Wen-Yen Chang

A Parallel Simulated Annealing Approach to Band Selection for High-Dimensional Remote Sensing Images

In this paper a parallel band selection approach, referred to as parallel simulated annealing band selection (PSABS), is presented for high-dimensional remote sensing images. The approach is based on the simulated annealing band selection (SABS) scheme which is originally designed to group highly correlated hyperspectral bands into a smaller subset of modules regardless of the original order in terms of wavelengths. SABS selects sets of correlated hyperspectral bands based on simulated annealing (SA) algorithm and utilizes the inherent separability of different classes to reduce dimensionality. In order to be effective, the proposed PSABS is introduced to improve the computational performance by using parallel computing technique. It allows multiple Markov chains (MMC) to be traced simultaneously and fully utilizes the parallelism of SABS to create a set of SABS modules on each parallel node. Two parallel implementations, namely the message passing interface (MPI) cluster-based library and the open multi-processing (OpenMP) multicore-based application programming interface, are applied to three different MMC techniques: non-interacting MMC, periodic exchange MMC and asynchronous MMC for evaluation. The effectiveness of the proposed PSABS is evaluated by NASA MODIS/ASTER (MASTER) airborne simulator data sets and airborne synthetic aperture radar (SAR) images for land cover classification during the Pacrim II campaign in the experiments. The results demonstrated that the MMC techniques of PSABS can significantly improve the computational performance and provide a more reliable quality of solution compared to the original SABS method.

Simulated annealing band selection approach for hyperspectral imagery

In hyperspectral imagery, greedy modular eigenspace (GME) was developed by clustering highly correlated bands into a smaller subset based on the greedy algorithm. Unfortunately, GME is hard to find the optimal set by greedy scheme except by exhaustive iteration. The long execution time has been the major drawback in practice. Accordingly, finding the optimal (or near-optimal) solution is very expensive. Instead of adopting the band-subset-selection paradigm underlying this approach, we introduce a simulated annealing band selection (SABS) approach, which takes sets of non-correlated bands for high-dimensional remote sensing images based on a heuristic optimization algorithm, to overcome this disadvantage. It utilizes the inherent separability of different classes embedded in high-dimensional data sets to reduce dimensionality and formulate the optimal or near-optimal GME feature. Our proposed SABS scheme has a number of merits. Unlike traditional principal component analysis, it avoids the bias problems that arise from transforming the information into linear combinations of bands. SABS can not only speed up the procedure to simultaneously select the most significant features according to the simulated annealing optimization scheme to find GME sets, but also further extend the convergence abilities in the solution space based on simulated annealing method to reach the global optimal or near-optimal solution and escape from local minima. The effectiveness of the proposed SABS is evaluated by NASA MODIS/ASTER (MASTER) airborne simulator data sets and airborne synthetic aperture radar images for land cover classification during the Pacrim II campaign. The performance of our proposed SABS is validated by supervised k-nearest neighbor classifier. The experimental results show that SABS is an effective technique of band subset selection and can be used as an alternative to the existing dimensionality reduction method.

Lateral variations of shallow shear-velocity structure in southwestern Taiwan inferred from short-period Rayleigh waves

Short-period fundamental-mode Rayleigh waves with periods of 1.1∼5.5 sec were used to investigate the lateral variations of shallow-depth shear-wave velocity structure up to a depth of 8 km under southwestern Taiwan. Through a priori regionalization, the region was divided into three subregions from the west to the east, and then the regionalized group velocity for each subregion was determined by a standard least-squares technique. By the structure inversion, the study region had obviously lateral velocity variations, which systemically increased from the west region to the east one. On the whole, the shear-wave velocity in the Western Foothills was higher than that in the Western Coastal Plain. Additionally, the three subregions all had a shear-wave velocity of less than 3 km/sec with the lowest one in the Western Coastal Plain near the coast, related to the thick sediments. These results were rather consistent with the geological features. For depths larger than 4 km, the velocity-gradient varying with depth in the Western Foothills was lower than that in the Western Coastal Plain. This is likely to interpret the reason that the seismic waves cannot be easily trapped within the Western Foothills; thus the short-period surface waves are poorly developed in that region.

