Acta Optica Sinica, Volume. 37, Issue 12, 1215006(2017)
Adaptive Tracking Algorithm for Aerial Small Targets Based on Multi-Domain Convolutional Neural Networks and Autoregression Model
Figures & Tables(8)
Fig. 3. Qualitative comparison of tracking algorithm performance. (a) seq1; (b) seq2; (c) seq3; (d) seq4; (e) seq5; (f) seq6; (g) seq7; (h) seq8
Fig. 4. Comparison of center location error. (a) seq1; (b) seq2; (c) seq3; (d) seq4; (e) seq5; (f) seq6; (g) seq7; (h) seq8
Table 1. Experimental test image sequences
View tableTable 1. Experimental test image sequences
Image sequence True target Track/Speed Main challenge seq1 1 Straight line/Faster Single pseudo target, moving fast seq2 2 Curve/Fast Single pseudo target, background interference seq3 3 Straight line/Fast Multiple pseudo targets rendezvous, background interference seq4 4 Curve/Fast Multiple pseudo targets rendezvous, background interference seq5 5 Curve/Faster Multiple pseudo targets rendezvous, background interference, moving fast seq6 6 Straight line/Fast Multiple pseudo targets rendezvous, similar noise seq7 7 Straight line/Slow Multiple pseudo targets rendezvous, similar noise seq8 8 Curve/Fast Multiple pseudo targets rendezvous, background interference
Table 2. Success rate%
View tableTable 2. Success rate%
Image sequence DLT DLTAR SAMF CNT DSST KCF MDNet Proposed seq1 51.9 32.9 35.6 55.7 55.4 35.6 55.6 100.0 seq2 92.4 71.7 51.5 51.8 49.5 50.5 51.7 99.5 seq3 57.6 28.4 53.5 52.8 53.2 53.5 52.3 95.3 seq4 16.5 7.2 13.9 19.5 13.7 10.5 21.7 96.4 seq5 18.2 8.1 12.7 19.3 12.4 9.8 21.9 95.8 seq6 38.6 16.4 36.4 39.8 36.3 28.7 40.2 98.7 seq7 37.9 16.2 61.2 39.4 35.8 21.1 41.7 99.1 seq8 17.7 8.1 13.2 20.7 11.2 10.3 22.8 96.6 Mean 41.4 23.6 34.8 37.4 33.4 27.5 38.5 97.7
Table 3. Mean center location errorpixel
View tableTable 3. Mean center location errorpixel
Image sequence DLT DLTAR SAMF CNT DSST KCF MDNet Proposed seq1 90.1 98.7 215.9 84.8 88.1 228.3 84.7 1.3 seq2 13.6 14.4 184.8 184.8 186.2 184.9 184.8 0.9 seq3 17.4 16.1 70.5 68.7 70.7 71.5 68.4 4.9 seq4 251.4 257.3 275.2 283.4 258.1 273.5 281.9 40.9 seq5 158.2 149.3 262.4 256.3 272.5 272.5 274.6 46.7 seq6 53.7 48.7 72.8 53.4 45.0 46.7 46.7 1.3 seq7 70.3 68.7 20.2 28.0 26.5 26.0 40.8 1.2 seq8 243.5 233.8 305.2 242.2 317.7 238.4 243.9 40.7 Mean 112.3 110.9 175.9 150.2 158.1 167.7 153.2 17.2
Table 4. Mean overlap rate%
View tableTable 4. Mean overlap rate%
Image sequence DLT DLTAR SAMF CNT DSST KCF MDNet Proposed seq1 34.97 32.28 19.66 34.81 32.38 20.64 35.48 64.63 seq2 61.84 59.26 21.37 31.89 26.00 21.40 32.80 64.12 seq3 56.64 58.25 33.04 30.26 33.49 33.40 31.56 60.08 seq4 5.89 5.14 5.57 5.64 6.01 6.02 7.72 31.56 seq5 6.69 7.02 5.10 5.49 5.91 6.14 7.88 30.47 seq6 6.59 6.81 7.32 9.18 9.18 9.76 15.49 39.18 seq7 6.47 7.85 16.30 17.17 15.76 16.68 15.76 39.21 seq8 6.52 7.24 5.39 5.59 6.01 6.03 8.28 46.29 Mean 23.20 22.98 14.22 17.50 16.84 15.00 19.37 46.94
Tools
Get Citation
Copy Citation Text
Suzhen Lin, Yao Zheng, Xiaofei Lu, Jianchao Zeng. Adaptive Tracking Algorithm for Aerial Small Targets Based on Multi-Domain Convolutional Neural Networks and Autoregression Model[J]. Acta Optica Sinica, 2017, 37(12): 1215006
Paper Information
Category: Machine Vision
Received: May. 27, 2017
Accepted: --
Published Online: Sep. 6, 2018
The Author Email: Lin Suzhen (lsz@nuc.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20