Calculation of the Sunbeam Centroid

The next step consists of sending a heliostat from a standby point to the target (which is predefined as an aimpoint, in this case number 9). Figure 6.7 shows zones of dif­ferent images obtained from the CCD camera in which: (a) the heliostat has nearly correct offset coordinates, (b) and (c) the heliostat has wrong offset coordinates. No­tice that the intensity of the shape is much higher than that of the other elements in the image, in such a way that the calculation of the sunbeam centroid is simplified using, for instance, threshold detection techniques, as done in the case of the target. Depending on the selected heliostat, the time elapsed in reaching the target will vary between 18 and 60 s, depending of the kind of heliostat and its position in the field. So, an upper bound of the time the system has to wait till performing the acquisition of a new image is 60 s, this being the first approximation adopted. After several tests, another algorithm was implemented, in such a way that after the first 18 s, each 5 s the center of the centroid was calculated and compared with the previous obtained one (when it appears in the image). When the difference between two con­secutive centers was small (less than a small constant selected to take into account the camera vibration due to the wind), the automatic system considered that the he – liostat was in its final position. This second algorithm worked well when images were obtained around solar midday. Nevertheless, it did not give good results when direct projections of Sun rays impinged on the tower (as happens early in the morn­ing in clear days, e. g. Fig. 6.10(a)), before the reflected shape of the Sun appeared in the image. Due to the automatic brightness/contrast adaptation mechanism of the CCD, once the reflected shape enters in the view field of the camera, the relative intensity of other Sun projections different from the main shape is largely dimin­ished (e. g. those due to the incidence of Sun rays on the side of the tower early in the morning or late in the evening). Another problem is the case of those heliostats which projection is out of the view field of the camera and does not appear in the

captured image. These heliostats are labeled as wrong pointing ones and their off­set can be corrected using a modification of the main algorithm commented later in the subsection. The calculation of the sunbeam centroid was also performed using a threshold detection technique (based on histogram information), as the intensity of the image was near 255. Different more complex image pre-processing algorithms were used, but the added complexity and processing time did not justify their use, because the obtained improvements were not relevant. In order to avoid that other Sun projections different from the main one should lead to wrong results, the pixels with intensities over the threshold are grouped according to their intensity level and the existence of neighbors with the same intensity (segmentation). The figure with largest area is selected as that corresponding to the sunbeam shape and its center is calculated. Regarding this criterion of largest area, in the application shown in this work a situation with most of the radiation falling on surfaces that are not essentially perpendicular to the line of sight of the camera does not appear due to the heliostat field-tower-target layout (this has been verified in the tests). In larger plants other criteria should be used, for instance those related with intensity weighting of the image, requiring more sophisticated hardware (cameras, frame grabbers, etc.) than those used in this subsection. Figure 6.12(a) shows the histogram of the image in Fig. 6.7(b), where the shape of the Sun projected by a heliostat can be observed. An intermediate buffer (or auxiliary file) is used to store the different elements of the image. The steps followed in the calculation of the sunbeam centroid were:

1. Determination of the threshold using the histogram.

2. Bottom-up/left-right scans are performed comparing each pixel with the thresh­old, in such a way that if the grey level of the pixel is below the threshold, it is saved in the auxiliary buffer as a black one (0 intensity). If the intensity of the pixel is greater than the threshold, a fixed intensity value is assigned to it (start­ing at 255 level) and to all the consequent pixels meeting the threshold condition, until a pixel below the threshold is found (and saved as a black one). The fixed intensity value is decremented and the algorithm continues looking for pixels fulfilling the threshold condition till finding the end of the row. This last pro­cessed row is stored into memory, in such a way that the procedure is repeated for the following row, not only comparing the intensities of the pixels with the threshold, but also with the intensities of the pixels belonging to the previous

Fig. 6.13 Approximated calculation of the vertical displacement [45]

processed row, in order to group and label with the same intensity level all the adjacent neighbors (backward modification of the assigned intensity is required in the current processing line to assign the same grey level to pre-processed ones belonging to the same object in the figure). Figure 6.12(b) shows the result (nega­tive image) of the application of the algorithm to the image shown in Figs. 6.8(b) and 6.12(c).

3. The completely processed rows are stored in an auxiliary buffer, also saving the number of pixels belonging to the used intensities (histogram), in such a way that when the process finishes, the grouped pixels with the predominant intensity should be selected as belonging to the main shape and the corresponding center will be calculated.

4. The intermediate buffer is scanned but only looking for pixels with the selected intensity (belonging to the shape after the segmentation process). The extreme pixels (up, low, left and right) are selected and an approximation to the centroid of the figure is obtained by using the center of the outer box including all the pixels of the shape (Fig. 6.12(b)), using a simple formula as the intersection of the central column ((right + left)/2) and row of the figure ((up + low)/2).

5. As is mentioned in the next subsection, once the centroid has been calculated, this is indicated in the image shown to the operator, in such a way that the decision of correcting the offsets is done by the operator, depending of the results given by the algorithm. In the mentioned case, in which the shape lies out of the view field of the camera, the system returns (0, 0) as the sunbeam centroid and the operator is asked about the inclusion of this heliostat in the group of wrong pointing ones, to be corrected with an alternative algorithm.