A Study of Decadal Coastal Changes on Western Taiwan Using a Time Series of ERS Satellite SAR Images

In this paper, coastal line changes were monitored and analyzed from a sequence of ERS-1/2 SAR images covering the years 1996 to 2005, totaling 44 images for each year. Waterlines were extracted using a multi-scale edge detection algorithm, and further refined by means of morphology. Substantial analysis was carried out in conjunction with ground survey and sonar bathymetric mapping. In addition, tidal records were used to ensure all the shore lines been calibrated to the same tidal level. Results showed that Waisanting Sandbar, a north-southward sandbar, experienced significant accretion and erosion, moving southward about 700 meters during a 10-year period, and shrinking to just one third of its 1996 size. The surrounding coastal waters and the estuary of the Peikang River receded substantially, moving inward toward the coastal flat. The water channel became even more heavily deposited as a result. Finally, Haifengdao Sandbar, another sandbar, moved southward about 1.5 km, although its size remained the same from 1996 to 2005. It also showed a clear tendency to receding inward. We conclude that satellite remote sensing by SAR, aided by ground tidal data, bathymetric maps, and optical images, provides an effective and efficient tool for understanding coastal processes over large areas of coverage and long time spans.

Hyperspectral band selection based on parallel particle swarm optimization and impurity function band prioritization schemes

Abstract In recent years, satellite imaging technologies have resulted in an increased number of bands acquired by hyperspectral sensors, greatly advancing the field of remote sensing. Accordingly, owing to the increasing number of bands, band selection in hyperspectral imagery for dimension reduction is important. This paper presents a framework for band selection in hyperspectral imagery that uses two techniques, referred to as particle swarm optimization (PSO) band selection and the impurity function band prioritization (IFBP) method. With the PSO band selection algorithm, highly correlated bands of hyperspectral imagery can first be grouped into modules to coarsely reduce high-dimensional datasets. Then, these highly correlated band modules are analyzed with the IFBP method to finely select the most important feature bands from the hyperspectral imagery dataset. However, PSO band selection is a time-consuming procedure when the number of hyperspectral bands is very large. Hence, this paper proposes a parallel computing version of PSO, namely parallel PSO (PPSO), using a modern graphics processing unit (GPU) architecture with NVIDIA’s compute unified device architecture technology to improve the computational speed of PSO processes. The natural parallelism of the proposed PPSO lies in the fact that each particle can be regarded as an independent agent. Parallel computation benefits the algorithm by providing each agent with a parallel processor. The intrinsic parallel characteristics embedded in PPSO are, therefore, suitable for parallel computation. The effectiveness of the proposed PPSO is evaluated through the use of airborne visible/infrared imaging spectrometer hyperspectral images. The performance of PPSO is validated using the supervised K-nearest neighbor classifier. The experimental results demonstrate that the proposed PPSO/IFBP band selection method can not only improve computational speed, but also offer a satisfactory classification performance.

Estimated Seismic Intensity Distributions for Earthquakes in Taiwan from 1900 to 2008

In this study we recreated peak ground accelerations (PGA) and peak ground velocity (PGV) distributions for Taiwan by applying the attenuation relations of Liu and Tsai (2005) to calculate the PGA and PGV values for 1989 Mw ≥5:0 earthquakes in a catalog of earthquakes from 1900 to 2008 with homogenized magnitude (Mw )( Chen and Tsai, 2008). We further combined the PGA and PGV values to obtain corresponding modified Mercalli intensity (MMI) values (Wald, Quitoriano, Heaton, Kanamori, et al., 1999) and their spatial distributions and recurrence intervals. We adopted a logarithmic functional form analogous to the Gutenberg-Richter relation for seismicity to represent the annual frequency of seismic intensity parameters: log10N a blog10PGA, log10N a blog10PGV, and log10N a bI. The regions with high PGA and PGV values are often associated with low b values in these equations. As it is well known that the Mw 7.45 Chi-Chi earthquake of 21 September 1999 had produced high PGA values (in excess of 0:9g) and PGV values (in excess of 300 cm=s), we used these relations to estimate the Poisson probability distributions in Taiwan for MMI ≥ VIII (i.e., PGA ≥ 485g) for recurrence intervals of 30, 50, and 100 years. The results show a wide range of differences in the Poisson probability of MMI ≥ VIII among different areas of Taiwan. For example, for a 50-year interval, this probability at 10 major cities in Taiwan is as follows: Taipei 0.67%, Hsinchu 2.15%, Taichung 5.24%, Chiayi 24.35%, Tainan 1.61%, Kaohsiung 0.04%, Hengchun 4.94%, Ilan 17.67%, Hualien 37.04%, and Taitung 9.82%. These esti- mates should be of interest to city planners, especially for earthquake preparedness planning.

