基于Faster R-CNN的高分辨率图像目标检测技术
|
摘要
为提升传统算法对高分辨率遥感图像中地物目标的检测效果,将深度学习目标检测框架快速区域卷积神经网络(faster regions with convolutional neural network,Faster R-CNN)应用于高分辨率遥感图像目标检测任务中。以机场为检测场景、飞机为检测目标进行实验,首先,利用高分辨率遥感图像数据集训练Faster R-CNN框架,得到相应的目标检测模型;然后,采用该模型对高分辨率遥感图像中的飞机目标进行检测;最后,对实验结果进行统计分析及评价。实验结果表明,Faster R-CNN模型能够全面而准确地检测飞机目标,最优F1分数值为0. 976 3,并且同一个模型可以对多种高分辨率遥感图像进行目标检测。
In order to improve the detection effect of the traditional algorithm on the ground objects in high resolution remote sensing images,this paper applies the deep learning object detection framework Faster R-CNN to the object detection task of high resolution remote sensing images. The airport and aircraft are used as the test scene and detection object for the experiment respectively,The Faster R-CNN framework is trained using the high-resolution remote sensing image data set to obtain the corresponding object detection model. The model is used to detect aircraft objects in high resolution remote sensing images and perform statistical analysis of the experimental results. The experimental results show that the Faster R-CNN model can entirely and accurately detect aircraft objects with an optimal F1 score of 0. 976 3,and the same model can be used for object detection of multiple high resolution remote sensing images.
In order to improve the detection effect of the traditional algorithm on the ground objects in high resolution remote sensing images,this paper applies the deep learning object detection framework Faster R-CNN to the object detection task of high resolution remote sensing images. The airport and aircraft are used as the test scene and detection object for the experiment respectively,The Faster R-CNN framework is trained using the high-resolution remote sensing image data set to obtain the corresponding object detection model. The model is used to detect aircraft objects in high resolution remote sensing images and perform statistical analysis of the experimental results. The experimental results show that the Faster R-CNN model can entirely and accurately detect aircraft objects with an optimal F1 score of 0. 976 3,and the same model can be used for object detection of multiple high resolution remote sensing images.
引文
[1]明冬萍,骆剑承,沈占锋,等.高分辨率遥感影像信息提取与目标识别技术研究[J].测绘科学,2005,30(3):18-20.Ming D P,Luo J C,Shen Z F,et al. Research on high resolution remote sensing image information extraction and target recognition technology[J]. Surveying and Mapping Science,2005,30(3):18-20.
[2]吴樊,王超,张红,等.基于知识的中高分辨率光学卫星遥感影像桥梁目标识别研究[J].电子与信息学报,2006,28(4):587-591.Wu F,Wang C,Zhang H,et al. Research on bridge target recognition based on knowledge of medium and high resolution optical satellite remote sensing images[J]. Journal of Electronics and Information Technology,2006,28(4):587-591.
[3]王文宇,李博.基于e Cognition的高分辨率遥感图像的自动识别分类技术[J].北京建筑工程学院学报,2006,22(4):26-29.Wang W Y,Li B. Automatic recognition and classification of high resolution remote sensing images based on e Cognition[J]. Journal of Beijing Institute of Civil Engineering and Architecture,2006,22(4):26-29.
[4]黄凯奇,任伟强,谭铁牛.图像物体分类与检测算法综述[J].计算机学报,2014,37(6):1225-1240.Huang K Q,Ren W Q,Tan T N. Summarization of image object classification and detection algorithm[J]. Journal of Computer,2014,37(6):1225-1240.
[5]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.Yin H P,Chen B,Chai Y,et al. Vision-based object detection and tracking:A review[J]. Journal of Automation,2016,42(10):1466-1489.
[6] Girshick R,Donahue J,Darrell T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Computer Vision and Pattern Recognition,2014:580-587.
[7] Peng X,Schmid C. Multi-region two-stream R-CNN for action detection[C]//European Conference on Computer Vision,2016:744-759.
[8] Girshick R. Fast R-CNN[C]//IEEE International Conference on Computer Vision,2015:1440-1448.
