GeoEtrim, SharpQ and Epix: Trio of Tools for Geospatial Image Analysis
Keywords: GeoEtrim, GeoSpot, GeoFigcon, SharpQ, epix, Geospatial analysis
Abstract. This paper presents an overview of three approaches developed in Matlab for geospatial analysis of images. GeoEtrim has two subpackages, GeoSpot and GeoFigcon. GeoSpot aims to perform bundle adjustment of stereo linear array remotely sensed images considering their interior and exterior orientation parameters. The correlation among orientation parameters, their validation and efficiency on the final accuracy can be estimated. The current version is available for SPOT-5 HRG level 1A stereoimages, reaching ±1 pixel accuracy at ICPs (Independent Check Points). GeoFigcon is the other sub-package of GeoEtrim, developed for estimation of georeferencing accuracy of orthoimages generated by various sensor-independent mathematical models and RFM (Rational Function Model). Using GeoFigcon, one can estimate the combined effect of the accuracy of transformation parameters estimated/updated by GCPs and DEM accuracy on the georeferencing accuracy of orthoimages. The experiments with IKONOS Geo, QuickBird OrthoReady Standard, OrbView-3 Basic and Pléiades-1A Primary prove that using RFM produces higher accuracy than using sensor-independent transformation models. Moreover, accuracy varies reducing from geometric centre of GCPs accommodating the high profile of topography. SharpQ generates pan-sharp images using the methods PCA (Principal Component Analysis), Brovey, and IHS (Intensity Hue Saturation), validates their quality with quantitative analysis by the methods CC (Correlation Coefficient), RMSE (Root Mean Square Error), RASE (Relative Average Spectral Error), SAM (Spectral Angle Mapper) and ERGAS (Erreur Relative Globale Adimensionnelle de Synthése). epix, the last member of the trio, can be used for estimation of effective GSD (Ground Sampling Distance) value of original or generated (such as pan-sharp) images, depending the principle of ESF (Edge Spread Function). So the real geometric resolution can be estimated for any kind of image. This trio is being still developed with the continuous research.