Process- and Computer Vision-based Detection of As-Built Components on Construction Sites

Automated progress monitoring becomes more and more common in the construction industry. Recent approaches often use new methods like Unmanned aerial vehicles (UAVs) for a capturing large construction sites. However, the used methods often lack accuracy due to occluded elements and/or reconstruction inaccuracies from using photogrammetric methods. This paper presents a novel approach for further improvement of element detection rates. 4D BIM semantic information is used, to generate precise "as-planned" geo- metric models. These models are used to render a building from all points of view during the monitoring phase. Based on this information, a more accurate and reliable estimation of all detected elements can be achieved.