Research Projects

In this undergraduate research project, our focus is on exploring the feasibility of applying unsupervised learning on radar images for the detection of collapsed buildings. We implement the Gray-Level Co-occurrence Matrix (GLCM) technique to extract clustering features from radar images, such as Angular Second Moment, Contrast, Entropy, and Homogeneity. The result reveals that the four features have insignificant correlations, thus more features or a dimension deduction process is recommended for further work.

I concurrently serve as a part-time research assistant while studying for my master's degree at National Yang Ming Chiao Tung University (NYCU). This is one of my side projects for the research assistant work. The objective of this study revolves around establishing a national gravity reference datum for various geophysical applications, such as precise leveling surveying. To achieve this goal, I developed the MATLAB scripts to collect the datasets and assess the accuracy of the relative and absolute gravity surveying.   

This is one of my side projects while studying for my master's degree at NYCU. Two superconducting gravimeters are operating in Taiwan, which is in Eighteen Peaks Mountain, Hsinchu as well as the Tatun Volcano Group (TVG). The objective is to identify transient signals arising from the fluid transmission processes within the TVG region. I have developed MATLAB scripts to decompose the tidal and seismic components from the gravity signals. The results are compared with the seismic waveforms and deformation analysis derived from the Persistent Scatterer Interferometric Synthetic Aperture Radar (PSInSAR) and the Global Navigation Satellite Systems (GNSS) time series. Preliminary analysis indicates minimal correlation among these observations. For further work, designing a filter or denoising process is suggested. Moreover, some of the empirical equations can be applied to the gravity signals to enhance the comparability with other observations.

In this study, Persistent Scatterer Interferometric Synthetic Aperture Radar (PSInSAR) is used to process ascending and descending Sentinel-1A images from April 2017 to January 2022. A two-dimensional velocity decomposition is appiled to determine the vertical and eastward deformation velocities. Before applying the decomposition, the contributions from the northward deformation velocities are computed from the Global Navigation Satellite Systems (GNSS) data and removed from the Line-of-Sight (LOS) deformation velocities. The results show the eastward deformation trends in the northeast area, the eastern Taipei Basin, and the southern Shanchiao Fault, which may be caused by tectonic activities. The Taipei Basin experienced land subsidence with rates up to 18.21 mm/yr and slight westward motions in its center. The basin underwent an uplifting period in 2019, which was related to the surface rebound caused by groundwater increase in the basin’s aquifers. The two fumaroles, Dayoukeng and Xiaoyoukeng in the TVG, underwent uplifting and subsidence in 2017–2022, which may be caused by the activity of the hydrothermal reservoirs under the volcano.

Geothermal energy serves as one of sustainable and low-emission energy sources with the potential to mitigate climate change and enhance energy security. It offers a viable substitute for conventional fossil fuels or electrical energy. The Hongchailin area in Ilan, Taiwan has been considered as a potential geothermal energy field in recent years. To investigate possible geothermal sources in Hongchailin, a dense seismic array comprising 186 geophones is deployed over a 5 × 4 km area covering the probable geothermal field between August 2022 and January 2023. A total of 41,095 P-arrivals are collected and used for seismic tomographic inversion. The velocity model shows several velocity anomaly zones in good spatial correlation with the resistivity model, although the resolvable depth of the model is limited to ~1 km. It demonstrates the active-source seismic tomography as a valuable geothermal exploration tool. Further, we employ unsupervised learning methods to classify and explore the resistivity-velocity relationships in each cluster. The preliminary results indicate a positive linear correlation for some regions but negative for some others, implying different materials such as rock composition or fluid content. These findings provide valuable insights for comprehensive understanding of geothermal resources in the Hongchailin area.

This project involves regular fieldwork for seismograph maintenance every three months, which is supervised by Hsin-Hua Huang in the Institute of Earth Sciences. Furthermore, I assess the data quality through monthly analyses of spectrograms, seismic signals, and the power spectral density plots.