The original Digital Surface Model (DSM) produced from image matching in Digital Photogrammetric Workstation (DPW) or from scanning by Airborne Laser Scanner (ALS) are the mixing representation for ground, buildings, trees and many others. We propose two mathematic morphology methods, Opening Operation and H-Dome Transformation, to reconstruct DSM in elevation space to separate the ground surface from those objects above ground. The procedure of the proposed methods includes: (1)Opening Operation to filter out the noise in elevation space. (2)H-Dome Transformation to detect the local regional maxima where the objects above ground locate. NCU area and ETH area are used for test data in this experiment. There are over 90% of buildings locate on the segmented areas. And for strict assessment by generating DTM from DSM, the results of two test data compared with manual measurement show that the mean errors are 0.21m and 0.22m, and the RIVISE are l.04m and l.45m, respectively.

Hough Transform is a sound method for detecting straight lines in digital images. Considering an image with lattice structure, we modify the procedure in detecting straight lines using Hough Transform. We propose a Principal Axis Analysis Method to speed up the extraction of straight lines and increase the accuracy of the detected lines. We first transfer the parameters to a one-dimensional angle-count histogram, and then analyze the dominated distribution of image features. In peak detection, the searching priority is set according to the principal axis. Experimental results indicate that the accuracy and the efficiency for all of the cases are improved. The efficiency is better improved when the objects in an image are with more regularity.

The greatest advantage of utilizing remotely sensed image data is being capable of observing larger land section and its information of land use. By means of analyzing image data within certain period of time, it becomes easier to find out the development and transitional trend of land use. This study applied the Landsat TM satellite images obtained during 1984-1995 to analyze the transition of green land in urban area. The image data set enable us to supervise green land transition in Taipei within the ten years. According to important constructions and land use record, our main research directions are:(1)Analysis of green land transition in Taipei city; (2)The influences of river straitening project of Ki-long River to neighbor area; (3)Analysis of wet land transition and trend of urban development in Wu-Gu. From the transition of green land in Taipei city between 1984-1995, we could find the affinity between green, civil region, and the promotion of policies. Besides, the influences of past large hydraulic constructions to neighbor green land are also investigated.

Synthetic Aperture Radar (SAR) is a microwave sensor that can penetrate the clouds and rains. It can be operated day and night, because it provides its own illumination. Consequently , such a sensor bears numerous advantages for remote sensing applications. But a SAR generates the images by coherently processing the backscattered signals. ,hence, they are highly susceptible to the speckle noise. The speckle substantially reduces the effectiveness of image interpretation and target recognition. To reduce the speckle level , while maintain the spatial resolution, in this paper, we use a set of multi-temporal ERS-2 SAR images. The six scenes of different dates were first co-registered with minimal geometric shift error. By taking the time-space correlation properties, a temporal filter is used by means of multivariate analysis. We found that the equivalent number of looks (ENL) of the resulting image is improved from 4 to 20. The process of multi-temporal process greatly enhances the capability of land cover classification and change detection using SAR image.

A simple model in determining the location of the shoreline on coastal zones by remote sensing data is presented in this study. The main concept involves using the reflectance difference between the coastal land and its adjacent waters in the visible and near infrared spectral bands. Concurrent shoreline position measurements by GPS are also employed to construct and improve our approach. Three sets of SPOT HRV images, acquired in the clear sky condition, are used in this study. The results show that the gradient criteria of the shoreline position are stable, and can be used to find the shoreline easily in the clear sky cases. More examples should be investigated for constructing a database of gradient values for further applications around tideland areas in the future.

In order to identify the spatial pattern of urban land use change, this research designed a study applying the technique of Experts' Knowledge Base. The result shows: with the aid of GIS spatial database, it is feasible to integrate various infrastructures of data (RS,GIS), and simplified traditional time-consuming and sophisticated process. The use of experts knowledge and experiences, accompanied with the mechanism of logical inference extract from database, we can rapidly and efficient perform urban land use change identification. On the other hand, for probing into the multi-temporal land use change form, the research also build up a model to semi-auto detect the classification output. The processing stage can be divided into two major parts: image classification and land used change detection. This study concludes, using this process can achieve the goal of semi-auto land use change detection, and it's suitable to analyze the urban land use change.