Novel View Synthesis with Neural Radiance Fields for Industrial Robot Applications

Markus Hillemann,Robert Langendörfer,Max Heiken,Max Mehltretter,Andreas Schenk,Martin Weinmann,Stefan Hinz,Christian Heipke,Markus Ulrich
2024-05-07
Abstract:Neural Radiance Fields (NeRFs) have become a rapidly growing research field with the potential to revolutionize typical photogrammetric workflows, such as those used for 3D scene reconstruction. As input, NeRFs require multi-view images with corresponding camera poses as well as the interior orientation. In the typical NeRF workflow, the camera poses and the interior orientation are estimated in advance with Structure from Motion (SfM). But the quality of the resulting novel views, which depends on different parameters such as the number and distribution of available images, as well as the accuracy of the related camera poses and interior orientation, is difficult to predict. In addition, SfM is a time-consuming pre-processing step, and its quality strongly depends on the image content. Furthermore, the undefined scaling factor of SfM hinders subsequent steps in which metric information is required. In this paper, we evaluate the potential of NeRFs for industrial robot applications. We propose an alternative to SfM pre-processing: we capture the input images with a calibrated camera that is attached to the end effector of an industrial robot and determine accurate camera poses with metric scale based on the robot kinematics. We then investigate the quality of the novel views by comparing them to ground truth, and by computing an internal quality measure based on ensemble methods. For evaluation purposes, we acquire multiple datasets that pose challenges for reconstruction typical of industrial applications, like reflective objects, poor texture, and fine structures. We show that the robot-based pose determination reaches similar accuracy as SfM in non-demanding cases, while having clear advantages in more challenging scenarios. Finally, we present first results of applying the ensemble method to estimate the quality of the synthetic novel view in the absence of a ground truth.
Computer Vision and Pattern Recognition,Artificial Intelligence,Robotics
What problem does this paper attempt to address?
The paper primarily aims to address several key issues of Neural Radiance Fields (NeRF) in industrial robot applications: 1. **Improvement of Camera Pose Estimation**: The paper proposes an alternative Structure from Motion (SfM) method to obtain more accurate camera poses. By mounting the camera on the end effector of an industrial robot and using the robot's kinematic information, accurate camera poses with metric scale are determined. 2. **Quality Assessment of Novel View Synthesis**: The study evaluates the quality of synthesized novel views in different scenarios (such as reflective objects, poor textures, and fine structures) and compares them with real images using metrics like PSNR and SSIM. 3. **Uncertainty Estimation**: An ensemble-based internal quality metric is proposed to predict the quality of synthesized novel views even in the absence of ground truth. Through these studies, the paper aims to demonstrate the advantages of the robot-based pose determination method in challenging situations compared to traditional SfM methods, and to explore the potential of NeRF methods in industrial applications and the importance of uncertainty estimation.