We propose the integration method that combines the Particle Image Velocimetry (PIV) technique and DEM subtraction, which is described to detect the active kinematics of the river morphology in this thesis. While PIV can provide estimates for the channel change direction and magnitude; vertical analysis such as river incision rate derivation could be performed using the height difference between multi-temporal surface models. It is observed that the proposed scheme is capable of identifying three river evolution stages from the extracted movements in the Taan case. These stages are: years 2001 to 2003, 2004 to 2006, and 2007 to 2010, respectively. The channel begins to have no stable path, and have regular path from 2001 to 2010. The maximum value of incision between 2001 and 2010 is about 15 m. This study demonstrates that both PIV and DSM subtraction are effective in river geomorphological change identification. The integration of these two approaches could provide more information when observing the evolution of river morphology.

Multi-lens multispectral camera system can be mounted on an Unmanned Aerial System (UAS) for high resolution remote sensing data acquisition. The adopted multi-lens structure allows it to acquire multi-band images with each independent camera and specific filter. However, this lead to significant band misregistration effects as the perspective centers, viewing angles, and lens distortion effects differ to each other. In this study, a Robust and Adaptive Band-to-Band Image Transform (RABBIT) is proposed to solve the band misregistration issues of multi-lens multispectral camera. The RABBIT utilizes a modified projective transform for transferring the multi-sensor geometry into one sensor geometry, in which all the necessary coefficients are derived from camera system calibration. Considering the coefficients uncertainty during the calibration, a simulation procedure is conducted to understand the systematic effects, and a robust and adaptive correction is thus developed to correct the various systematic errors. Several datasets were collected from three state-of-the-art multi-lens multispectral camera system for the accuracy analysis and reliability evaluation. The experiments show that RABBIT can achieve 0.2-0.4 pixels band co-registration accuracy, better than the other proposed method, and is robust to different cameras, different camera calibration approaches, and different flight missions.

Global warming has been a worldwide issue and significantly increasing icecap melting rate over polar area. Consequently the sea level rises continuously and poses a fundamental threat to whole human beings. Since the mass loss of Greenland ice sheet (GrIS) is highly correlated to the velocity of glacier movement, this study aims to monitor the impact of global warming by tracking glacier terminus displacement over GrIS using spaceborne remote sensing techniques which are widely applied in cryosphere monitoring for its continuous and efficient data collection ability. However, as many techniques are limited by the rapid changing and instable landscape dynamics of glacier surface, the present study utilizes the pixel-offset (PO) method with optical and synthetic aperture radar (SAR) images to track glacier deformation. In addition, by comparing the combinations of different quality images, processing parameters and pre-processing methods, we found the pan channel of Landsat-8 images and the log-processed HH polarization SAR images show the most promising results.

Fine Particulate Matter (PM2.5) is one of the pollutants which affects human health and has been attracting attention in recent years. Due to the uneven distribution and limited number of monitoring sites, observations from the monitoring stations are not capable of depicting the variability of PM2.5 over intra-urban areas in Taiwan. Land Use Regression (LUR) is one of the solution to deal with this challenge and has been widely applied in western cities recently. However, there are limited cases related to LUR in Taiwan. In this work, 6-year observations of 24 air quality monitoring stations located in the Northern Air Quality Zone of Taiwan operated by Environmental Protection Agency were collected. Compared with previous studies, several culture-specific emission sources such as temples, and Chinese restaurants, and a long-term satellite-based vegetation index were further considered to develop the LUR model at monthly and yearly resolutions for estimating the spatial-temporal variability of PM2.5. The R2 of the resultant models are 0.75 and 0.85 for monthly and yearly resolutions, respectively. Moreover, temples, and Chinese restaurants were both selected as important predictors in the developed models. The results of external data verification confirmed the robustness of the model performance. Finally, spatial-temporal variability of fine particle estimated using the developed models show a slightly decreasing during the study period over high polluted areas clustered in the urban cores of Taipei and Taoyuan cities.

Bridge is an important infrastructure construction for human living. Its surface’s crack and concrete delamination are important indicators that can reflect its safety status. In conventional, the bridge inspection was mainly conduct visually. Some of the process are dangerous and time-consuming to the inspector. Meanwhile, the judgement is subjective. In this research, we utilize object-based image analysis for automatic crack detection from UAV acquired images. Most case studies present more than 90% producer accuracy and more than 80% user accuracy. Later, based on the detected crack position on the image, we convert the original image into epipor one. Then, utilize image matching to detect conjugate point and perform space forward intersection to obtain the three-dimensional information of crack. For the concrete delamination, the approximate location of concrete delamination is manual selected. Then, we conduct dense image matching to obtain 3D point cloud and convert to a digital surface model (DSM) by subtract a fitted plane. By means of the elevation difference, we can detect the delamination area and calculate its volume. In the future works, the change detection and analysis of crack’s three-dimensional information between two different times can assist the safety inspection of bridges.