Monitoring concrete pouring with knowledge graph-enhanced computer vision: a case study from Munich, Germany

This paper introduces a novel approach for monitoring concrete pouring. Traditional manual tracking methods are tedious, while automated solutions, such as Computer Vision (CV)-enabled methods, are challenged with occulted data and limited adaptability to diverse crane behaviour patterns. We propose a knowledge graphenhanced CV method that combines context knowledge with object recognition. This approach analyses tower crane behaviours and their interactions with workers, truck mixers, and building elements, providing a detailed and resilient interpretation of concrete pouring progress. Preliminary findings reveal the method's capacity to interpret incomplete data and comprehend complex site dynamics, demonstrating promising potential in a realworld scenario.