Open AccessAutomated georeferencing of Diwata-2 multispectral imagery using feature matching
Author(s) -
P Brotoisworo,
R.K.D. Aranas,
MJ Felix
Publication year - 2022
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2214/1/012027
Subject(s) - multispectral image , scale invariant feature transform , remote sensing , artificial intelligence , computer science , computer vision , pixel , feature (linguistics) , orthophoto , mean squared error , matching (statistics) , image resolution , feature extraction , geography , mathematics , linguistics , philosophy , statistics
The Diwata-2 microsatellite is a 56 kg optical microsatellite that features different optical sensors such as the medium resolution Spaceborne Multispectral Sensor (SMI) and a high-resolution sensor called High Precision Telescope (HPT). This research aims to develop an automated georeferencing workflow for coregistered multiband HPT and SMI imagery with a GSD of 5 meters and 127 meters respectively. Georeferencing is done using a mixture of FAST and SIFT feature matching algorithms where HPT imagery used SIFT descriptors and detectors and SMI used FAST detectors and SIFT descriptors. The research used freely available Landsat- 8 imagery which was processed into cloud-free mosaics as reference data. Accuracy assessment is automated using AROSICS software which performs geometric correction and calculates the RMSE of local shifts based on image tie points which are automatically generated using the Diwata image as secondary image and Landsat-8 as reference image. Output of the feature matching algorithms has achieved high accuracy with an average RMSE value of 2.55 meters for HPT imagery and 65.96 meters for SMI imagery. Both RMSE values are close to half the GSD of both sensors which indicate sub-pixel precision. Future research for this method includes additional method for keypoint filtering to further increase the accuracy of the feature matching.