Mobile asset tracking for dynamic 3D crane workspace generation in real time

Dynamic workspace modeling is important for providing real-time assistance including hazard detection, collision avoidance and 3D visualization for operators of heavy equipment such as cranes. This information helps the operators to rapidly perceive the surrounding environment and carry out manipulative tasks safely and efficiently. 2D image data can be collected quickly but lacks depth information; on the other hand, point cloud data captures complete 3D information but requires a long period of time to acquire and process the data. To address these shortcomings, this study proposes a hybrid point cloud-vision approach to track the location of mobile assets (e.g., machinery, workers) and to build a dynamic 3D model of the crane workspace. The initial site geometry is captured in a point cloud format using a laser scanner, where mobile assets are represented by 3D oriented bounding boxes. A wide-angle camera mounted on the crane boom continuously monitors the locations of moving objects. The acquired information from the camera and laser scanner is then fused to incrementally update the 3D location of object bounding boxes. The proposed approach is validated in a series of realistic crane lift scenarios in the field.