Error Propagation in Satellite Multi-Image Geometry

This paper describes an investigation of the source geospatial error in digital surface models (DSMs) constructed from multiple satellite images. In this study uncertainty geometry is separated into two spatial components; global that affects absolute position surface, and local varies point to point. The component caused by inaccuracy imaging process, mainly due orientation (pose) during image collection. A key result a new algorithm for determining geoposition DSM takes account pose covariance each information used weigh evidence computation DSM. use significantly decreases overall position. also approach prediction surface. observed variance within single stereo reconstruction defines horizontal error. fused set elevations pairs at location vertical These accuracy predictions are compared ground truth provided LiDAR scans same geographic region interest.