基于卷积神经网络的遥感图像目标检测
|
摘要
针对遥感图像中的目标检测问题,采用基于卷积神经网络的目标检测框架对目标进行提取,针对该网络制作了包含三类遥感图像中常见目标的目标检测数据集。为了解决遥感图像目标旋转角度较大的问题,将空间变换网络融入超快区域卷积神经网络,提出了一种具有旋转不变性自学习能力的目标检测模型。通过与传统的目标检测方法进行对比分析,探究了不同方法对遥感图像目标检测的实际效果。相对于传统的目标检测方法,融合了空间变换网络的卷积神经网络所提取的特征具有更好的旋转不变特性,从而能够达到更高的检测精度。
Aiming at the problem of object detection in remote sensing images,the Faster-Rcnn network based on the convolutional neural network models is used to extract the features of the object area.An object detection dataset containing three kinds of common targets in remote sensing images is made to train this network.In addition,in order to solve the problem of large rotation angle of remote sensing images,a target detection model with a rotation invariance self-learning ability is proposed,which integrates the spatial transformation network into the Faster RCNN framework.By the analysis and comparison with the traditional object detection methods,the true effects of object detection in remote sensing images by different methods are explored.The features extracted by the convolutional neural networks based on the spatial transformation networks possess stronger orientation robustness than those by the traditional methods,which makes it possible to obtain a high detection precision.
Aiming at the problem of object detection in remote sensing images,the Faster-Rcnn network based on the convolutional neural network models is used to extract the features of the object area.An object detection dataset containing three kinds of common targets in remote sensing images is made to train this network.In addition,in order to solve the problem of large rotation angle of remote sensing images,a target detection model with a rotation invariance self-learning ability is proposed,which integrates the spatial transformation network into the Faster RCNN framework.By the analysis and comparison with the traditional object detection methods,the true effects of object detection in remote sensing images by different methods are explored.The features extracted by the convolutional neural networks based on the spatial transformation networks possess stronger orientation robustness than those by the traditional methods,which makes it possible to obtain a high detection precision.
引文
[1]Han J W,Zhang D W,Cheng G,et al.Object detection in optical remote sensing images based on weakly supervised learning and high-level feature learning[J].IEEE Transactions on Geoscience and Remote Sensing,2015,53(6):3325-3337.
[2]Shao W,Yang W,Liu G,et al.Car detection from high-resolution aerial imagery using multiple features[C]//IEEE International Geoscience and Remote Sensing Symposium,2012:4379-4382.
[3]Zhang F,Du B,Zhang L P.Saliency-guided unsupervised feature learning for scene classification[J].IEEE Transactions on Geoscience and Remote Sensing,2015,53(4):2175-2184.
[4]Xiao Z F,Liu Q,Tang G F,et al.Elliptic Fourier transformation-based histograms of oriented gradients for rotationally invariant object detection in remotesensing images[J].International Journal of Remote Sensing,2015,36(2):618-644.
[5]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[6]Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition,2014:580-587.
[7]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.
[8]Jaderberg M,Simonyan K,Zisserman A,et al.Spatial transformer networks[EB/OL].(2016-02-04)[2018-07-25].https://arxiv.org/abs/1506.02025.
[9]Girshick R.Fast R-CNN[C]//IEEE International Conference on Computer Vision(ICCV),2015:1440-1448.
[10]Uijlings J R R,van de Sande K E A,Gevers T,et al.Selective search for object recognition[J].International Journal of Computer Vision,2013,104(2):154-171.
[11]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[EB/OL].(2015-04-10)[2018-07-25].https://arxiv.org/abs/1409.1556.
[12]He K M,Zhang X Y,Ren S Q,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:770-778.
[13]Zhu H G,Chen X G,Dai W Q,et al.Orientation robust object detection in aerial images using deep convolutional neural network[C]//IEEE International Conference on Image Processing(ICIP),2015:3735-3739.
[14]Zhu M M,Xu Y L,Ma S P,et al.Airport detection method with improved region-based convolutional neural network[J].Acta Optica Sinica,2018,38(7):0728001.朱明明,许悦雷,马时平,等.改进区域卷积神经网络的机场检测方法[J].光学学报,2018,38(7):0728001.
[15]Wei Y M,Quan J C,Hou Y Q Y.Aerial image location of unmanned aerial vehicle based on YOLO v2[J].Laser&Optoelectronics Progress,2017,54(11):111002.魏湧明,全吉成,侯宇青阳.基于YOLO v2的无人机航拍图像定位研究[J].激光与光电子学进展,2017,54(11):111002.
[16]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]Shao W,Yang W,Liu G,et al.Car detection from high-resolution aerial imagery using multiple features[C]//IEEE International Geoscience and Remote Sensing Symposium,2012:4379-4382.
[3]Zhang F,Du B,Zhang L P.Saliency-guided unsupervised feature learning for scene classification[J].IEEE Transactions on Geoscience and Remote Sensing,2015,53(4):2175-2184.
[4]Xiao Z F,Liu Q,Tang G F,et al.Elliptic Fourier transformation-based histograms of oriented gradients for rotationally invariant object detection in remotesensing images[J].International Journal of Remote Sensing,2015,36(2):618-644.
[5]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[6]Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition,2014:580-587.
[7]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.
[8]Jaderberg M,Simonyan K,Zisserman A,et al.Spatial transformer networks[EB/OL].(2016-02-04)[2018-07-25].https://arxiv.org/abs/1506.02025.
[9]Girshick R.Fast R-CNN[C]//IEEE International Conference on Computer Vision(ICCV),2015:1440-1448.
[10]Uijlings J R R,van de Sande K E A,Gevers T,et al.Selective search for object recognition[J].International Journal of Computer Vision,2013,104(2):154-171.
[11]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[EB/OL].(2015-04-10)[2018-07-25].https://arxiv.org/abs/1409.1556.
[12]He K M,Zhang X Y,Ren S Q,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:770-778.
[13]Zhu H G,Chen X G,Dai W Q,et al.Orientation robust object detection in aerial images using deep convolutional neural network[C]//IEEE International Conference on Image Processing(ICIP),2015:3735-3739.
[14]Zhu M M,Xu Y L,Ma S P,et al.Airport detection method with improved region-based convolutional neural network[J].Acta Optica Sinica,2018,38(7):0728001.朱明明,许悦雷,马时平,等.改进区域卷积神经网络的机场检测方法[J].光学学报,2018,38(7):0728001.
[15]Wei Y M,Quan J C,Hou Y Q Y.Aerial image location of unmanned aerial vehicle based on YOLO v2[J].Laser&Optoelectronics Progress,2017,54(11):111002.魏湧明,全吉成,侯宇青阳.基于YOLO v2的无人机航拍图像定位研究[J].激光与光电子学进展,2017,54(11):111002.
[16]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.