Aircraft Recognition and Tracking Device


English \| Ελληνικά
Aircraft Recognition and Tracking Device




Project




Simulation


In order to simulate the system's algorithms, actual data of a real aircraft flight were taken from a real scenario of target attack. The reference position trajectory of the aircraft (reference signal) is known in three-dimensional space (3D), in which the complete simulation scenario in 3D form was designed. Then, a flight path from two (2) different angles of view was received, so two (2) simulation scenarios were created. These angles were selected in such a way that system's performance would be able to be examined under conditions of variable aircraft size (first scenario) and under conditions of variable aircraft speed (second scenario). In both scenarios, it was assumed an enemy aircraft attacking a military airport at six (6) second time window. Moreover, the positions of the original flight path were processed under scale in order that these positions fit in camera's angle of view. Reference Signal Color Code (yellow: past, red: present)

Scenario #1 - Aircraft with variable size

In the first sequence of frames, the distance from one target's position to another is fixed at 20 pixels and its size is increasing after every five (5) positions (with one (1) pixel step per position), in each dimension (0.2 pixels/position). After setting the camera to maximum magnification and having given the coordinates of the target, the following measurements were taken. Figure 1.1 depicts the reference signal.	Fig. 1.1 - Scenario #1: Variable size
The next figure (Fig. 1.2) shows the limits of recognition's algorithm operation. From the first (1st) to the tenth (10th) aircraft's position on the screen, the object is very small, so the algorithm can not identify it. However, recognition is achieved at the eleventh (11th) position, but the system requires three (3) system cycles to identify it as a "valid" target. Thus, after the fourteenth (14th) position onwards the target is successfully identified and verified. The algorithm is tested for the growing size of the target, which at the time of its recognition was 6x6 pixels (minimum recognized size), while by the end of the simulation was varying from 12x12 pixels to 14x14 pixels (maximum recognized size).	Fig. 1.2 - Scenario #1: Reference signal / Recognition algorithm output
As shown in Figure 1.3, the time required for the target to be recognized, is 2.62 seconds (10th position). However, system begins to converge towards the aircraft after 1.33 seconds (in the 3.95th second of the simulation - 14th position); the deviation is 0.334° at X-axis and it falls down to 0.138° just before the simulation ends at the 6.14th second. As for the deviation at Y-axis is 0.026° at the 2.98th second of the simulation.	Fig. 1.3 - Scenario #1: Reference signal / System output for X and Y-axis
Figure 1.4 shows the distance of the target from the implied that passes through the center of the camera. The average error was measured at 0.247479°. Ιt is worth of noting that here is not crucial to locate the aircraft precisely nor to track it with great precision, but to achieve target recognition in the time frame of six (6) seconds within the field of the camera (with the highest precision possible), which was eventually succeeded.	Fig. 1.4 - Scenario #1: Target distance from the center of the system

Scenario #2 - Aircraft with variable speed

Fig. 2.1 - Scenario #2: Variable speed		In the second scenario, a frame sequence was constructed, in which the size of the target was fixed at 29x10 pixels; its speed was set to increase with one (1) pixel step every position. Figure 2.1 shows the reference signal with which the second simulation scenario was carried out.
Fig. 2.2 - Scenario #2: Reference signal / Recognition algorithm output		The operating limits of the recognition algorithm are distinguished when observing Figure 2.2. When the aircraft is moving slowly, from first (1st) to sixteenth (16th) position, the algorithm successfully recognizes it (except for the first three (3) positions at which it performs calculations). Two (2) positions before the end of the simulation, the algorithm is unable to identify the aircraft. That's because, in the seventeenth (17th) position, the target moved with higher speed (about 29 pixels from its previous position) above the speed tolerance of optical algorithm. It should be noticed that the real speed limit (of the aircraft) cannot be defined, because the relevant speed of the system is different from that of the aircraft, as magnification and target distance from observer (system) changes. So by "speed", it is meant the "speed" (in pixels) that target moves on screen. Therefore, the conclusion that was draw is that the maximum "speed", with which the system can move during tracking, is 29 pixels/machine cycle (≈160 milliseconds approximately) (Fig. 2.2).
Fig. 2.3 - Scenario #2: Reference signal / System output for X and Y-axis		As it was seen before (Fig. 2.2), the recognition algorithm works with great precision. The time that is required to recognize the target is 0.71 seconds (Fig. 2.3). After 1.86 seconds (in the 2.57th second of the simulation) the system begins to converge towards the aircraft; the deviation is about 0.154° per axis.
Fig. 2.4 - Scenario #2: Target distance from the center of the system		Having set the magnification to its maximum value, the field of camera view is 2° at X-axis and 1° at Y-axis; bearing that in mind, it is concluded that the deviation is relatively small in proportion to the large magnification. This is illustrated at Figure 2.4. It is obvious that the majority of points are within the middle circles, which indicates a deviation smaller than 0.23°. The average error of target position from the imaginary line passing through the center of the camera (and the system) is 0.197951°.

Conclusion

In conclusion, both scenarios were carried out successfully. So the system is capable of recognizing and tracking a possible incoming threat within the six (6) second time frame.