Quality Evaluation on Object Coordinates Determined by FORMOSAT-2 Stereo Pairs in Polar Area

• Main Authors: Yu-WenCheng, 鄭鈺雯
• Other Authors: Jaan-Rong Tsay
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/78020885973277368474

碩士 === 國立成功大學 === 測量及空間資訊學系碩博士班 === 98 === FORMOSAT-2 satellite developed by Taiwan has high spatial resolution (2m panchromatic images) and high temporal resolution (revisit time interval of one day). In this study, the data process programs developed are adopted to use FORMOSAT-2 images to get polar terrain features and determine the terrain surface changes. By means of analyzing the data quality index including accuracy, reliability and conditional number, the quality of object coordinates determined by FORMOSAT-2 images is examined and analyzed. The inherent problems and limitations to the determination of accurate DSM in polar area are also discussed, and some solutions are given. In the process of simulated frame images in the polar region, if the baseline length of images is increased from 100km to 800km, the horizontal and vertical accuracy is increased about 2m and 35m, respectively. The internal reliability value of horizontal coordinates is decreased 1m and external reliability value is decreased 1, and also, the internal and external reliability values of vertical coordinate are decreased 106 and 54, respectively. They show that the observation errors can be found more easily and they have less effect on the determination of correct function values of unknowns when the baseline length is increased. Compared to the results determined by two images taken on the same orbit, four images taken on two neighboring orbits have better accuracy with the standard deviation decrease of about 31m and 40m on horizontal and vertical coordinates, respectively. In addition, the internal and external reliability values of horizontal coordinates are decreased 3 and 3, and the internal and external reliability values of vertical coordinate are decreased 7 and 4, respectively. In the test areas in polar region, no ground control points are available. Therefore, three stereo models are computed respectively by using stereo pairs taken on three days, where 11 and 23 points determined from the stereo pairs taken on first day are selected and used as GCPs and check points in the computations of the stereo pairs on the second and third day. The model on the second day has the RMSD on X, Y and Z with ±15.57m, ±8.90m and ±28.18m, respectively. The RMSD on X, Y and Z of the points on the model of the third day is ±46.70m, ±18.9m and ±43.48m, respectively. Because the check points and ground control points in polar areas can not be acquired, the DSMs determined by real FORMOSAT-2 images are compared by means of relative registration of any two DSMs. Although the real volume can not be measured in the polar area, the fast-change in the terrain suface of polar region can still be detected by registration of any two DSMs determined by FORMOSAT-2 images taken on different days.