Fig 1.
Proposed target recognition framework.
Fig 2.
Off-board recognition system.
Fig 3.
On-board/real-time recognition system.
Table 1.
CNN configurations.
Fig 4.
A and B represent waypoints while the search area was 40m*40m.
Fig 5.
Simulation environment, UAV moving from one-way point to another while searching for the target a) UAV taking off b) searching target c) target recognized.
Fig 6.
UAV used in the experiment.
Fig 7.
Comparing the supervised learning classifiers’ results on five different evaluation sets a) F1 score for training sets b) F1 score for testing sets.
Fig 8.
Recognition with a confidence score.
Table 2.
Average training results of classifiers for the five evaluation tests.
Table 3.
Average testing results of classifiers for the five evaluation tests.
Fig 9.
Average F1 score and processing time for all configurations.
Fig 10.
MSM for comparing two sets of images [31].
Fig 11.
Typical LeNet-5 architecture [35].
Fig 12.
VGG-16 model for bird species classification [36].
Table 4.
Average testing results of classifiers for the 5 evaluation tests.
Fig 13.
Average F1 score and processing time for comparison.