SAR图像舰船目标联合检测与方向估计
|
摘要
提出了一种单阶段、快速的SAR(synthetic aperture radar)图像舰船目标检测与方位向估计的方法,它在输入图像之后,进行一次前向计算即可直接输出图像中舰船的位置、类别和方位向信息,可完全端到端地进行训练和推理。该方法以SSD(single shot detector)为基础,通过特征金字塔网络充分利用高层语义特征和底层细节特征,使底层与高层都有了类别信息,解决了小尺寸目标在高层会被忽略、底层容易预测出错的问题;通过新设计的损失函数,降低数量较多的易分类样本的损失权重,避免其覆盖了数量较少的难分类样本的损失,使目标函数更快、更好地收敛;通过新增加的方位向估计模块,在增加少量计算量的条件下,在完成检测任务的同时完成方位向估计。通过公开的数据集验证了所提方法可快速准确地完成对舰船目标检测和方位向估计。
This paper proposes a fast single stage synthetic aperture radar(SAR) ship detection and azimuth estimation method. It can output the location, type and orientation of the object in the image after a forward process, which is completely end to end for training and inference. This method is based on the single shot detector(SSD). The feature pyramid network makes full use of the high-level semantic features and low-level position features, which make the bottom and top layers have class information. This can solve the following two problems: Small targets are easy ignored at the top layer and the bottom layer would predict the wrong class. The loss function reduces the weight of huge number easy classified examples, which can avoid dominating the hard classified examples. This can make the objective function converge better and faster. By adding the new azimuth estimation module, the method can perform the two tasks simultaneously with a small increase in calculation. By the experiments on the opened SAR ship detection dataset, we can find that the proposed method can detect ships and estimate the orientation rapidly and accurately.
This paper proposes a fast single stage synthetic aperture radar(SAR) ship detection and azimuth estimation method. It can output the location, type and orientation of the object in the image after a forward process, which is completely end to end for training and inference. This method is based on the single shot detector(SSD). The feature pyramid network makes full use of the high-level semantic features and low-level position features, which make the bottom and top layers have class information. This can solve the following two problems: Small targets are easy ignored at the top layer and the bottom layer would predict the wrong class. The loss function reduces the weight of huge number easy classified examples, which can avoid dominating the hard classified examples. This can make the objective function converge better and faster. By adding the new azimuth estimation module, the method can perform the two tasks simultaneously with a small increase in calculation. By the experiments on the opened SAR ship detection dataset, we can find that the proposed method can detect ships and estimate the orientation rapidly and accurately.
引文
[1] Friedman W C C,Pichel K S,Clemente-Colon W G,et al.Automatic Detection of Ships in RadarSAT-1 SAR Imagery[J].Canadian Journal of Remote Sensing,2001,27 (5):568-577
[2] Viola P,Jones M,Rapid Object Detection Using a Boosted Cascade of Simple Features[C].CVPR Colorado,USA,2001
[3] 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,IEEE Computer Society,Columbus,USA,2014
[4] Girshick R.Fast R-CNN[C].IEEE International Conference on Computer Vision (ICCV),Santiago,Chile,2015
[5] Ren S,He K,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,2017,39:1 137-1 149
[6] Dai J,Li Y,He K,et al.R-FCN:Object Detection via Region-based Fully Convolutional Networks[C].CVPR,Las Vegas,USA,2016
[7] Redmon J,Divvala S,Girshick R,et al.You Only Look Once:Unified,Real-Time Object Detection[C].CVPR.Las Vegas,USA,2016
[8] Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[C].ECCV 2016,Amsterdam,Netherlands,2016
[9] Simonyan K,Zisserman A.Very Deep Convolutional Networks for Large-scale Image Recognition[C].CVPR,Columbus,USA,2014
[10] Cai Z,Fan Q,Feris R S,et al.A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection[C].European Conference on Computer Vision,The Netherlands,2016
[11] Hariharan B,Arbelaez P,Girshick R,et al.Hypercolumns for Object Segmentation and Fine-grained Localization[C].CVPR,Boston,USA,2015
