摘要
提出了一种特征融合结合软判决的飞机检测方法。以区域卷积神经网络为基本框架,依次采用L2范数归一化、特征连接、尺度缩放和特征降维来融合多层特征。为了降低网络在目标高度重叠时的漏检率,引入软判决来改进传统的非极大值抑制方法。实验结果表明,所提方法能够准确快速地检测到飞机,得到检测率为94.25%、虚警率为5.5%、平均运行时间为0.16 s的实验结果。与现有的其他检测方法相比,所提方法的各项指标均得到显著提升。
An airplane detection method is proposed based on feature fusion and soft decision, in which the region-based convolutional neural network is used as the basic framework and the L2 normalization, feature connection, scaling, and dimensionality reduction are in turn used to fuse the multi-layer features. The soft decision, which can improve the traditional non-maximum suppression method, is introduced in order to reduce the detection-omission-rate of grids in the case of significant overlap of targets. The experimental results show that the proposed method can be used to detect airplanes accurately and quickly with a detection rate of 94.25%, a false alarm rate of 5.5%, and the average running time of 0.16 s. Compared with those of the other existing detection methods, each index of the proposed method is significantly improved.
引文
[1] Song M Z, Qu H S, Jin G. Weak ship target detection of noisy optical remote sensing image on sea surface[J]. Acta Optica Sinica, 2017, 37(10): 1011004. 宋明珠, 曲宏松, 金光. 含噪光学遥感图像海面弱小舰船目标检测[J]. 光学学报, 2017, 37(10): 1011004.
[2] Li W, Xiang S M, Wang H B, et al. Robust airplane detection in satellite images[C]. IEEE International Conference on Image Processing, 2011: 2821-2824.
[3] Bo S K, Jing Y J. Region-based airplane detection in remotely sensed imagery[C]. International Congress on Image and Signal Processing, 2010: 1923-1926.
[4] Liu L, Shi Z W. Airplane detection based on rotation invariant and sparse coding in remote sensing images[J]. Optik-International Journal for Light and Electron Optics, 2014, 125(18): 5327-5333.
[5] Chen X Y, Xiang S M, Liu C L, et al. Aircraft detection by deep belief nets[C]. Asian Conference on Pattern Recognition, 2013: 54-58.
[6] Wu H, Zhang H, Zhang J F, et al. Fast aircraft detection in satellite images based on convolutional neural networks[C]. IEEE International Conference on Image Processing, 2015: 4210-4214.
[7] Xin P, Xu Y L, Tan H, et al. Fast airplane detection based on multi-layer feature fusion of fully convolutional networks[J]. Acta Optica Sinica, 2018, 38(3): 0315003. 辛鹏, 许悦雷, 唐红, 等. 全卷积网络多层特征融合的飞机快速检测[J]. 光学学报, 2018, 38(3): 0315003.
[8] Li Y, Fu K, Sun H, et al. An aircraft detection framework based on reinforcement learning and convolutional neural networks in remote sensing images[J]. Remote Sensing, 2018, 10(2): 243.
[9] Zhang W C, Sun X, Fu K, et al. Object detection in high-resolution remote sensing images using rotation invariant parts based model[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(1): 74-78.
[10] Wang G L, Wang X C, Fan B, et al. Feature extraction by rotation-invariant matrix representation for object detection in aerial image[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(6): 851-855.
[11] Pan S J, Yang Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10): 1345-1359.
[12] Ren S Q, He K M, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[13] Girshick R B. Fast R-CNN[C]. International Conference on Computer Vision, 2015: 1440-1448.
[14] Kong T, Yao A, Chen Y, et al. HyperNet: towards accurate region proposal generation and joint object detection[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2016: 845-853.
[15] Liu W, Rabinovich A, Berg A C, et al. ParseNet: looking wider to see better[EB/OL]. (2015-11-23)[2018-01-15]. https://arxiv.org/abs/1506.04579
[16] Rothe R, Guillaumin M, Van Gool L, et al. Non-maximum suppression for object detection by passing messages between windows[C]. Asian Conference on Computer Vision, 2014: 290-306.