The Rectification of High Resolution Remote Sensing Satellite Imagery
In a flat terrain, geometric correction can be implemented by transformation of planimetric control points. While in a hilly area, relief displacement of terrain features are significant and need to be compensated. A 3-dimensional transformation should be introduced and the ortho-rectification procedure is adopted.
2.2 Image Resampling
In image resampling, in general, 3 different approaches can be applied, namely: the nearest neighbourhood, the bilinear interpolation and the cubic convolution. They possess their own characteristics but mainly vary in computation time and fidelity. For those who needs more information about resample techniques can refer to interpolation textbooks.
3. Data Source
3.1 Satellite Image
The image used in this study is the IKONOS satellite image. IKONOS was launched successfully on 24 September 1999, from California, USA. The orbit height is 681km above sea level, travel at a speed of 7 km/sec. The revision period ranges from 1 to 3 days depending on the imaging angle. The optical scanner provides image swath of 11km. The sensor system allows simultaneously collection of 1 meter resolution black and white (panchromatic) images and 4 meter resolution color (multispectral) images. The sun-synchronous orbit drive the satellite flies over 24°N latitude at 10:00 a.m. everyday, which allows the image shadows falling onto about the same direction of north-west, and hence improves the efficiency of visual interpretation.
Two IKONOS scenes were used in this paper. The first scene is the forbidden city, Beijin. The imaging date was 10/21/1999 and the data volume is 2006 * 2006 pixels. The second scene is Yuan-shan, Taipei. The imaging date was 10/21/1999, and the data volume is 2101 * 2101 pixels. Both scenes were downloaded from the Internet provided by the Space Imaging Co. Thess image have been radiometrically adjusted to improve the radiometric quality by the producer before uploading to the web site. From a general user point of view, the radiometry can be of little improvement, whilst the geometric accuracy is of concerned as long as the control materials are available. For the second scene, the image frame covers part of northern Taipei area, namely Yuan-shan. The geomorphology in this area is mostly flat terrain, while a small portion of hilly mountain located at the right middle with elevation of 110 meter above sea. Tan-shui river flows through at left of the image and Kee-lung river flows from right middle to north. Highway, metropolitan railway, football field and one of the landmark of Taipei, the Grant Hotel, are clearly visible in this image. Due to the image is an Internet sample, the orbit parameters such as satellite position of exposure, geometry of scanner, tilt angle as well as sun angle are unknown. This makes it impossible to apply orbit parameters in precisely determining the geometry of imaging.
3.2 Control Data
The control data used in this research were derived from two different sources. In the Forbidden City case, due to the lack of ground truth, the control data were read from a 1/15,000 tourism map. It is clear that this map's accuracy is not sufficient for rectification purpose of the IKONOS image. But the geometric condition of the rectangularity for the shape of the Forbidden City was evidently a major constrain from visual inspection of this tourism map. In the Yuan-shan case, a digital 1/1,000 topographic map produced by the Bureau of Urban Development of Taipei City Government was used. It provides approximately 30cm planimetric accuracy and 50cm horizontal accuracy. Compare with the ground resolution of 1 meter of IKONOS image, this digital map provides sufficient control data.
4. Experiences and Results
4.1 Case 1
Case 1 refers to the relative geometric rectification of the Forbidden City image. As previously stated, the control points were derived from a tourism map. This map was scanned and converted to a digital form. By superimposing a digital regular grid net on top of the scanned map, relative control points of the test area were derived. In particularly, the right angle corners of the Forbidden City Wall were abstracted and recorded.
Figure 1 shows the original image. It is evident that the corners of the Wall are in skew due to the satellite view angle at the imaging instant. Figure 2 presents the relatively geometric corrected image in which the rectangularity of the Wall corners were maintained.
