# 8.2 Black Spot module testing (2D testing)

111

comparable. These large data sets show how the centroids vary with the changes that can be found in an office environment during a typical day.

It should be noted how drift could affect the performance of the system. First, assume that the system is calibrated only at start up and therefore any drift throughout the day would degrade the performance of the system. large drift in the measured centroid positions from calibration would lead to a large degradation in performance. Therefore it is important that any drift measured is cyclic and that the maximum magnitude of the drift

is small.

Each data set was captured using a shutter period of 300 µs with a 2 m separation between camera and beacon. In each of the three data sets, a marker near the centre of the FOV was used, as these provide the best performance as discussed in Section 8.2.3. The minimum standard deviations of each of the centroids over the entire data set was also calculated in X and Y. Results are shown in Table 8.2. The lowest standard deviation over the entire data set was from Module 4^{1 }and this shows that the centre marker has a standard deviation of 0.0141 pixels in X and 0.0234 pixels in Y. The standard deviations in pixels can be converted to metres as the focal length of the lens and distance between the LEDs and the camera module are known. The lowest standard deviation corresponds to 28.2 µm in X and 46.8 µm in Y at 2 m. The worst result was from Module 2 and shows a standard deviation of 0.0507 pixels in X and 0.0842 pixels in Y corresponding to 101.5 µm in X and 168.4 µm in Y. These values indicate that even with only one calibration period at the beginning of a 7 hour shift (and assuming the data set is representative), the 2D centroid variation is low.

In the future, the intention is that the system will continuously recalibrate itself and it is assumed that any drift can be accounted during calibration. ssuming that the system re-calibrates every 2 minutes. Using the data sets available this corresponds to 100 sam- ples of data. For each sample, the standard deviation of the 100 data points surrounding this sample was calculated. This produced a matrix of standard deviations. The columns represent the centroid number and the rows represent time. Therefore, the element at the m^{th }row and n^{th }column represents the standard deviation of the centroids from the n^{th }marker. This standard deviation is calculated using the 100 data points surrounding the

m^{th }

instant. This was done for each data module.

The graph in Figure 8.10 shows an example of this data showing the standard deviation as it varies over time for Module 4. The values shown are the mean of the standard deviations of the markers and are calculated by taking the mean of each row of the matrix of data. The standard deviations shown are of the centroids’ Y coordinates. The graph shows that the mean standard deviation is generally small but there are a number of large spikes in the

^{1}Modules are numbered corresponding to their bus ID. Hence, they are numbered 2, 3, and 4.