[9] He K,Zhang X,Ren S,et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,37(9):1904-1916.
[10] Li J N,Liang X D,Shen S M. Scale-aware fast R-CNN for pedestrian detection[EB/OL].(2015-10-28). https://arxiv.org/pdf/1510. 08160. pdf.
[11] Ren S,He K,Girshick R,et al. Faster R-CNN:Towards realtime object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems,2015:91-99.
[12] Jiang H,Learned-Miller E. Face detection with the Faster RCNN[C]//IEEE International Conference on Automatic Face and Gesture Recognition,2017:650-657.
[13] Simonyan K,Zisserman A. Very deep convolutional networks for large-scale image recognition[EB/OL].(2014-09-04). https://arxiv. org/pdf/1409. 1556. pdf.
[14] He K,Zhang X,Ren S,et al. Deep residual learning for image recognition[C]//IEEE Computer Vision and Pattern Recognition,2016:770-778.
[15] Szegedy C,Liu W,Jia Y,et al. Going deeper with convolutions[C]//IEEE Computer Vision and Pattern Recognition,2015:1-9.
[16] Lin T Y,Maire M,Belongie S,et al. Microsoft COCO:Common objects in context[C]//European Conference on Computer Vision,2014:740-755.
[2]吴樊,王超,张红,等.基于知识的中高分辨率光学卫星遥感影像桥梁目标识别研究[J].电子与信息学报,2006,28(4):587-591.Wu F,Wang C,Zhang H,et al. Research on bridge target recognition based on knowledge of medium and high resolution optical satellite remote sensing images[J]. Journal of Electronics and Information Technology,2006,28(4):587-591.
[3]王文宇,李博.基于e Cognition的高分辨率遥感图像的自动识别分类技术[J].北京建筑工程学院学报,2006,22(4):26-29.Wang W Y,Li B. Automatic recognition and classification of high resolution remote sensing images based on e Cognition[J]. Journal of Beijing Institute of Civil Engineering and Architecture,2006,22(4):26-29.
[4]黄凯奇,任伟强,谭铁牛.图像物体分类与检测算法综述[J].计算机学报,2014,37(6):1225-1240.Huang K Q,Ren W Q,Tan T N. Summarization of image object classification and detection algorithm[J]. Journal of Computer,2014,37(6):1225-1240.
[5]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.Yin H P,Chen B,Chai Y,et al. Vision-based object detection and tracking:A review[J]. Journal of Automation,2016,42(10):1466-1489.
[6] Girshick R,Donahue J,Darrell T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Computer Vision and Pattern Recognition,2014:580-587.
[7] Peng X,Schmid C. Multi-region two-stream R-CNN for action detection[C]//European Conference on Computer Vision,2016:744-759.
[8] Girshick R. Fast R-CNN[C]//IEEE International Conference on Computer Vision,2015:1440-1448.
[9] He K,Zhang X,Ren S,et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,37(9):1904-1916.
[10] Li J N,Liang X D,Shen S M. Scale-aware fast R-CNN for pedestrian detection[EB/OL].(2015-10-28). https://arxiv.org/pdf/1510. 08160. pdf.
[11] Ren S,He K,Girshick R,et al. Faster R-CNN:Towards realtime object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems,2015:91-99.
[12] Jiang H,Learned-Miller E. Face detection with the Faster RCNN[C]//IEEE International Conference on Automatic Face and Gesture Recognition,2017:650-657.
[13] Simonyan K,Zisserman A. Very deep convolutional networks for large-scale image recognition[EB/OL].(2014-09-04). https://arxiv. org/pdf/1409. 1556. pdf.
[14] He K,Zhang X,Ren S,et al. Deep residual learning for image recognition[C]//IEEE Computer Vision and Pattern Recognition,2016:770-778.
[15] Szegedy C,Liu W,Jia Y,et al. Going deeper with convolutions[C]//IEEE Computer Vision and Pattern Recognition,2015:1-9.
[16] Lin T Y,Maire M,Belongie S,et al. Microsoft COCO:Common objects in context[C]//European Conference on Computer Vision,2014:740-755.