Accuracy of frame type hyperspectral image aerotrianguliation

Abstract

First of all, the received images were calibrated. Aerotriangulation (AT) were done with each band of the spectrum. During AT tie points (connection points between the images) were placed in images automatically and the square error calculated. Standart error was estimated for all 58 image bands separately. All image processing and calculations were done by Rikola hyperspectral image software, Rikola Ltd coregistering, EnsoMOSAIC Rikola coregistering applications, Microsoft Excel. The results of the analysis of the processing has been applied and the rollup methods.Best accuracy of aerotriangulation is obtained using the 526 spectral band, standard error is 0.18 pixel, so best accuracy we can achieve if we use 526 spectral band to create hyperspectral orthophoto. More and more hyperspectral images are used in the various areas worldwide. Hyperspectral images are commonly used to monitor various objects. The main objective was to assess the accuracy of the production of aerotriangulation using different spectral bands of the images. The study was conduction in about 1000 ha of Ringaudai village near Kaunas city located in the middle of Lithuania. In the research area there are different types of land cover: forests, agricultural land and urban area. The remotely sensed data were collected on 2016.09 11. Oblique aerial photographs were taken by a RIKOLA hyperspectral camera (58 spectral bands from 502 nm every 4th band to 903 nm with spectral resolution of 8-12 nm) from ultra-light airplane X32 BEKAS flying altitude 1000 m above sea level and with 65% of side and 75% forward overlaps. Resolution of aerial photographs 0.5 m. Coordinates of every frame of image center were collected by GPS integrated in aerophotography system with accuracy ~2m.