基于SSD改进的遥感图像目标检测算法
An Improved Object Detection Algorithm Based on SSD in Remote Sensing Image
摘要: 随着光学遥感技术的发展，遥感图像的目标检测方法也在逐步完善。SSD是一种单级目标检测模型，本文是基于SSD算法应用与遥感图像的目标检测任务并且针对遥感图像与正常图像的区别有针对性的改进。首先，目标检测分类模块和检测模块同时进行优化，由于检测模块内部简单样本和困难样本分布不均匀带来的梯度更新不均衡，导致检测模块收敛慢于分类，针对这个问题提出了一种新的损失函数可以缓解梯度更新不均衡，有效地在训练过程中加快模型的收敛并提高精度。同时，提出了Laplace-NMS方法，对于目标密集情况时后处理效果更好。本文提出的损失函数相对于SSD算法采用的提高了3.47%，同时本文提出的Laplace-NMS算法相对于NMS算法有0.78%的提升。
Abstract: With the development of optical remote sensing technology, the object detection method of remote sensing image is gradually improved. SSD is a single-stage target detection model. This paper is based on the application of SSD algorithm and the object detection task of remote sensing images, and aimed at the difference between remote sensing images and normal images to make targeted improvements. Firstly, the classification module and the detection module of object detection are optimized simultaneously. Because of the uneven distribution of simple samples and difficult samples in the detection module, the gradient update is not balanced, which leads to the slower convergence of the detection module than the classification. In order to solve this problem, a new loss function is proposed, which can alleviate the imbalance of gradient updating and improve the accuracy. At the same time, the Laplace-NMS method is proposed, which has better post-processing effect when the target is dense. The loss function proposed in this paper is improved by 3.47% compared with the SSD algorithm, and the Laplace-NMS algorithm proposed in this paper is improved by 0.78% compared with the NMS algorithm.
文章引用：卢启祥. 基于SSD改进的遥感图像目标检测算法[J]. 计算机科学与应用, 2021, 11(5): 1579-1587. https://doi.org/10.12677/CSA.2021.115163

参考文献

[1] Girshick, R., Donahue, J., Darrell, T. and Malik, J. (2014) Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, 24-27 June 2014, 580-587. [Google Scholar] [CrossRef] 
[2] Girshick, R. (2015) Fast R-CNN. arXiv:1504.08083 [cs.CV] [Google Scholar] [CrossRef] 
[3] Ren, S., He, K., Girshick, R., et al. (2017) Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39, 1137-1149. [Google Scholar] [CrossRef] 
[4] Lin, T.Y., Dollár, P., Girshick, R., et al. (2016) Feature Pyra-mid Networks for Object Detection. arXiv:1612.03144 [cs.CV] [Google Scholar] [CrossRef] 
[5] Cai, Z. and Vasconcelos, N. (2017) Cascade R-CNN: Delving into High Quality Object Detection. arXiv:1712.00726 [cs.CV] [Google Scholar] [CrossRef] 
[6] Cao, J., Cholakkal, H., Anwer, R.M., et al. (2020) D2det: To-wards High Quality Object Detection and Instance Segmentation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16 November 2020, 11485-11494. [Google Scholar] [CrossRef] 
[7] Redmon, J., Divvala, S., Girshick, R., et al. (2015) You Only Look Once: Unified, Real-Time Object Detection. arXiv:1506.02640 [cs.CV] [Google Scholar] [CrossRef] 
[8] Liu, W., Anguelov, D., Erhan, D., et al. (2016) SSD: Single Shot MultiBox Detector. arXiv:1512.02325 [cs.CV] [Google Scholar] [CrossRef] 
[9] Fan, Q., Zhuo, W., Tang, C.K., et al. (2020) Few-Shot Object Detection with Attention-RPN and Multi-Relation Detector. Proceedings of the IEEE/CVF Conference on Computer Vi-sion and Pattern Recognition, Seattle, 13-19 June 2020, 4013-4022. [Google Scholar] [CrossRef] 
[10] Neubeck, A. and Gool, L.J.V. (2006) Efficient Non-Maximum Suppression. 18th International Conference on Pattern Recognition (ICPR’06), Hong Kong, 20-24 Au-gust 2006, 850-855. [Google Scholar] [CrossRef] 
[11] Simonyan, K. and Zisserman, A. (2015) Very Deep Convolutional Networks for Large Scale Image Recognition. arXiv:1409.1556 [cs.CV]
[12] Cheng, G., Han, J., Zhou, P. and Guo, L. (2014) Multi-Class Geospatial Object Detec-tion and Geographic Image Classification Based on Collection of Part Detectors. ISPRS Journal ofPhotogrammetry and Remote Sensing, 98, 119-132. [Google Scholar] [CrossRef] 
[13] Russakovsky, O., Deng, J., Su, H., et al. (2015) ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision, 115, 211-252. [Google Scholar] [CrossRef]