Automatic detection of mound structures in airborne laser scanning data


This paper describes a processing chain for the semi-automatic mapping of grave mounds from airborne laser scanning (ALS) data. In a digital terrain model (DTM) of the ALS ground points, the automatic method slides a template mound over all positions, and assigns confidence scores to anything resembling a mound. The integer scores range from 1 to 6. By using mound templates with gradually increasing radii from 1.0 to 16 m, the method is able to detect all mounds in this range, provided they are visible and well-formed in the DTM. Despite a high number of false positives, the method is a useful tool in semi-automatic, detailed mapping of known grave fields, especially when the number of ground points per square metre is sufficiently large. The method is also able to identify the location of a previously unknown grave field. We discuss possible improvements of the method. The highest potential for better detection performance is in ALS acquisition in the early spring or the late autumn, when leaves are not present.