighbors and zoom factor on pixel swapping based analyses. When combined, the error with which objects of varying shape were represented was typically greatly reduced relative to that observed from the application of the methods individually. For example, the root mean square error in mapping the boundary of an aeroplane represented in relatively fine spatial resolution imagery decreased from 14.41 m with contouring and 4.35 m with pixel swapping to 3.07 m when the approaches were combined.

Combining Hopfield Neural Network and Contouring Methods to Enhance Super-Resolution Mapping

The mixed pixel problem may be reduced through the use of a soft image classification and super-resolution mapping analyses. Here, the positive attributes of two popular super-resolution mapping methods, based on contouring and the Hopfield neural network, are combined. For a binary classification scenario, the method is based on fitting a contour of equal class membership to a pre-final output of a standard Hopfield neural network. Analyses of simulated and real image data sets show that the proposed method is more accurate than the standard contouring and Hopfield neural network based methods, with error typically reduced by a factor of two or more. The sensitivity of the Hopfield neural network based approaches to the setting of a gain function is also explored.

An Anisotropic Spatial Modeling Approach for Remote Sensing Image Rectification

Abstract Rectification of a remote sensing image is commonly done by applying polynomial regression models to image coordinates and map coordinates of ground control points. A major drawback of the polynomial regression model is that it does not capture the random characteristic of terrain elevation. In fact, the distortion of a remote sensing image is attributed to the variation of terrain elevation and orbital parameters, the variations being random in nature. A more effective approach of remote sensing image rectification is a stochastic approach that takes into account the spatial variation structure of terrain elevation. This article presents an anisotropic spatial modeling approach of image rectification using ordinary kriging estimation. By considering the residuals of polynomial trend mapping as anisotropic random fields, the proposed approach models separately the spatial variation structures of the residuals in X and Y directions, and employs the ordinary kriging method for spatial interpolation of the residual random fields. By means of a cross validation procedure, residuals of image rectification by the polynomial trend mapping, the multiquadric interpolation function, and the ordinary kriging approaches are compared. The ordinary kriging approach yields smallest variances and root-mean-squared of mapping errors.

Assessing the impact of climate change on annual typhoon rainfall—a stochastic simulation approach

Rainfall amount drawn by typhoon events accounts for a significant portion of annual rainfall in Taiwan. Changes in typhoon rainfall due to climate change may have severe consequences for water resources management. A stochastic simulation approach is proposed for evaluation of changes in typhoon rainfall under certain climate change scenarios. The number of typhoon events and total rainfall of individual typhoon events are, respectively, considered as random variables of the Poisson and Gamma distributions. Climate change scenarios were set by varying various degrees of changes in average number of typhoon events annually and the mean of event-total rainfall. Using stochastic simulation, basin-wide annual typhoon rainfalls were simulated for the Shihmen Reservoir watershed in northern Taiwan. It is found that 10% increases in average annual number of typhoon events and mean event-total rainfall will result in 18% increase in the annual typhoon rainfall of 5-year return period, whereas the annual typhoon rainfall of 10-year return period will increase by 15% under the same climate change scenario. Such increases may cause significant increase in reservoir sediment and pose challenges to reservoir management.

A Multivariate Model for Coastal Water Quality Mapping Using Satellite Remote Sensing Images.

This study demonstrates the feasibility of coastal water quality mapping using satellite remote sensing images. Water quality sampling campaigns were conducted over a coastal area in northern Taiwan for measurements of three water quality variables including Secchi disk depth, turbidity, and total suspended solids. SPOT satellite images nearly concurrent with the water quality sampling campaigns were also acquired. A spectral reflectance estimation scheme proposed in this study was applied to SPOT multispectral images for estimation of the sea surface reflectance. Two models, univariate and multivariate, for water quality estimation using the sea surface reflectance derived from SPOT images were established. The multivariate model takes into consideration the wavelength-dependent combined effect of individual seawater constituents on the sea surface reflectance and is superior over the univariate model. Finally, quantitative coastal water quality mapping was accomplished by substituting the pixel-specific spectral reflectance into the multivariate water quality estimation model.

