基于Faster-RCNN的遥感图像飞机检测算法
|
摘要
CCCV2017发布遥感图像飞机数据集,用于评测飞机检测算法.针对该遥感图像数据集中的飞机朝向不确定、图像覆盖范围广、图像背景复杂度高,导致飞机检测难度大、检测算法准确率和算法泛化能力低等问题,提出了基于Faster-RCNN的飞机检测改进算法.首先,通过对图像采用翻转以及角度旋转等方式对数据集进行合理的扩增;然后,在扩增后的数据集上,使用深度残差网络对图像进行特征提取,针对数据集中飞机目标的长宽比特点优化区域建议网络;同时,为了防止训练集中正负样本不均衡,采用在线困难样本挖掘方法对数据进行训练.在CCCV2017数据集上评估表明,改进后的Faster-RCNN算法极大提高了初始的Faster-RCNN算法性能,在测试集上m AP达到了89.93%.在NWPUVHR-10、NWPU-RESISC45、UCAS-AOD遥感图像飞机数据集测试表明,该改进模型同样具有良好的性能,从而验证了该模型具有良好的鲁棒性和泛化能力.
CCCV2017 releases remote sensing image airplane dataset for evaluating airplane detection algorithm.Due to the uncertainty of the orientation of airplanes in remote sensing images and the images with a wide coverage and high background complexity,airplane detection is difficult,the precision and the generalization ability of the model are low.This paper proposes an improved airplane detection algorithm based on Faster-RCNN.First of all,the dataset is reasonably augmented by flipping and rotating the images;then,on the augmented dataset,the residual network is used to extract features from the images and the region proposal network is optimized based on the characteristics of the aspect ratio of airplanes;at the same time,in order to prevent imbalance between positive and negative samples in the training set,the online hard example mining method is used to train the data.The evaluation on the CCCV2017 dataset shows that the improved Faster-RCNN algorithm greatly improved the performance of the initial Faster-RCNN algorithm.In the test set,the m AP(mean Average Precision,m AP) has reached 89.93%.Tests on NWPU VHR-10,NWPU VHR-45,and UCAS-AOD remote sensing image datasets show that the improved model also has good performance,which verifies that the model has good robustness and generalization ability.
CCCV2017 releases remote sensing image airplane dataset for evaluating airplane detection algorithm.Due to the uncertainty of the orientation of airplanes in remote sensing images and the images with a wide coverage and high background complexity,airplane detection is difficult,the precision and the generalization ability of the model are low.This paper proposes an improved airplane detection algorithm based on Faster-RCNN.First of all,the dataset is reasonably augmented by flipping and rotating the images;then,on the augmented dataset,the residual network is used to extract features from the images and the region proposal network is optimized based on the characteristics of the aspect ratio of airplanes;at the same time,in order to prevent imbalance between positive and negative samples in the training set,the online hard example mining method is used to train the data.The evaluation on the CCCV2017 dataset shows that the improved Faster-RCNN algorithm greatly improved the performance of the initial Faster-RCNN algorithm.In the test set,the m AP(mean Average Precision,m AP) has reached 89.93%.Tests on NWPU VHR-10,NWPU VHR-45,and UCAS-AOD remote sensing image datasets show that the improved model also has good performance,which verifies that the model has good robustness and generalization ability.
引文
[1]尹宏鹏,陈波,柴毅,等.基于视觉的目标检测与跟踪综述[J].自动化学报,2016,42(10):1466-1489.
[2] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International journal of computer vision,2004,60(2):91-110.
[3] DALAL N,TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the IEEE Computer SocietyConference on Computer Vision&Pattern Recognition. San Diego,2005,1(12):886-893.
[4] MALLAT S. A wavelet tour of signal processing[M]. Beijing:China Machine Press,2010.
[5] VAPNIK V,CORTES C. Support vector networks[J]. Machine learning,1995,20(3):273-297.
[6] CHENG J,GREINER R. Comparing bayesian network classifiers[J]. IEEE transactions on vehicular technology,2013,63(5):2002-2012.
[7] LECUN Y,BOTTOU L,BENGIO Y,et al. Gradient-based learning applied to document recognition[J]. Proceedings of theIEEE,1998,86(11):2278-2324.
[8] UIJLINGS J R,SANDE K E,GEVERS T,et al. Selective search for object recognition[J]. International journal of computervision,2013,104(2):154-171.
[9] GIRSHICK R,DONAHUE J,DARRELL T. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Columbus,2014:580-587.
[10] GIRSHICK R. Fast R-CNN[DB/OL].[2018-10-22]. https://arxiv.org/pdf/1504.08083.pdf.
[11] REN S Q,HE K M,GIRSHICK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEEtransactions on pattern analysis&machine intelligence,2017,39(6):1137-1149.
[12] REDMON J,DIVVALA S,GIRSHICK R,et al. You only look once:unified,real-time object detection[C]//Proceedings of the IEEEConference on Computer Vision and Pattern Recognition. Las Vegas,2016:779-788.
[13] HE K M,ZHANG X Y,REN S Q,et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference onComputer Vision and Pattern Recognition. Las Vegas,2016:770-778.
