Feature-based next-best-view selection for scanning 3D objects

This thesis presents a method to observe a 3D object’s surface that needs to be scanned and find the next positions for a sensor or camera with a minimum number of iterations. This process is the Next Best View (NBV). We propose a method that uses a point cloud density measurement technique to determine the NBV positions. The point cloud density technique assesses the point cloud density which is determined from point cloud acquired from scanning the object's surface. The point cloud obtained from the scanned surface has different densities on each part of the object. The point cloud density measurement is used as an indicator to determine the completion of the scanning process. Our proposed algorithm verifies that the acquired point cloud is good and covers every part of the object whenever the quality measurement is satisfied and meets criteria. We tested our method in a simulation environment. Our proposed algorithm gives a lower number of iterations compared to the state-of-the-art methods that are used as the baseline algorithm.