Page 72 - RAC_CIAW_ a_I_n_01_2021.pdf

Fig. 3. Example of radar screen used in the simulations

the PPI screen until the algorithm finds
a match or until all possibilities along
the screen are tested. During the search
process, the test image (i.e., the PPI
screen) is read and converted to gray-
scale. This serves to eliminate possible
color variations, performing only the
analysis of the pixel intensity. The tem-
plate is also processed in grayscale for
the comparison. To implement the al-
gorithm, the libraries Numpy, cv2 and
Pillow were used. The implementation
is shown in Figure 4.
Five threshold levels were as-
sessed: 0.3, 0.4, 0.5, 0.6, 0.7. Recall
that the computed PCCs are compared
with the threshold levels in order to de-
cide if a match was found or not (see Fig. 4 Triggering mechanism implementation in Python
Section III). Each threshold level was
assessed using the set of 30 different scenarios. The PPI and there is a match with the template. False-Pos-
values corresponding to the confusion matrix for each itive (FP) is the case in which the triggering command
threshold level are compiled in Table 1. The perfor- is not present in the PPI, but there is a match with the
mance of the triggering mechanism for each threshold template. True- Negative (TN) occurs when the trig-
level is also depicted in Figure 5. gering command is not present in the PPI and there is
The situation of True-Positive (TP) refers to the no match with the template. Finally, the False-Negative
case where the triggering command is present in the (FN) occurs when the triggering command is in the PPI

72 REVISTA ACADÊMICA CIENTÍFICA DO CIAW