Welcome to Francis Academic Press

Academic Journal of Computing & Information Science, 2023, 6(4); doi: 10.25236/AJCIS.2023.060409.

The GMIoU Loss for Accurate Rotated Object Detection in Remote Sensing Image

Author(s)

Zhiwei Zhang, Wei Sun

Corresponding Author:
Zhiwei Zhang
Affiliation(s)

College of Information Engineering, Shanghai Maritime University, Shanghai, China

Abstract

How to accurately metric the oriented bounding box loss in remote sensing image object detection has been a hot research topic in this field. The main reason for not being able to accurately calculate the loss of rotating bounding boxes is the complexity of calculating IoU between rotating bounding boxes. This paper is inspired by KFIoU and explores on this basis, and finds that KFIoU needs to be more to metric the merging volume of rotating bounding boxes, and the variation in the value is much different from the actual one. In this paper, we propose a faster and more accurate loss function GMIoU, which calculates the bounding box concatenation volume by Gaussian mixture model and the intersection volume by Kalman filer, and GMIoU solves the problem of inaccuracy of KFIoU in the case of no intersection of rotating bounding boxes and improves the numerator and denominator of KFIoU. The large differences between numerator and denominator of KFIoU in orders of magnitude. Meanwhile, to prevent the angle periodicity problem from affecting the model training effect, a classification-based DCL angle classification module is introduced in the head of the model to improve the model generalization ability and angle prediction accuracy. The experimental results show that GMIoU is closer to SkewIoU regarding variation trend, and the convergence speed is faster compared with, for example, KFIoU. The training effect improves the model's accuracy for large aspect ratio objects and shows good results on square-like objects.

Keywords

remote sensing image, rotating object, 2D Gaussian distribution, GMIoU

Cite This Paper

Zhiwei Zhang, Wei Sun. The GMIoU Loss for Accurate Rotated Object Detection in Remote Sensing Image. Academic Journal of Computing & Information Science (2023), Vol. 6, Issue 4: 68-76. https://doi.org/10.25236/AJCIS.2023.060409.

References

[1] NIE Guang tao, HUANG Hua, A Survey of Object Detection in Optical Remote Sensing Images. ACTA AUTOMATICA SINICA 47.08(2021):1749-1768. doi:10.16383/j.aas.c200596.

[2] Girshick Ross, et al. "Rich feature hierarchies for accurate object detection and semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2014.

[3] He Kaiming, et al. "Spatial pyramid pooling in deep convolutional networks for visual recognition." IEEE transactions on pattern analysis and machine intelligence 37.9 (2015): 1904-1916.

[4] Girshick Ross. "Fast r-cnn." Proceedings of the IEEE international conference on computer vision. 2015.

[5] Ren Shaoqing, et al. "Faster r-cnn: Towards real-time object detection with region proposal networks." Advances in neural information processing systems 28 (2015).

[6] Redmon Joseph, et al. "You only look once: Unified, real-time object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.

[7] LIAO Yurong1, et al. Research progress of deep learning-based object detection of optical remote sensing image. Journal on Communications 43.05(2022):190-203.

[8] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in Neural Information Processing Systems,pp. 91–99, 2015.

[9] Yang X, Liu Q, Yan J, et al. R3Det: Refined single-stage detector with feature refinement for rotating object[J/OL].(2020-2-21)[2021-10-09]. 

[10] Yang Xue, et al. "Scrdet: Towards more robust detection for small, cluttered and rotated objects." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019.

[11] LI Z H, WANG Z P, HE Y T. Aerial photography dense small target detection algorithm based on adaptive cooperative attention mechanism[J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(8):327944(in Chinese). doi: 10.7527/S1000- 6893.2022.27944

[12] YANG X, YAN J C, HE T, et al. On the arbitrary-oriented object detection: Classification based approaches revisited [J/OL]. (2021-3-27)[2021-10-09].

[13] YANG X, HOU L P, ZHOU Y, et al. Dense label encoding for boundary discontinuity free rotation detection[J/OL].(2021-5-25).2021-10-09]

[14] Rezato Figurehi H, Tsoi N, Gwak J Y, et al. Generalized intersection over union: A metric and a loss for bounding box regression[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 658-666.

[15] Chen Dong, Miao Duoqian, Control Distance IoU and Control Distance IoU Loss for Better Bounding Box Regression, Pattern Recognition, Volume 137,2023,109256, ISSN 0031-3203,

[16] Zheng Z., et al. "Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression." arXiv (2019).

[17] Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression [J]. arXiv preprint arXiv:2101.08158, 2021. 

[18] Chen Z. et al. (2020). PIoU Loss: Towards Accurate Oriented Object Detection in Complex Environments. In: Vedaldi, A., Bischof, H., Brox, T., Frahm, JM. (eds) Computer Vision – ECCV 2020. ECCV 2020. Lecture Notes in Computer Science, vol 12350. Springer, Cham.

[19] Xue Yang, Junchi Yan, Qi Ming, Wentao Wang, Xiaopeng Zhang, and Qi Tian. Rethinking rotated object detection with gaussian wasserstein distance loss. In International Conference on Machine Learning, pp. 11830–11841. PMLR, 2021c.

[20] Xue Yang, Xiaojiang Yang, Jirui Yang, Qi Ming, Wentao Wang, Qi Tian, and Junchi Yan. Learning high-precision bounding box for rotated object detection via kullback-leibler divergence. Advances in Neural Information Processing Systems, 34, 2021d.

[21] Yang X., Zhou Y., Zhang G., Yang J., Wang W., Yan J. & Tian Q. (2022). The kfiou loss for rotated object detection. arXiv preprint arXiv:2201. 12558, 2022