Bundle Adjustment of Multi-station Spherical Panorama Images with GPS Positioning

• Main Authors: Kuan-YingLin, 林冠穎
• Other Authors: Yi-Hsing Tseng
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/66770851909022987867

碩士 === 國立成功大學 === 測量及空間資訊學系 === 102 === In this study, we developed a portable panoramic image mapping system (PPIMS). It is designed for some areas that vehicle-based mapping systems are not allowed to enter, such as rugged terrains, forest areas, and etc. PPIMS is a specially designed platform equipped with eight cameras and a GPS receiver that it can capture eight images simultaneously with e-GPS positioning. When images are taken from multiple stations, a large amount of images are needed to handle and measure. Finding targets among images becomes a puzzled task. This study proposes a new concept of photogrammetry by using panoramic images. It not only breaks the limitation of FOV in traditional photogrammetry but also resolves the problem that handling a lot of images are confusing. Eight images captured with PPIMS can form an spherical panorama image (SPI), which is called PPIMS SPI. The PPIMS SPIs are then used for photogrammetric triangulation and mapping instead of using the original images. Because the collinearity condition is not rigorously kept that causes the gaps on the cutting edges of overlapped images. This study first proposes using SPIs in bundle adjustment to solve the orientation of stations and coordinates of conjugate points in the object space. Then we aim at the deficiency of PPIMS SPI to fix the imperfect geometry of PPIMS SPIs by coordinates of object points solved. And solve the orientation of stations and coordinates of conjugate points in the object space again until the convergence. Two experiments using PPIMS SPIs for bundle adjustment are done, and both their results are validated with check points. One is on the test field in Tzu-Chiang Campus, and another one is in the Eternal Golden Castle (EGC), which has the much complicated landscape. On the test field in Tzu-Chiang Campus, the RMSD values in three directions decreases significantly from (±0.321m, ±0.741m, ±0386m) to (±0.209m, ±0.376m, ±0.270m) after applying corrections for PPIMS SPIs. In EGC project, the RMSD values in three directions become consistent and are about 0.1 m after applying corrections for PPIMS SPIs. Both of them confirm bundle adjustment of SPI is possible, and applying corrections for PPIMS SPIs is necessary and effective for bundle adjustment. We also compare the results using SPIs with original images to understand their real differences. On the test field in Tzu-Chiang Campus, the maximum difference of rotation angles is about 0.03 degrees. Ignoring two unreliable check points, the maximum difference of check point coordinates is about 0.06 meters. In EGC project, the maximum difference of rotation angles is about 0.03 degrees, and all difference of check point coordinates are under 0.04 meters. The above confirms both results using SPIs are consistent with original images, that presents SPIs can replace original images in bundle adjustment.