FORMOSAT-2 Mission: Current Status and Contributions to Earth Observations

This paper presents an overview of the current status and data applications of FORMOSAT-2, Taiwaneses first earth observation satellite mission. Highlights of its contributions to monitoring of global natural disasters and earth environmental changes will be illustrated. The FORMOSAT-2 satellite successfully complements existing high spatial resolution imaging satellites such as SPOT-5, IKONOS, and QuickBird, among others, with its unique capability of daily revisits worldwide. The FORMOSAT-2 follow-up program to ensure data continuity to the user community is briefly introduced.

Mapping Geo-Hazard by Satellite Radar Interferometry

This paper describes the application examples of satellite interferometric synthetic aperture radar (InSAR) in exploring potential geo-hazards and mapping geo-related disasters, including earthquakes, landslides, and land subsidence, triggered by natural forces and/or human activities. Satellite images were acquired from the Japan Aerospace Exploration Agencys (JAXAs) ALOS-PALSAR satellite, the Euroeapn Space Agencys (ESAs) Envisat-ASAR satellite, and the German Aerospace Centers (DLRs) TerraSAR-X satellite, covering frequency bands of L, C, and X, respectively. The study areas include Taiwan and Vietnam because both regions are prone to greater risks of geo-hazards and natural disasters. A series of C-band and L-band both ascending and descending mode SAR imagery data were used to map the deformation and subsidence of the environmentally sensitive areas. With the historical information and fast updated data sets, long-term and short-term subsidence patterns can be effectively and efficiently obtained. Excellent agreements were obtained as compared with Global Positioning System (GPS) and leveling measurements. Stacking of dual-beam mode as ascending and descending mode could extract the vertical displacement more easily and reduce the influence from the horizontal displacement. In summary, the advance of satellite radar interferometry provides a vital mapping to detect and identify geo-hazard and potential geo-disaster in general.

Probability for Simulating Future Earthquakes with Mw ≥6.0 in Taiwan for Seismic Hazard for the Earthquake Catalog from 1900 to 2008

In this study, we use extreme value theory based on Gumbel‐equation derivations to estimate the Gutenberg–Richter a and b parameters for Taiwan. Data are from the augmented, homogenized (in terms of moment magnitude), historic catalog for Taiwan. The island is divided into grids of 0.2° latitude by 0.2° longitude, and Gumbel type 1 statistical analysis is applied. The values of a and b are then used to determine the probability of large earthquakes ( M w≥6.0) occurring at each grid. The results show two relatively high probability paths for large earthquakes, one extending from Hsinchu southward to Taichung, Chiayi, and Tainan in western Taiwan and the other from Ilan southward to Hualian and Taitung in eastern Taiwan, both of which are characterized by low b ‐values. It indicates that future earthquakes can be expected along these paths characterized by low b ‐values. Additionally, maximum peak ground acceleration and maximum peak ground velocity (determined from respective attenuation laws and a gridding regimen of 0.1° latitude by 0.1° longitude for Taiwan) follow similar paths to that of the low b ‐values.

A Seismic Structure Study in the Kaoping Area, Southwestern Taiwan

The difference between S ‐wave and S ‐to‐ P ‐wave conversion ( S P phase) arrival times is enhanced with rectilinear motion detector filtering to describe alluvial‐sediment thickness in the Kaohsiung–Pingtung (Kaoping) plains area. A more complete understanding of the underground structures of the Kaoping area is provided in this paper and explains why the surrounding regions in Taiwan experience more earthquakes than the Kaoping area. Data are based on seismic activity recorded by the portable array for numerical data acquisition (PANDA) for the period from 1995 to 1997. The difference between S ‐wave and S P ‐phase arrival times shows that the sedimentary layer is thicker along the west and southwest coasts. P ‐wave travel‐time residuals, high‐frequency attenuation parameters Kappa, and quality factor Q P , Q S , and coda waves confirm this result. We also determined the orientation of the Chaochou fault using the first motion of P ‐wave arrivals. To the east of the Chaochou fault, stress trends southeast–northwest, while to the west, it trends northeast–southwest. The change in stress trends east and west of Chaochou fault suggests the presence of a highly fluid accretionary wedge in the Kaoping area. The Chaochou fault forms a seismically active tectonic boundary with the uplift of the hanging wall leading to westward tilting of the basement of the Kaoping plains. We demonstrate that these features are the reason there are relatively few earthquakes in the Kaoping area. The presence of a highly fluid accretionary wedge is indicated by a thick alluvial layer in the west and southwest Kaoping coasts; the Peikung High acts as the indenter that may allow seismic energy to escape and reduce the number of earthquakes in the region. Online Material: Figures illustrating calculations of Kappa, Q c , P ‐ and S ‐wave spectra, Q P , and Q S from ground‐motion data.