Koki Iwao

An Automated Method for Global Urban Area Mapping by Integrating ASTER Satellite Images and GIS Data

We present an automated classification method for global urban area mapping by integrating satellite images taken by Visible and Near-Infrared Radiometer of Advanced Spaceborne Thermal Emission and Reflection radiometer (ASTER/VNIR) and GIS data derived from existing urban area maps. The method consists of two steps. First, we extracted urban areas from ASTER/VNIR satellite images by using an iterative machine-learning classification method known as Learning with Local and Global Consistency (LLGC). This method is capable of automatically performing classification with a noisy training dataset, in our case, low-resolution urban maps. Therefore, we were able to perform supervised classification of ASTER/VNIR images without using labor-intensive visual interpretation. Second, we integrated the LLGC confidence map with other maps by logistic regression. The logistic regression complemented misclassifications in the LLGC map and provided useful information for further improvement of the model. In an experiment including 194 scenes of ASTER/VNIR images, the integrated maps were developed at a resolution of 15 m resolution, which is much finer than existing maps with resolutions of 300 to 1000 m. The maps achieved an overall accuracy of 90.0% and a kappa coefficient of 0.565, both of which are higher than or almost equal to the values for major existing global urban area maps.

Unsupervised Global Urban Area Mapping via Automatic Labeling from ASTER and PALSAR Satellite Images

In this study, a novel unsupervised method for global urban area mapping is proposed. Different from traditional clustering-based unsupervised methods, in our approach a labeler is designed, which is able to automatically select training samples from satellite images by propagating common urban/non-urban knowledge through the unlabeled data. Two kinds of satellite images, captured by the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Phased Array L-band Synthetic Aperture Radar (PALSAR), are exploited here. In this method, spectral features are first extracted from the original dataset, followed by coarse prediction of urban/non-urban areas via weak classifiers. By developing an improved belief-propagation based clustering algorithm, a confidence map is obtained and training data are selected via weighted sampling. Finally, the urban area map is obtained by employing the Support Vector Machine (SVM) classifier. The proposed method can generate urban areamaps at a resolution of 15 m, while the same settings are used for all test cases. Experimental results involving 75 scenes from different climate zones show that our proposed method achieves an overall accuracy of 84.4% and a kappa coefficient of 0.628, which is competitive relative to the supervised SVM method.

Inter-Band Radiometric Comparison and Calibration of ASTER Visible and Near-Infrared Bands

The present study evaluates inter-band radiometric consistency across the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) visible and near-infrared (VNIR) bands and develops an inter-band calibration algorithm to improve radiometric consistency. Inter-band radiometric comparison of current ASTER data shows a root mean square error (RMSE) of 3.8%–5.7% among radiance outputs of spectral bands due primarily to differences between calibration strategies of the NIR band for nadir-looking (Band 3N) and the other two bands (green and red bands, corresponding to Bands 1 and 2). An algorithm for radiometric calibration of Bands 2 and 3N with reference to Band 1 is developed based on the band translation technique and is used to obtain new radiometric calibration coefficients (RCCs) for sensor sensitivity degradation. The systematic errors between radiance outputs are decreased by applying the derived RCCs, which result in reducing the RMSE from 3.8%–5.7% to 2.2%–2.9%. The remaining errors are approximately equal to or smaller than the intrinsic uncertainties of inter-band calibration derived by sensitivity analysis. Improvement of the radiometric consistency would increase the accuracy of band algebra (e.g., vegetation indices) and its application. The algorithm can be used to evaluate inter-band radiometric consistency, as well as for the calibration of other sensors.

Classification of multi-scene high-spatial resolution images by using information obtained from temporal low-spatial resolution images

Elaborate jobs are required for land cover classification using multi-scene high-spatial resolution satellite images, like selecting training area on each scene with sufficient a priori knowledge. The classification method proposed in this paper is assumed to use both high-spatial resolution images and time-series low-spatial resolution images. It can automatically produce training data set on each scene, optimized considering land-cover features to the scene. Moreover, it prevents from deteriorating into low accuracy classification result, by referring to the class candidate information derived from time-series low-spatial resolution images. Experiments were conducted that used Landsat TM and NOAA AVHRR images as high-spatial and low-spatial resolution images respectively. Validation results by using three visually interpreted TM images demonstrate the optimization of training data set improved the classification accuracy from 56.0% to 66.2%, and the class candidate in formation did from 61.9% to 66.2%.