Small sample vehicle target recognition using component model for unmanned aerial vehicle

被引：0
作者：
Niu Y. [1 ]
Zhu Y. [1 ]
Li H. [1 ]
Wang C. [1 ]
Wu L. [1 ]
机构：
[1] College of Intelligence Science and Technology, National University of Defense Technology, Changsha
来源：
Guofang Keji Daxue Xuebao/Journal of National University of Defense Technology | 2021年 / 43卷 / 01期
关键词：
Component model; Image segmentation; Small sample learning; Target recognition; Unmanned aerial vehicle;
D O I：
10.11887/j.cn.202101016
中图分类号：
学科分类号：
摘要：
Detecting and recognizing targets on the ground is one of the typical tasks of UAVs (unmanned aerial vehicles), but it is limited by the task particularity so that it is often difficult to obtain sufficient data about target samples to achieve highly reliable target recognition. In view of this problem, a small-sample vehicle target recognition method based on the component model was proposed, which combined the cognitive characteristics of human beings to improve the perception ability of ground targets. The possible region of the target was extracted by visual saliency detection and objectness detection, and then the GrabCut segmentation method based on the Graph theory and the maximum between-class variance was used to segment the target and to extract the components from the target. A component recognition method based on a probability map model was used to perform component recognition by sparsely representing a component outline as a conditional random field and performing probabilistic reasoning. The Bayesian-based target recognition method was used to determine whether the target was a vehicle. Verification on real images captured by the UAV showed that the algorithm can detect and identify the vehicle target with high accuracy under the condition of fewer samples, poorer illumination and certain occlusion. At the same time, the recognition method can achieve the effect of certain interpretability. © 2021, NUDT Press. All right reserved.
引用
收藏
页码：117 / 126
页数：9
相关论文
共 22 条
  • [1] SHEN Lincheng, NIU Yifeng, ZHU Huayong, Theories and methods of autonomous cooperative control for multiple UAVs, (2018)
  • [2] ZHU Yuting, Mixed-initiative recognition and tracking of ground vehicles using UAV images, (2018)
  • [3] GIRSHICK R, DONAHUE J, DARRELL T, Et al., Rich feature hierarchies for accurate object detection and semantic segmentation, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 580-587, (2014)
  • [4] REN S Q, HE K M, GIRSHICK R, Et al., Faster R-CNN: towards real-time object detection with region proposal networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, 39, 6, pp. 1137-1149, (2017)
  • [5] REDMON J, DIVVALA S, GIRSHICK R, Et al., You only look once: unified, real-time object detection, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 779-788, (2016)
  • [6] LIU W, ANGUELOV D, ERHAN D, Et al., SSD: single shot multibox detector, Proceedings of the European Conference on Computer Vision, pp. 21-37, (2016)
  • [7] CAI Z W, VASCONCELOS N., Cascade R-CNN: delving into high quality object detection, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6154-6162, (2018)
  • [8] ZHANG S F, WEN L Y, BIAN X, Et al., Single-shot refinement neural network for object detection, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4203-4212, (2018)
  • [9] SINGH B, DAVIS L S., An analysis of scale invariance in object detection-SNIP, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3578-3587, (2018)
  • [10] GEORGE D, LEHRACH W, KANSKY K, Et al., A generative vision model that trains with high data efficiency and breaks text-based CAPTCHAs, Science, 358, 6368, (2017)
123