This content is not included in your SAE MOBILUS subscription, or you are not logged in.
SRGAN-TQT, an Improved Motion Tracking Technique for UAVs with Super-Resolution Generative Adversarial Network (SRGAN) and Temporal Quad-Tree (TQT)
ISSN: 0148-7191, e-ISSN: 2688-3627
Published May 26, 2022 by SAE International in United States
Annotation ability available
Event: AeroCON 2022
Unmanned Aerial Vehicles (UAVs) are gaining significant popularity due to wide-scale applications in civilian and military use. Unmanned Aerial Vehicles are most commonly used for surveillance. Object tracking is one of the most important things that an autonomous UAV has to perform. However, the accuracy of the object tracking model degrades when the object fades away to some distance or if the input images have low resolution. High-resolution cameras are expensive and increase the overall cost of the UAV. The concept of SRGAN-TQT (Super-Resolution Generative Adversarial Network - Temporal Quad-Tree), an improved object tracking pipeline for UAVs in the presence of low-resolution cameras or distant objects, provides a cost-effective solution with enhanced accuracy to perform object tracking. Implementation of Super-Resolution - Generative Adversarial Networks (SRGANs) and Temporal Quad-Tree (TQT) along with state-of-the-art object detection algorithms serve as the backend of the pipeline. This approach uses Deep Neural Network-based GANs to upsample images precisely. Temporal Quad-Tree (TQT) is a motion tracking technique that is an extension of the popular Quad-Tree segmentation algorithm. The Temporal Quad-Tree algorithm is present to reduce the computational complexity and give a highly reliable tracking algorithm. This consequently omits the requirement of a high-resolution camera for UAVs while increasing the object tracking capability.
CitationMore, D., Acharya, S., and Aryan, S., "SRGAN-TQT, an Improved Motion Tracking Technique for UAVs with Super-Resolution Generative Adversarial Network (SRGAN) and Temporal Quad-Tree (TQT)," SAE Technical Paper 2022-26-0021, 2022, https://doi.org/10.4271/2022-26-0021.
- Ledig , C. , Theis , L. , Huszar , F. , Caballero , J. et al. 2017 10.1109/CVPR.2017.19
- Jilani , B.A. , Rabie , T. , and Baziyad , M. Autonomous Motion Tracking for Dynamic Objects Using a Temporal Quad-Tree Algorithm Advances in Science and Engineering Technology International Conferences (ASET) 2019 1 5 10.1109/ICASET.2019.8714279
- Ranade , S. , Rosenfeld , A. , and Prewitt , J. 1980 IEEE Transactions on Reliability - TR.
- Muhsin , Z.F. , Rehman , A. , Altameem , A. , Saba , T. et al. Improved Quadtree Image Segmentation Approach to Region Information Imaging Science Journal 62 2014 56 62 10.1179/1743131X12Y.0000000063
- Bradski , G.R. 1998 10.1.1.14.7673
- Fukunaga , K. and Hostetler , L.D. The Estimation of the Gradient of a Density Function, with Applications in Pattern Recognition IEEE Trans. Inf. Theory 21 1975 32 40 10.1109/TIT.1975.1055330
- Simonyan , K. and Zisserman , A. 2014
- He , K. , Zhang , X. , Ren , S. , and Sun , J. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification IEEE International Conference on Computer Vision (ICCV) 2015 , 1026 1034
- He , K. , Zhang , X. , Ren , S. , and Sun , J. 2016 770 778 10.1109/CVPR.2016.90
- Maas , A.L. , Hannun , A.Y. , and Ng , A.Y. Rectifier Nonlinearities Improve Neural Network Acoustic Models Proceedings ICML 2013 - 30th International Conference on Machine Learning (ICML) Atlanta, Georgia , June 16- 21, 2013 10.1.1.693.1422
- Zhu , P. , Wen , L. , Du , D. , Bian , X. et al. Detection and Tracking Meet Drones Challenge IEEE Transactions on Pattern Analysis and Machine Intelligence 10.1109/TPAMI.2021.3119563
- Frazier , G. 1993 446 452 10.1145/170791.170896
- Coll , B. and Morel , J.-M. Non-Local Means Denoising Image Processing On Line. 2011 1 10.5201/ipol.2011.bcm_nlm