[12] Kong T,Yao A,Chen Y,et al.HyperNet:Towards Accurate Region Proposal Generation and Joint Object Detection[C].CVPR,Las Vegas,USA,2016
[13] Liu W,Rabinovich A,Berg A C.ParseNet:Looking Wider to See Better[C].CVPR,Boston,USA,2015
[14] Bell S,Zitnick C L,Bala K,et al.Inside-Outside Net:Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks[C].IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,NV,USA,2016
[15] Shrivastava A,Gupta A,Girshick R.Training Region Based Object Detectors with Online Hard Example Mining[C].CVPR,Las Vegas,USA,2016
[16] Lin T Y ,Goyal P ,Girshick R ,et al.Focal Loss for Dense Object Detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017(99):2 999-3 007
[17] Krizhevsky A,Sutskever I,Hinton G.Image Net Classification with Deep Convolutional Neural Networks[J].Advances in Neural Information Processing Systems,2012,25(2):1 097-1 105
[18] Erhan D,Szegedy C,Toshev A,et al.Scalable Object Detection Using Deep Neural Networks[C].IEEE Conference on Computer Vision and Pattern Recognition,Washington D C,USA,2014
[19] Fu C Y,Liu W,Ranga A,et al.DSSD :Deconvolutional Single Shot Detector[C].CVPR,Honolulu,USA,2017
[20] Lin T Y,Dollár P,Girshick R,et al.Feature Pyramid Networks for Object Detection[C].CVPR,Honolulu,USA,2017
[21] Poirson P,Ammirato P,Fu C Y,et al.Fast Single Shot Detection and Pose Estimation[C].Fourth International Conference on 3D Vision,IEEE,Cornell,USA,2016
[2] Viola P,Jones M,Rapid Object Detection Using a Boosted Cascade of Simple Features[C].CVPR Colorado,USA,2001
[3] 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,IEEE Computer Society,Columbus,USA,2014
[4] Girshick R.Fast R-CNN[C].IEEE International Conference on Computer Vision (ICCV),Santiago,Chile,2015
[5] Ren S,He K,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,2017,39:1 137-1 149
[6] Dai J,Li Y,He K,et al.R-FCN:Object Detection via Region-based Fully Convolutional Networks[C].CVPR,Las Vegas,USA,2016
[7] Redmon J,Divvala S,Girshick R,et al.You Only Look Once:Unified,Real-Time Object Detection[C].CVPR.Las Vegas,USA,2016
[8] Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[C].ECCV 2016,Amsterdam,Netherlands,2016
[9] Simonyan K,Zisserman A.Very Deep Convolutional Networks for Large-scale Image Recognition[C].CVPR,Columbus,USA,2014
[10] Cai Z,Fan Q,Feris R S,et al.A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection[C].European Conference on Computer Vision,The Netherlands,2016
[11] Hariharan B,Arbelaez P,Girshick R,et al.Hypercolumns for Object Segmentation and Fine-grained Localization[C].CVPR,Boston,USA,2015
[12] Kong T,Yao A,Chen Y,et al.HyperNet:Towards Accurate Region Proposal Generation and Joint Object Detection[C].CVPR,Las Vegas,USA,2016
[13] Liu W,Rabinovich A,Berg A C.ParseNet:Looking Wider to See Better[C].CVPR,Boston,USA,2015
[14] Bell S,Zitnick C L,Bala K,et al.Inside-Outside Net:Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks[C].IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,NV,USA,2016
[15] Shrivastava A,Gupta A,Girshick R.Training Region Based Object Detectors with Online Hard Example Mining[C].CVPR,Las Vegas,USA,2016
[16] Lin T Y ,Goyal P ,Girshick R ,et al.Focal Loss for Dense Object Detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017(99):2 999-3 007
[17] Krizhevsky A,Sutskever I,Hinton G.Image Net Classification with Deep Convolutional Neural Networks[J].Advances in Neural Information Processing Systems,2012,25(2):1 097-1 105
[18] Erhan D,Szegedy C,Toshev A,et al.Scalable Object Detection Using Deep Neural Networks[C].IEEE Conference on Computer Vision and Pattern Recognition,Washington D C,USA,2014
[19] Fu C Y,Liu W,Ranga A,et al.DSSD :Deconvolutional Single Shot Detector[C].CVPR,Honolulu,USA,2017
[20] Lin T Y,Dollár P,Girshick R,et al.Feature Pyramid Networks for Object Detection[C].CVPR,Honolulu,USA,2017
[21] Poirson P,Ammirato P,Fu C Y,et al.Fast Single Shot Detection and Pose Estimation[C].Fourth International Conference on 3D Vision,IEEE,Cornell,USA,2016