High resolution remote sensing image ship target detection technology based on deep learning

Abstract

With the development of China’s high-resolution special projects and the rapid development of commercial satellite, the resolution of the mainstream satellite remote sensing images has reached the sub-meter level. Ship target detection in high-resolution remote sensing images has always been the focus and hotspot in image understanding. Real-time and effective detection of ships play an extremely important role in marine transportation, military operations and so on. Firstly, the full-factor ship target sample library of high-resolution image is synthetically prepared. Then, based on the Faster R-CNN framework and Resnet model, optimize the parameters of the model to achieve accurate results. The simulation results show that the detection model trained in this paper has the highest recall rate of 98.01% and false alarm rate of 0.83%. It can be applied to the practical application of ship detection in remote sensing images.

References

[1] GaoC X, SangN. Bulletin of Surveying and Mapping, 2014, S1: 108(in Chinese)

[2] XiaoL P, CaoJ, GaoX Y. Opto-Electronic Engineering, 2007, 34: 6(in Chinese)

[3] ZhaoY H, WuX Q, WenL Y. Opto-Electronic Engineering, 2008, 35: 8

[4] M Busta, L Neumann and J Matas, An End-to-End Trainable Scene Text Localization and Recognition Framework, IEEE International Conference on Computer Vision, 2204 (2017).

[5] Alex Krizhevsky, Ilya Sutskever and Geoffrey E Hinton, Imagenet Classification with Deep Convolutional Neural Networks, The 25th International Conference on Neural Information Processing Systems, 1097 (2012).

[6] EikvilL, AurdalL, KorenH. ISPRS Journal of Photogrammetry and Remote Sensing, 2009, 64: 65

[7] EveringhamM, EslamiS M, GoolL, WilliamsC K, WinnJ, ZissermanA. International Journal of Computer Vision, 2015, 111: 98

[8] R Girshick, Fast R-CNN, IEEE International Conference on Computer Vision, 1440 (2015).

[9] K. Simonyan and A. Zisserman, Very Deep Convolutional Networks For Large-Scale Image Recognition, arXiv:1409.1556v6, 2015.

[10] C Szegedy, W Liu, Y Jia, P Sermanet, S Reed, D Anguelov, D Erhan, V Vanhoucke and A Rabinovich, Going Deeper with Convolutions, IEEE Conference on Computer Vision and Pattern Recognition, 1 (2015).

[11] R Girshick, J Donahue, T Darrell and J Malik, Rich Feature Hierarchies for Accurate Object Detection and Segmentation, IEEE Conference on Computer Vision and Pattern Recognition, 580 (2014).

[12] D M Gonzalez, M Volpi, N Komodakis and D Tuia, Rotation Equivariant Vector Field Networks, arXiv:1612.09346v3, 2017.

[13] J F Henriques and A Vedaldi, Warped Convolutions: Efficient Invariance to Spatial Transformations, The 34th International Conference on Machine Learning, 1461 (2017).

[14] KembhaviA, HarwoodD, DavisL S. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33: 1250

[15] LeitloffJ, RosenbaumD, KurzF. O Meynberg and P Reinatz. Remote Sensing, 2014, 6: 11315

[16] LiuK, MattyusG. IEEE Geosci. Remote Sensing Letters, 2015, 12: 1

[17] TangT Y, ZhouS L, DengZ P, ZouH X, LeiL. Sensors, 2017, 17: 336

[18] G S Xia, X Bai, J Ding, Z Zhu, S Belongie, J Luo, M Datcu, M Pelillo and L Zhang, DOTA: A Large-Scale Dataset for Object Detection in Aerial Images, arXiv:1711.10398v3, 2019.

[19] Y Zhou and O Tuzel, Voxelnet: End-to-End Learning for Point Cloud Based 3D Object Detection, arXiv:1711.06396, 2017.