The shape of an installed solar concentrator (e.g. a heliostat) may differ from its original design due to manufacturing defects, structural/wind loads and thermal expansion. By measuring the shape of a solar concentrator, it is possible to account for the deviation in optical performance from its original design point. A method to measure concentrator shape needs to be fast, accurate, and not involve contact or interference with the reflective surface of the concentrator. State-of-the-art techniques include flux mapping, photogrammetry, and deflectometry using conventional cameras.
This paper presents a study to characterise solar concentrator shapes using light-field imaging. Conventional cameras capture the light intensity of a point in a scene at a single point in the sensor, creating a two-dimensional image. A light field camera features multiple micro-lenses placed between the main lens and the sensor, providing many small images from slightly different angles in a single shot. This information is used to reconstruct the position of a light source in space. The advantage of this new technique to the ones mentioned above is that the light field camera is robust and self-contained, which allows easy-to-use application in heliostat fields.
In this study, light-field camera measurements were performed with flat mirrors and a curved mirror under laboratory conditions. In order to resolve the surface of the mirror surfaces, several methods to impose contours of the mirror surface have been studied, including dirt, small water droplets, scattering of low-power laser light, and paper-marks. A wide range of camera-to-mirror distances between 43 cm and 5 m have been studied. Greater distances allow the capture of the entire surface, but decrease the precision of depth measurements. In order to obtain high precision measurements while being able to capture the entire surface, a compositing strategy has been developed, combining several light-field image measurements. The overall accuracy of the system was improved further by averaging measurements over several image frames. Subsequently, the reconstructed surface points have been fed to a ray-tracing algorithm realized in Matlab/Python.
Results in this study are able to resolve the shape of small concentrators to sub-millimetre precision when taking pictures at a distance of 0.4 m.