Précédent|Prochain

Object detection in remote sensing images using densely connected recursive feature pyramids

  • role： First author第一作者
  • Affiliation：

    College of Combat Support, Rocket Force University of Engineering, Xi'an 710025, China

  • Email：504998692@qq.com
  • Introduction：吕奕龙，研究方向为遥感目标识别与检测。E-mail： 504998692@qq.com
LYU Yilong,  
  • role： Corresponding author通信作者
  • Affiliation：

    College of Combat Support, Rocket Force University of Engineering, Xi'an 710025, China

  • Email：proflimin@163.com
  • Introduction：李敏，研究方向为人工智能与模式识别。E-mail： proflimin@163.com
LI Min*,  
  • Affiliation：

    College of Combat Support, Rocket Force University of Engineering, Xi'an 710025, China

WU Zhaoqing,  
  • Affiliation：

    College of Combat Support, Rocket Force University of Engineering, Xi'an 710025, China

HE Yujie

résumé

In recent years, the multiscale utilization of input sample features has gradually become a research hotspot in the field of target detection. However, remote sensing target detection suffers from some problems, such as small target size, easy confusion with similar objects, and extensive background interference. Therefore, a remote sensing target detection algorithm based on dense connection recursive feature pyramids is proposed. First, the feature fusion mode is improved to use the features of remote sensing images fully. The traditional feature fusion method is only pixel-by-pixel addition, which is simple and rough to calculate and cannot effectively screen features. Therefore, canonical correlation analysis is used to replace the simple pixel-by-pixel additive fusion mode to enhance the effectiveness of feature fusion. Moreover, this method does not add any new parameters. Second, the multireceptive field (MRF) mechanism was added to enhance the feature extraction of small-scale targets, and the features of different receptive fields were extracted and fused by dilated convolution of different sizes to enhance network perception. Given the increase in receptive field types, the richness of features that can be extracted is greatly enhanced, which is conducive to the improved transmission of effective information. In addition, our proposed MRF module is a multibranch convolution module, which is intended to mimic the human visual receptive field mechanism. Then, the feature recurrence form is improved to solve the generalization problem of a multiscale remote sensing target, and the dense connection structure is introduced to enhance the feature fusion density. A dense connection improves network performance because the feature level increases, and the feature richness is enhanced accordingly. Compared with the original recursive feature pyramid, the utilization of the backbone network is remarkably improved. The backbone network and the feature information of high and low layers are fully utilized. Finally,Based on our proposed methods above, this study changes the way of feature recursion and designs a dense connection structure of the recursive feature pyramid. That is, it adds dense connections between multiscale features and each layer of the backbone network to improve the efficiency of feature extraction and utilization. In summary, the network design in this study includes a top-down fusion subnetwork, bottom-up path enhancement subnetwork, and feature recursive fusion subnetwork. Experimental results show that the average accuracy of the proposed pyramid model can be improved by 9.9% on the general dataset MS-COCO2017. On the remote sensing dataset NWPU VHR, the average accuracy of the proposed algorithm can be improved by 1.1%. On the remote sensing dataset DIOR, the average accuracy of the proposed algorithm can be increased by 2.2%, which is higher than other feature pyramid models and detection algorithms. On the large-scale remote sensing dataset DOTA, the average accuracy of the proposed algorithm can be increased by 1.8%. Experimental results show that the proposed method can outperform other feature pyramid models and detection algorithms. It achieves not only high precision detection of remote sensing targets but also has good performance on the benchmark dataset COCO. Therefore, the proposed method is advanced.

mots-clés

remote sensing image;object detection;feature pyramid network;feature recursive;densely connected

