The potential of WorldView-2 potentialities for ortho-image production within the" Control with Remote Sensing Programme" of the European Commission

Imagery acquired by the WorldView-2 (WV2) sensor is of potential interest for the Control with Remote Sensing (CwRS) Programme of the European Commission and therefore needs to be assessed. In details, the horizontal accuracy of the orthoimages which can be derived from WV2 imagery have to be concerned, recalling that in order to qualify WV2 as a Very High Resolution (VHR) prime sensor (i.e. a sensor suitable for measuring parcel areas to the accuracy requested by the Common Agriculture Policy regulation), the CwRS guidelines requires that the one-dimensional Root Mean Square Error (RMSE) (i.e. in the East and North components) estimated on external Check Points (CPs) for any orthoimage should not exceed 2.5 m. This paper summarizes the results regarding the orientation tests of five totally overlapped WorldView-2 Panchromatic (WV2 PAN) images acquired over the Maussane Test Site (Provence, Southern France), two Pan-sharpened (WV2 PANSHP) images (COSE_MODE_1) acquired over Cosenza Test Site (Southern Italy) and nine Pan-sharpened scenes (COSE_MODE_3) again acquired over Cosenza Test Site; these tests have been carried out with Geomatica (PCI Geomatics), ERDAS Imagine 2011 and SISAR software, using both rigorous model and Rational Polynomial Functions (RPFs) model with Rational Polynomial Coefficients (RPCs). The Hold-Out-Validation (HOV) accuracy assessment method was considered, computing the RMSE of the residuals between the estimated and the reference positions of the CPs for each horizontal component (East, North), varying the number of the Ground Control Points (GCPs). In addition the Leave-One-Out Cross Validation (LOOCV) method has been used to identify possible outliers. Considering the obtained results, at first the following conclusions as regards the orientation can be summarized: • the orientation accuracy displays a significant dependence on the off-nadir angle (higher accuracy for lower off-nadir angle) • the best orientation accuracy is reached with the RPFs model using the RPCs supplied with the imagery metadata and applying a shift refinement even with 4 GCPs • the orientation accuracy is practically software independent and appears suited for the orthoimages generation with the one-dimensional RMSE below 2.5 meter in each horizontal component • concerning Cosenza Test Site, it appears by far more convenient to process the two long strips (up to 28 km) (COSE_MODE_1) instead of the corresponding separate nine scenes (COSE_MODE_3), since in the former case the achieved orientation accuracy is practically the same whereas the number of the required GCPs is much lower Secondly, as regards the subsequent orthoimages validation, the already mentioned requirement, that is one-dimensional RMSE below 2.5 meter in each horizontal component, can be reached, provided a good quality DSM is used. In details, for Maussane Test Site all the orthoimages satisfy the previous requirement, whereas sincea DSM with appropriate accuracy was not available for Cosenza Test Site, the results were not considered significantly representative of the real potentiality of WV2 sensor, so that they are not presented. In this last respect, some tests were carried out with the overlapped images of the Maussane Site Test just to evaluate the potentialities of WV2 stereopairs for DSM generation suited for orthoimages production. The results showed that this could be an effective additional resource in case of lack of an external good quality DSM, when proper stereopairs are available.

AASTRAND Paer;  BONGIORNI Marco;  CRESPI M.;  NOWAK DA COSTA Joanna;  FRATARCANGELI F.;  PIERALICE F.;  WALCZYNSKA Agneszka; 

2012-10-19

ELSEVIER SCIENCE BV

JRC67516

0303-2434,   

http://www.sciencedirect.com/science/article/pii/S0303243412001316,    https://publications.jrc.ec.europa.eu/repository/handle/JRC67516,   

10.1016/j.jag.2012.06.003,