Generally, atmospheric haze has important ramifications in remote sensing data. In this work,the theaerial photographs of ADS-40 image at Choshui River agricultural areas are divided into 4 regions which are thick clouds region, haze region, shadow region and normal region through a hybrid region detect model of ATSA (Automated Threshold Selection Algorithm) and MLC (Maximum Likelihood Classification) methods, respectively. In addition, for the different regions of atmospheric haze effects, this work will also develop an efficient hierarchical algorithm for mixed modes which are correct and verifies the image properly. These modes are as below: 1. Mosaic thick cloud region: prophase images provides new information for thick cloud region; 2. Restore haze region: spectral information was invested by DCPA (Dark Channel Prior Algorithm) method, and 3. Enhance shadow region: the Retinex algorithms were applied to enhance these shadow regions. After that, this work check the original and processed image quality assessments of this research processing with RMSE (Root Mean Square Error), PSNR (Peak Signal to Noise Ratio) and SSIM (Structure Similarity Image Measurement) indicators, respectively. It shows that the processed image with this hierarchical mixed mode of this study is better than those of the original image of ADS-40 aerial photographs. The index of RMSE reduced from 88.02 (original image) to 17.63 (processed image), PSNR index value improved from 9.24 (original image) to 23.20 (processed image) and SSIM value upgraded from 0.74 (original image) to 0.97 (processed image), respectively. Finally, in this study, the outcomes showed that the hierarchical mixed mode for image processing procedure exceeds that of the original image of agricultural areas evaluation from ADS-40 image.

With the rapid development of information technology (IT) and communication technology (CT),people can obtain every kind of digital data more convenient than before. The consequence is that the “copyright protection” which prevents digital data from been duplicated illegally should be paid much more attention. In the past, digital watermarking technology has been successfully applied to the “copyright protection” of various kinds of digital multimedia data, however the researches and  applications of applying digital watermarking to geo-information data are still very inadequate. In this study, the influence of the digital watermarking on the photogrammetric images is firstly analyzed. A novel digital watermarking algorithm based on the keypoints is proposed especially for photogrammetric images. The robustness of the embedded digital watermark and the impact on photogrammetric images are finally evaluated and analyzed. The experiments show that the proposed algorithm has high robustness on watermarking and little influence on the quality of the photogrammetric images. The success rate of image matching has been improved by 3% using the proposed watermarking algorithm.

Three-dimensional (3D) digital city is an emerging topic in the development and application of Geo-information. Building model is one of the most important components in a 3D digital city system. This paper presents viable strategies for generating 3D building models of different levels of details from assorted sources of spatial data, including large-scale topographic maps, aerial photographs, satellite images, close-range images and range data acquired with terrestrial laser scanners. Different approaches integrating photogrammetry, remote sensing and computer vision technologies are proposed according to different data and model requirements. The developed algorithms are designed for 3D modeling of complicated building pattens in the cities of Taiwan. The results demonstrated in this paper indicate that the proposed strategies and methods can generate 3D building models of different levels of details from different sources effectively. In addition, accuracy and cost analyses were also performed in order to provide more comprehensive references for future digital city implementations and related applications.

Absolute dynamical topography data derived from multi-satellite altimetry were used to analyze the formation and the transformation of oceanic eddies in the sea off southwestern Taiwan from 1992 to 2011. To track and detect eddies, we developed an integration filtering algorithm based on connected component labeling and the Okubo-Weiss parameter. Area, relative vorticity, nonlinearity parameter, amplitude, central velocity, lifetime, propagation pathway and kinetic energy of eddies are determined by the algorithm. The results indicate that eddies are almost generated around 100 km within the center of 21.5ºN and 120ºE. The number of warm eddy is more than that of cold eddy. Most warm eddies propagate to southwest along about 500 meter isobaths but cold eddies linger between 118oE and 121oE. The area, amplitude, lifespan, and eddy kinetic energy of warm eddies are greater than those of cold eddies. Eddies in the sea off southwestern Taiwan possess very high nonlinear property.

This study concentrated on satellite image’ correction processes and its application on water pollution estimation. In addition to the existing established Geometric Correction, the image correction was evaluated by combining radiation with atmospheric corrections. In SVR model’s input variables, we optimized from not only the original four bands in SPOT 4 and 5, but also the derived bands refer to different author' recommendations, Additionally, the Digital Elevation Model (DEM) was also used as one of the input variables. Therefore, two different model estimation procedures were evaluated: average river pollution index from simulated single river pollutant's score (RPI-Si) and average river pollution index from simulated single river pollutant's concentration (RPI-C). Finally, average river pollution index which predicted from the above compared best method was Co-Kriging (CK) with calculated RPI from few measured water quality sampling stations was also evaluated. The results showed that the short wave infrared band of SPOT image also has the positive contribution on the prediction results of river pollution index. Among different river pollution index estimation procedures, RPI_Si is the best way to estimate the final average water quality index. Furthermore, if we can interpolate the data of monitoring stations with the data predicted from SVR model for the display of river pollution index maps, the results indicated more improvement than that of only using SVR model. Conclusionally, when we combine the precised traditional river water quality monitoring data with the roughly advanced image predicted data from SVR model will be the feasible way as so far in order to establish a sketch map of surface river pollution index.