The objective of this investigation has two folds. One is to simulate ROCSAT-2 images and the other is to develop a applicable model for geometric correction to ROCSAT-2 images. We will use orthorectified aerial image and digital terrain model as referance data to perform imaging simulation. This is due to the unavailability of ROCSAT-2 image. For ROCSAT-2 image and on-board data simulation, we need to establish a satellite-orbit-model. This model can be divided to four parts. The first one is coordinate-transformation-model which transforms all coordinate systems in this investigation. The second one is satellite-f1y-model which describes the satel­lite orbit. The third one is simulation-model which simulates ROCSAT-2 image and on-board data. The fourth one is on-board-data-model which simulates on- board data by assuming ROCSAT-2 flying case and adding errors to satellite position and attitude data. The errors include systematic errors and random errors. We compute orientation parameters and position using simulation data. In single image case, we use on-board data to establish initial orbit and correct the orbit with ground control points. Then we use least squares filtering to fine tune the orbit and perform the validation. In stereo images case, we adjust 3-D position in object space by using least squares filtering after correcting orbits of two images. Then the validation is performed. Experimental results indicate that the systematic errors may be mostly compensated while the random errors consistently influence the positioning accuracy.

Digital cameras are getting widely used due to their operation conveniences and real-time processing capabilities. Although digital cameras have been considered suitable to photogrammetric tasks, their metric characteristics are not clear and stable as traditional metric cameras. A real-time on-the-job calibration is therefore, required before a digital camera can be applied for a photogrammetric task. The techniq ue of self­calibration is considered an effective means for digital camera calibration, but a large amount of image measurements are required. To solve this problem, we designed a new text-coded target and proposed an automatic process to detect, position, and identify the targets. Automatic camera calibration can be achieved by integrating this automatic measuring process and bundle adjustment computation. The experimental results are encouraging and optimally show the feasibility of the proposed method.

Developing a 3D city spatial information system to vividly model a city is currently an intensive research area in the field of digital photogrammetry. In this study area, extracting building information is the most important work. This study proposed a semi-automatic CSG model-based building extraction method to acquire precise, reliable and complete 3D data of buildings. In this paper how to apply CSG theory for building models and related model operations are illustrated, so that we can employ it in the semi-automatic model-based building extraction. To investigate how the CSG model fits real buildings we analyzed and classified the characteristics of common building shapes. Based on the investigation, a CSG building model base containing simple and compound primitives is proposed. By combining the primitives we can con­struct complex building models. Each building primitive will be associated with a set of parameters that can be modified to perform model-image fitting. The methods and tests of model operation for model-image fitting will be introduced. This study draws a blueprint for semi-automatic building extraction.

Semi-automatic model-based building extraction is the major research topic of digital photogrammetry. The affectiveness of model-image fitting algorithm is the key of semi-automatic model-based building extraction. In this paper a building is modeled by some combinative primitives, and model-image fitting is proceeded primitive by primitive. After fitting, complex buildings can be modeled by combining simple primitives defined by the theory of Constructive Solid Geometry (CSG). A model primitive can be trans­formed into the object coordinate system to form its size and position with the shape and pose parameters. With the known photo exterior and interior orientation parameters the primitive model and image features can be transformed into the photo coordinate system. The fitting is evaluated based on the correspondence between the projected model wire frame and the edge pixels. The least-squares fitting algorithm is applied to solve the optimal fitting and estimate the final model parameters. Ten houses composed of box and gable­roof primitives are tested for demonstration. The experimental results are compared with the data of manual measurements.

There exist some errors of the retrieved surface reflectance by using an atmospheric correction model based on Lambertian surface assumption, since the natural targets are not Lambertian. In this study, 6S is used to estimate the errors. Plowed field is select and a linear BRDF model is used. Wavelength at 0.63ìm is considered. Errors varied with solar zenith angle and aerosol optical depth are studied. The results show that the errors are larger at back-scattering region than fore-scattering region. These errors cannot be neglected, even in clear sky. The error can be up to 0.13 at view angle -60°, when solar zenith angle is 60° and the sky is hazy. When the solar zenith angle and/or sky changes from clear to hazy, the errors generally increase.