2D Geometric Correction of IKONOS Imagery Using Genetic Algorithm

The number of high resolution space imageries, in the civilian market is growing fast.This images have great interest in the photogrammetric and remote sensing communities. The problem with this images for many users, at the present time is the lack of sensor calibration information and precise ephemeris data. Consequently it is not possible to apply the precise mathematical models such as orbital parameter model to determine the dynamic geometry of them at the time of imaging. Therefore for, there is a need for an alternative approach to extract 2D and 3D spatial information from this images. This paper presents a mathematical solution to extract planimetric information as accurate as possible from IKONOS. This method uses Genetic Algorithm technique to find the best terms of global polynomial that provides the best fitness model of the imagery to the ground space. Using the output of global polynomial in multiquadric transformation gives better results in term of accuracy. The result of this method is compared with other 2D mathematical models such as global polynomial, 2D direct linear transformation and multi-quadric models. Also the model is applied to the IRS-1C imagery. IKONOS, Geometric correction, Genetic Algorithm, DLT, Polynomial, multiquadric transformation 1Introduction the new generation of commercial high resolution space imageries such as IKONOS have provided a considerable progress in photogrammetric mapping and map revision. One of the great promises of IKONOS is its potential to generate 3D spatial information products in meter accuracy and to produce large scale topographic maps such as 1:10000 scale and even larger. IKONOS is a pushbroom linear array imaging system, which can take the high-resolution images of the earth using satellite's flexible pointing capability. Accurate geometric correction of IKONOS imageries is an important matter in using these imageries for map production and spatial information extraction. There have been many investigations for 3D spatial information extraction from pushbroom scanners over the past decade. Mathematical models were developed to determine the dynamic geometry of the cross-track stereo images taken by SPOT and IRS-1C/1D by Valadan and Petrie (1998), Radhadevi et al. (1998) and Valadan and Foomani (1999). There were also mathematical models developed to determine the exterior orientation parameters of the along-track stereo images taken by MOMS-02 such as Ebner et al. (1992), Fraser and Shao (1996) and Valadan (1997). In many applications and conditions, 2D geometric correction of satellite imageries is required. Having just a mono image, requesting just a planimetric map, producing fast and accurate geometric correction of image, having no information of camera model and etc. are the reasons to use 2D geometric correction of satellite imagery. At the time of writing this paper, space imaging has refused to release information on the camera model for IKONOS, as well as data on the precise in-flight and attitude of the imaging sensor. Also due to the high expense of stereo images most of users prefer to get the mono images and extract required information after geometric correction. 2D mathematical models seem to be the easiest and the most rapid way for geometric correction of IKONOS imageries. Finding the best method for fast, easy and accurate correction of IKONOS imageries is an important issue in this field. There are several mathematical models for 2D geometric correction of satellite imageries such as global polynomial, piecewise polynomial, pointwise polynomial and multi-quadric transformation,2D projective transformation. Each of these methods has their advantages and disadvantages and is suitable for geometric correction of satellite imageries depending on the topographic specification of the ground and the nature of the features existing on the image. In this paper three new methods for 2D geometric correction of satellite imageries has been introduced and has been tested on IKONOS and IRS-1C images. These methods are construction a combination of global polynomial and Symposium on Geospatial Theory, Processing and Applications, Symposium sur la théorie, les traitements et les applications des données Géospatiales, Ottawa 2002 improved 2D projective, of Global polynomial and genetic algorithm (GA) as well as, Multiquadric and genetic algorithm. 22D mathematical models Some mathematical models has been overviewed as follows: 2-1Global polynomial This model describe the relation between image and object space independent of sensor geometry with the bellow general equation Eq(1),Eq(2): ∑ ∑ = = 0 0 n