Forest canopy closure is an important index on ecological study and forest manage­ment. Forest canopy closure is closely related to erosion control, water and soil conserva­tion, watershed runoff estimation, which is also used to assess wildlife habitat and forest biomass. In the early years, aerial photography were usually used to interprete forest canopy closure. However after the launch of Landsat-l in 1972, using remote sensing image to estimate forest canopy closure is widely studied. The study area of this research locates at Chi-Tou, Experimental Forest of National Taiwan University. A Landsat Thematic Mapper scene of this area was used in the study. In this study, geometric correction and radiometric correction were done first to improve the quality of the TM imagery. After preprocessing, seventy-two plots with a size of 3*3 pixels were drawn from the TM imagery. Fourteen vegetation indices including TMl、TM2、TM3、TM4、TM5、TM7、GVI、NDVI、R43、R54、R32、VI43、VI45、VI42. Those indices were used as input data for the fuzzy classification approach and the minimum distance pattern classification approach to estimate forest canopy closure, and the results of these two approaches were compared.

Image matching is one of the most important tasks in automated extraction of digital terrain models in digital photogrammetry. Area-based matching usually works well for the smooth terrain with low relief, especially when low resolution image pairs are used. However, area-based matching is not feasible for high resolution image pairs where artifi­cial buildings exist. On the other hand, feature matching has better chance to cope with the problem. Hence, we propose a scheme which extracts image features first then orga­ nizes a feature matching strategy, and develops a new feature-based matching scheme, i.e., Azimuth-Moment Matching method, to correspond the potential conjugate features. Experi­mental results indicate that for a pair of high resolution aerial photographs the proposed scheme may accomplish the task that conventional templet matching dose not.

Besides conventional methods, remote sensing technology is another method to quickly map water depths of broad coastal zones. The images of visible band are available analyzed to obtain the information of shallow water depths owing to discriminating reflec­tion of electromagnetic waves over the coastal zone with varied water depths. The visible band method is not suitable to eastern coastal zone of Taiwan due to its water depths almost beyond the penetrating depth of visible light and either western one because of high water turbidity. In this paper, the wave spectrum methodology is applied to bathymetry of shallow water. SPOT images were used to derive wave spectra and estimate water depths because of different lengths of waves over varied water depths. The bathymetry of shallow water is derived from wave spectra via the general dispersion relation of water waves. It bases on the hypothesis that waves keep their frequency uniform when they propagate from deep water into shallow water. Comparing with in situ water depths, the average error of estimating shallow water depths by wave spectrum bathymetric method is 15%. The moving average processing removes the random errors of water depths estimating by wave spectrum bathymetric method. Then the average error of water depths reduces to 8% ~ 10%. Wave spectrum bathymetric method takes advantages of remote sensing technology to obtain the synoptical bathymetry of shallow water in safety, economically and quickly. This method is a feasible approach over coastal zones of Taiwan with acceptable accuracy.

The land use of the aboriginal reserve around Shi Juo Village, Chia Yi County had been changed significantly as a result of rapid expansions of tea plantation along A-Li Shan highway. It is the reason that a study of land development in the area is of great interest. The theory of land rent was employed in this study and the empirical data were gener­ated by applying image classification technique of remote sensing and the spatial analytical capability of GIS. In addition, an econometric approach was used, which has an attempt to use five GIS data(i.e., altitude, slope, aspect, distance from near by trucking road, and distance from CBD) in stead of financial factors of land development.

For the geometric analysis of airborne side-looking SAR imagery, we use radargrammetric conditions that comprise range and Doppler equations. Second-order polynomials are adopted to describe the time-varying position parameters of a radar carrier. When estimat­ing the position parameters, both pixel spacing and squint angle are also treated as vari­ables. This simultaneous parameter estimation is an article's highlight. Together with a given digital terrain model, the estimated orientation parameters are used to rectify SAR slant-range imagery by an indirect (bottom-up) method. The reported algorithms for the orientation parameter estimation and the ortho-image generation yield root mean square errors of approximate 1.5 pixels, based on 15 independent check points. The paper is concluded by remarking on how to improve control point coordinate measurements.