Establishing acceptance regions for L-moments based goodness-of-fit tests for the Pearson type III distribution

Goodness-of-fit tests based on the L-moment-ratio diagram for selection of appropriate distributions for hydrological variables have had many applications in recent years. For such applications, sample-size-dependent acceptance regions need to be established in order to take into account the uncertainties induced by sample L-skewness and L-kurtosis. Acceptance regions of two-parameter distributions such as the normal and Gumbel distributions have been developed. However, many hydrological variables are better characterized by three-parameter distributions such as the Pearson type III and generalized extreme value distributions. Establishing acceptance regions for these three-parameter distributions is more complicated since their L-moment-ratio diagrams plot as curves, instead of unique points for two-parameter distributions. Through stochastic simulation we established sample-size-dependent 95% acceptance regions for the Pearson type III distribution. The proposed approach involves two key elements—the conditional distribution of population L-skewness given a sample L-skewness and the conditional distribution of sample L-kurtosis given a sample L-skewness. The established 95% acceptance regions of the Pearson type III distribution were further validated through two types of validity check, and were found to be applicable for goodness-of-fit tests for random samples of any sample size between 20 and 300 and coefficient of skewness not exceeding 3.0.

A Feature-Space Indicator Kriging Approach for Remote Sensing Image Classification

An indicator kriging (IK) approach for remote sensing image classification is proposed. By introducing indicator variables for categorical data, the work of image classification is transformed into estimation of class-dependent probabilities in feature space using ordinary kriging. Individual pixels are then assigned to the class with maximum class probability. The approach is distribution free and yields perfect classification accuracies for training data provided that collocated data in feature space do not exist. Technical considerations regarding implementation of IK such as indicator semivariogram modeling and handling of collocated data in feature space are also described. The IK, Gaussian-based maximum likelihood, nearest neighbor, and support vector machine (SVM) classifiers were applied to study areas within the Shimen reservoir watershed (case A: FORMOSAT-2) and Taipei city (case B: SPOT 4). The results show that the overall accuracies of the proposed IK classifier and SVM can achieve higher than 97% for training data and 81% for testing data. (The overall accuracies of IK are a little higher than those of SVM.) IK and SVM are found to be superior to the other two classifiers in terms of overall accuracies for both training and testing data. The proposed IK classifier has the following advantages: 1) It can deal with anisotropic problem in feature space; 2) it is a nonparametric method and needs not to know the type of probability distribution; and 3) it yields 100% classification accuracy for the training data provided that collocated data in feature space do not exist.

Assessing Uncertainty in LULC Classification Accuracy by Using Bootstrap Resampling

Supervised land-use/land-cover (LULC) classifications are typically conducted using class assignment rules derived from a set of multiclass training samples. Consequently, classification accuracy varies with the training data set and is thus associated with uncertainty. In this study, we propose a bootstrap resampling and reclassification approach that can be applied for assessing not only the uncertainty in classification results of the bootstrap-training data sets, but also the classification uncertainty of individual pixels in the study area. Two measures of pixel-specific classification uncertainty, namely the maximum class probability and Shannon entropy, were derived from the class probability vector of individual pixels and used for the identification of unclassified pixels. Unclassified pixels that are identified using the traditional chi-square threshold technique represent outliers of individual LULC classes, but they are not necessarily associated with higher classification uncertainty. By contrast, unclassified pixels identified using the equal-likelihood technique are associated with higher classification uncertainty and they mostly occur on or near the borders of different land-cover.

A Conceptual Model of Surface Reflectance Estimation for Satellite Remote Sensing Images Using in situ Reference Data

For satellite remote sensing, radiances received at the sensor are not only affected by the atmosphere but also by the topographic properties of the terrain surface. As a result, atmospheric correction alone does not yield output images that truly reflect terrain surface properties, namely surface reflectance (bidirectional reflectance factor, BRF) of objects on the earth surface. Following the concept of the radiometric control area (RCA)-based path radiance estimation method, we herein propose a statistical approach for surface reflectance estimation utilizing DEM data and surface reflectance of selected radiometric control areas. An algorithm for identification of shaded samples and a shape factor model were also developed in this study. The proposed RCA-based surface reflectance estimation method is capable of achieving good reflectance estimates in a region where elevation varies from 0 to approximately 600 m above the mean sea level. However, further study is recommended in order to extend the application of the proposed method to areas with substantial terrain variation.