[M-SD40] Micro-satellite and its constellation for next-generation remote sensing

[MSD40-05] Automated Georeferencing Algorithm for Microsatellite's Multispectral Images

*San Lin Phyo¹, Hline Htet Win - -¹, Ye Min Htay - -¹, Junichi Kurihara - -¹ (1.Hokkaido University)

Keywords:Image Matching, Geometric affine transformation, Georeferencing

Georeferencing means that the internal coordinate system of a map or aerial photo image can be related to a geographic coordinate system. It is crucial to make satellite and aerial as well as terrestrial imagery useful for mapping. Wrong georeferencing leads to inaccurate map projections and measurements that require post-processing correction. The objective of georeferencing is to establish a relationship between image pixel coordinates and a corresponding geographic coordinate system. Microsatellites utilize their on-board sensors for georeferencing but the error in attitude measurements and satellite position remains because of their limited sensor’s performance. Therefore, this research intends to develop an autonomous optimum algorithm which can provide high accuracy for multispectral image georeferencing and automated processing by applying image matching techniques and Google Earth Engine. The objectives of this research are to identify important parameters (brightness, viewing angle, local solar time, etc.) affecting the georeferencing accuracy and to clarify geographical and spectral dependence of the georeferencing accuracy. At first, the data set (such as geolocation, time range, types of bands) of the satellite query image which is required to do georeferencing are obtained. Based on these data set, the suitable base images are searched and obtained using Google Earth Engine. After that, the features matching process between the two images are developed. And then, the false matching points called outliers are rejected and correct matching points known as inliers are collected. Based on these inlier points, the homography matrix which requires for affine transformation is manipulated. The query image is transformed to the base image’s coordinate frame and the pixel coordinates(x,y) of query image are converted to the geographic coordinates(latitude, Longitude) based on the geolocation of base image. Finally, the georeferenced query image is obtained. In this seminar, regarding image matching techniques, the geometric affine transformation and geographic coordinates transformation will be presented and the georeferencing results will be discussed.