[1]LIN Shubin,WU Guishan,YAO Wenyong,et al.Unmanned aerial vehicles object tracking based on illumination adaptive dynamic consistency[J].CAAI Transactions on Intelligent Systems,2022,17(6):1093-1103.[doi:10.11992/tis.202110023]
Copy
CAAI Transactions on Intelligent Systems[ISSN 1673-4785/CN 23-1538/TP] Volume:
17
Number of periods:
2022 6
Page number:
1093-1103
Column:
学术论文—机器学习
Public date:
2022-11-05
- Title:
-
Unmanned aerial vehicles object tracking based on illumination adaptive dynamic consistency
- Author(s):
-
LIN Shubin1; 2; WU Guishan1; 2; YAO Wenyong3; YANG Wenyuan4
-
1. School of Computer Science, Minnan Normal University, Zhangzhou 363000, China;
2. Key Laboratory of Data Science and Intelligence Application of Fujian Provincial Universities, Minnan Normal University, Zhangzhou 363000, China;
3. School of Foreign Studies, Minnan Normal University, Zhangzhou 363000, China;
4. Fujian Key Laboratory of Granular Computing and Application, Minnan Normal University, Zhangzhou 363000, China
-
- Keywords:
-
computer vision; object tracking; unmanned aerial vehicle; machine learning; correlation filtering; illumination adaptive; dynamic constraint; consistency evaluation
- CLC:
-
TP391.4
- DOI:
-
10.11992/tis.202110023
- Abstract:
-
The tracking of unmanned aerial vehicles (UAVs) is often confronted with illumination change scenes. However, UAV tracking methods mainly achieve robust tracking under sufficient illumination. In this study, a UAV tracking method that uses dynamic consistency evaluation based on illumination adaptability and cross-frame semantic perception is proposed to realize efficient UAV object tracking under insufficient illumination. First, an adaptive illumination module was designed to recognize dim scenes and compensate for the illumination intensity of the dim image. Subsequently, aobject template was built to train a filter with object perception to perform correlation operations; furthermore, the consistency evaluation was performed using the response information between the cross-frames. Finally, a dynamic constraint strategy was developed, and the response difference of the tracker was constrained to maintain temporal smoothness. When the tracking performance of this method was compared to that of nine state-of-the-art algorithms on the UAVDark135 and UAV123 datasets, the results indicated that the algorithm had better tracking performance.