Accuracy Assessment of World View-2 Satellite Imagery for Planimetric Maps Production

Abstract

A planimetric map of scale 1:10,000 meets the requirements of a large segment of user’s maps for instance urban city planners and various GIS implementations. Nowadays, a very high-resolution satellite images, such as World View02 (WV02) with spatial resolution of 0.5 m, are very important to produce planimetric maps or update existing ones. Main aim of this research is the assessment of WV02 image for production of the planimetric photomap of scale 1:10,000 with class 1 according to ASPRS standards (ASPRS give accuracy tolerances for map scales at 1:20,000 or larger, this accuracy reported as Class 1, 2, or 3). The investigation includes, studying the best-fit mathematical model (order of polynomial transformation model) that can be used to perform geometric correction for the used image. As well as, examine the effect of the ground control points (GCPs) configuration on the accuracy that can reached from photomap by using the best polynomial order. The Root Mean Square Error (RMSE) resulting at the checkpoints (CPs) will be evaluated. Before the study of impact of the mentioned effects, will be studying the possibility of obtaining a photomap with scale of 1:10,000 and determining the class of this map by using raster satellite image directly (raw image). Through it will compare the coordinates of GCPs observed by using Differential Global Positioning System (DGPS) on the raster WV02 satellite image with respect to its true position on the ground. Taking into consideration this comparison will be conducted according to international standards (National Standard for Spatial Data Accuracy (NSSDA) and American Society for Photogrammetry and Remote Sensing (ASPRS) standards). Evidenced by the results that have been accessible, it cannot obtained on photomap with scale of 1:10,000 of class1 according to ASPRS standard from raw WV02 satellite image, because the RMSE was 4.709 m, this value is largest of allowable error value for this class of the scale. Further, the extracted results showed that using a 1st order polynomial for WV02 image correction with the 14 GCPs that well distributed is slightly superior to other order polynomials (2nd and 3rd order) with a total RMSE of 0.790 m at the 8 ground CPs. On the other hand, using 13 GCPs well distributed (covers the wholly raster of the used image) for the correction process with the same polynomial order, the total RMSE obtained is 0.894 m obtained at 9 CPs, which is less than the value of two pixel size (user-threshold value) of WV02 image. As well as, according to NSSDA and ASPRS standards, this result satisfies the requirements of large-scale maps production accuracy (larger than 1:10,000). In addition, by decreasing the number of the GCPs (using 9 GCPs until 4) the reliability of the results decreases (i.e., the horizontal error increased, approximately 1.4 m are obtained at CPs), but at the same time can get a photomap within scale of 1:10,000.