A versatile and efficient data fusion methodology for heterogeneous airborne LiDAR and optical imagery data acquired under unconstrained conditions

Overview & Synopsis

The fusion of airborne LiDAR and aerial/satellite optical imagery datasets allows representing observed scenes in 3-D with a better precision and completeness. Such a fusion has shown strong advantages for large-scale building extraction as well as other remote sensing applications, enabling to overcome the challenges from each data source. However, existing fusion methods do not cope with datasets collected from different platforms with different acquisition configurations, at different moments, and having different spatial resolutions and levels of detail. Such acquisition conditions are referred to as the unconstrained conditions. In addition, building extraction methods reported over the years have achieved relatively significant results by assuming building shapes, enforcing geometrical constraints, or limiting on specific urban areas. Such as sumptions are no longer applicable when dealing with large-scale datasets. This research workis devoted to the development of a versatile coarse-to-fine registration method between datasets acquired in such an unconstrained context. In addition, an efficient building extraction method has been proposed, providing a high accuracy level while being an unsupervised method dedicated to large-scale applications. Relevant results have been achieved upon rigorous assessments of the proposed methods, namely highly desirable accuracy compared to existing methods.