Fig. 5: Surface reconstruction capabilities and NURBS study. | Nature Communications

Fig. 5: Surface reconstruction capabilities and NURBS study.

From: Automatic heliostat learning for in situ concentrating solar power plant metrology with differentiable ray tracing

Fig. 5

a Heliostat surface reconstructions from focal spot images depending on their count (left column) and the distance to the receiver (right column). The imaged centered at the top shows the ground truth. The visualizations show deviations in mrad of the surface normals from the ideal planar surface. b Quantitative assessment of the reconstruction quality. For distances of up to 100 m, learning NURBS surfaces can be reconstructed accurately with four or more training images. c Surface reconstruction and irradiance quality with respect to the learning NURBS parameter count, polynomial degree, as well as the distance to the receiver. 11 × 11 learning NURBS excel at predicting irradiances, while 7 × 7 NURBS result in more precise surface reconstructions.

Back to article page