[14] SHRIVASTAVA A,GUPTA A,GIRSHICK R. Training region-based object detectors with online hard example mining[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas,2016:761-769.
[15] DAI J,LI Y,HE K,et al. R-FCN:object detection via region-based fully convolutional networks[DB/OL].[2018-10-22].https://arxiv.org/pdf/1605.06409.pdf.
[16] ZEILER M D,FERGUS R. Visualizing and understanding convolutional networks[DB/OL].[2018-10-22]. https://arxiv.org/pdf/1311.2901.pdf.
[17] SIMONYAN K,ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[DB/OL].[2018-10-22].https://arxiv.org/pdf/1409.1556.pdf.
[18] CHENG G,HAN J W,ZHOU P C,et al. Multi-class geospatial object detection and geographic image classification based oncollection of part detectors[J]. Isprs journal of photogrammetry&remote sensing,2014,98(1):119-132.
[19] CHENG G,HAN J W. A survey on object detection in optical remote sensing images[J]. Isprs journal of photogrammetry&remote sensing,2016,117:11-28.
[20] CHENG G,ZHOU P C,HAN J W. Learning rotation-invariant convolutional neural networks for object detection in VHR opticalremote sensing images[J]. IEEE transactions on geoscience&remote sensing,2016,54(12):7405-7415.
[21] CHENG G,HAN J W,LU X Q. Remote sensing image scene classification:benchmark and state of the art[J]. Proceedings ofthe IEEE,2017,105(10):1865-1883.
[22] ZHU H G,CHEN X G,DAI W Q,et al. Orientation robust object detection in aerial images using deep convolutional neuralnetwork[C]//Proceedings of the IEEE International Conference on Image Processing. Quebec,2015:3735-3739.
[2] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International journal of computer vision,2004,60(2):91-110.
[3] DALAL N,TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the IEEE Computer SocietyConference on Computer Vision&Pattern Recognition. San Diego,2005,1(12):886-893.
[4] MALLAT S. A wavelet tour of signal processing[M]. Beijing:China Machine Press,2010.
[5] VAPNIK V,CORTES C. Support vector networks[J]. Machine learning,1995,20(3):273-297.
[6] CHENG J,GREINER R. Comparing bayesian network classifiers[J]. IEEE transactions on vehicular technology,2013,63(5):2002-2012.
[7] LECUN Y,BOTTOU L,BENGIO Y,et al. Gradient-based learning applied to document recognition[J]. Proceedings of theIEEE,1998,86(11):2278-2324.
[8] UIJLINGS J R,SANDE K E,GEVERS T,et al. Selective search for object recognition[J]. International journal of computervision,2013,104(2):154-171.
[9] GIRSHICK R,DONAHUE J,DARRELL T. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Columbus,2014:580-587.
[10] GIRSHICK R. Fast R-CNN[DB/OL].[2018-10-22]. https://arxiv.org/pdf/1504.08083.pdf.
[11] REN S Q,HE K M,GIRSHICK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEEtransactions on pattern analysis&machine intelligence,2017,39(6):1137-1149.
[12] REDMON J,DIVVALA S,GIRSHICK R,et al. You only look once:unified,real-time object detection[C]//Proceedings of the IEEEConference on Computer Vision and Pattern Recognition. Las Vegas,2016:779-788.
[13] HE K M,ZHANG X Y,REN S Q,et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference onComputer Vision and Pattern Recognition. Las Vegas,2016:770-778.
[14] SHRIVASTAVA A,GUPTA A,GIRSHICK R. Training region-based object detectors with online hard example mining[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas,2016:761-769.
[15] DAI J,LI Y,HE K,et al. R-FCN:object detection via region-based fully convolutional networks[DB/OL].[2018-10-22].https://arxiv.org/pdf/1605.06409.pdf.
[16] ZEILER M D,FERGUS R. Visualizing and understanding convolutional networks[DB/OL].[2018-10-22]. https://arxiv.org/pdf/1311.2901.pdf.
[17] SIMONYAN K,ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[DB/OL].[2018-10-22].https://arxiv.org/pdf/1409.1556.pdf.
[18] CHENG G,HAN J W,ZHOU P C,et al. Multi-class geospatial object detection and geographic image classification based oncollection of part detectors[J]. Isprs journal of photogrammetry&remote sensing,2014,98(1):119-132.
[19] CHENG G,HAN J W. A survey on object detection in optical remote sensing images[J]. Isprs journal of photogrammetry&remote sensing,2016,117:11-28.
[20] CHENG G,ZHOU P C,HAN J W. Learning rotation-invariant convolutional neural networks for object detection in VHR opticalremote sensing images[J]. IEEE transactions on geoscience&remote sensing,2016,54(12):7405-7415.
[21] CHENG G,HAN J W,LU X Q. Remote sensing image scene classification:benchmark and state of the art[J]. Proceedings ofthe IEEE,2017,105(10):1865-1883.
[22] ZHU H G,CHEN X G,DAI W Q,et al. Orientation robust object detection in aerial images using deep convolutional neuralnetwork[C]//Proceedings of the IEEE International Conference on Image Processing. Quebec,2015:3735-3739.