References

  1. 1.
    Cai Z W and Vasconcelos N. 2018. Cascade R-CNN: delving into high quality object detection//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE: 6154-6162
  2. 2.
    Chen K, Pang J M, Wang J Q, Xiong Y, Li X X, Sun S Y, Feng W S, Liu Z W, Shi Z P, Ouyang W L, Loy C C and Lin D H. 2019. Hybrid task cascade for instance segmentation//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE: 4969-4978
  3. 3.
    Chen K, Wang J Q, Pang J M, Cao Y H, Xiong Y, Li X X, Sun S Y, Feng W S, Liu Z W, Xu Z R, Zhang Z, Cheng Z D, Zhu C C, Cheng T H, Zhao Q J, Li B Y, Lu X, Zhu R, Wu Y, Dai J F, Wang J D, Shi J P, Ouyang W L, Loy C C and Lin D H. 2019. MMDetection: open mmlab detection toolbox and benchmark. arXiv: 1906.07155
  4. 4.
    Cheng G, Han J W and Lu X Q. 2017. Remote sensing image scene classification: benchmark and state of the art. Proceedings of the IEEE, 105(10): 1865-1883
  5. 5.
    Ghiasi G, Lin T Y and Le Q V. 2019. NAS-FPN: learning scalable feature pyramid architecture for object detection//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE: 7029-7038
  6. 6.
    Girshick R, Donahue J, Darrell T and Malik J. 2014. Rich feature hierarchies for accurate object detection and semantic segmentation//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE: 580-587
  7. 7.
    Han Y S, Ma S P, He L Y, Li C H, Zhu M M and Xu Y L. 2021. Regional object detection of remote sensing airport based on improved deep neural network. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(7): 1470-1480
  8. 8.
    Lin T Y, Dollár P, Girshick R, He K M, Hariharan B and Belongie S. 2017a. Feature pyramid networks for object detection//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE: 936-944
  9. 9.
    Lin T Y, Goyal P, Girshick R, He K M and Dollár P. 2017b. Focal loss for dense object detection//Proceedings of 2017 IEEE International Conference on Computer Vision. Venice: IEEE: 2999-3007
  10. 10.
    Lin T Y, Maire M, Belongie S, Hays J, Perona P, Ramanan D, Dollár P and Zitnick C L. 2014. Microsoft COCO: common objects in context//Proceedings of the 13th European Conference on Computer Vision. Zurich: Springer: 740-755
  11. 11.
    Liu S, Qi L, Qin H F, Shi J P and Jia J Y. 2018. Path aggregation network for instance segmentation//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE: 8759-8768
  12. 12.
    Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu C Y and Berg A C. 2016. SSD: single shot multibox detector//Proceedings of the 14th European Conference on Computer Vision. Amsterdam: Springer: 21-37
  13. 13.
    Qiao S Y, Chen L C and Yuille A. 2021. DetectoRS: detecting objects with recursive feature pyramid and switchable atrous convolution//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE: 10208-10219
  14. 14.
    Radosavovic I, Kosaraju R P, Girshick R, He K M and Dollár P. 2020. Designing network design spaces//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE: 10425-10433
  15. 15.
    Redmon J, Divvala S, Girshick R and Farhadi A. 2016. You only look once: unified, real-time object detection//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE: 779-788
  16. 16.
    Ren S Q, He K M, Girshick R and Sun J. 2015. Faster R-CNN: towards real-time object detection with region proposal networks//Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal: MIT Press: 91-99
  17. 17.
    Sun Q S, Zeng S G, Liu Y, Heng P A and Xia D S. 2005. A new method of feature fusion and its application in image recognition. Pattern Recognition, 38(12): 2437-2448
  18. 18.
    Tan M X, Pang R M and Le Q V. 2020. EfficientDet: scalable and efficient object detection//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE: 10778-10787
  19. 19.
    Tian Z, Shen C H, Chen H and He T. 2019. Fcos: fully convolutional one-stage object detection//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE: 9626-9635
  20. 20.
    Yao Q L, Hu X and Lei H. 2019. Object detection in remote sensing images using multiscale convolutional neural networks. Acta Optica Sinica, 39(11): 1128002
  21. 21.
    Yao Y Q, Cheng G, Xie X X and Han J W. 2021. Optical remote sensing image object detection based on multi-resolution feature fusion. Journal of Remote Sensing (in Chinese), 25(5): 1124-1137
  22. 22.
    Yu F and Koltun V. 2016. Multi-scale context aggregation by dilated convolutions//Proceedings of the 4th International Conference on Learning Representations. San Juan: ICLR
  23. 23.
    Zhang H K, Chang H, Ma B P, Wang N Y and Chen X L. 2020. Dynamic R-CNN: towards high quality object detection via dynamic training//Proceedings of the 16th European Conference on Computer Vision. Glasgow: Springer: 260-275

Lire l'article complet

The above content is generated by Large Model Translation. The translated content is for reference only. We do not assume any commercial or legal responsibilty for any consequences arising from the use of our website