基于深度卷积网络的遥感影像建筑物分割方法
|
摘要
大规模可见光遥感图像的建筑物提取是遥感图像分析领域中的一项重要工作,但是,在真实环境下,城市建筑物的尺寸范围变化大、颜色纹理轮廓复杂,加上树木等造成的遮挡,以及光照等原因,影响建筑物分割的精度。因此,文中设计并实现了两种端到端全卷积神经网络的分割方法,并在两个网络模型中加入剪裁层用以解决预测结果中产生的拼接痕迹问题,同时将IOU评价标准变形加入损失函数中,来提高模型分割精度。两个模型以不同尺度的遥感影像作为网络的输入,将网络模型输出结果采用加权的方式进行融合,进一步提高遥感影像建筑物识别和分割精度。在公开的Inria遥感影像数据集上的实验证明了该方法在遥感影像建筑物分割上的有效性。
Building extraction technique based on large-scale optical remote sensing images plays an important role in the field of remote sensing image analysis. But in the real environment,due to the big range of urban building's size,the complexity of building's colors,texture and contour,the occlusion of trees,as well as the illumination intensity,the precision of building segmentation is decreased. In order to improve the accuracy of building segmentation,two kinds of end-to-end full convolution networks(FCN) are proposed and realized,then crop layer is added to these two models to solve the visible boundary on predicted patches. Meanwhile IOU index are added into the loss function to improve the segmentation accuracy. These two networks use different scale images as input,and the two output images are fused in a weighted way. The experiment on Inria aerial imagery dataset shows that this method is effective in building segmentation of remote sensing imagery.
Building extraction technique based on large-scale optical remote sensing images plays an important role in the field of remote sensing image analysis. But in the real environment,due to the big range of urban building's size,the complexity of building's colors,texture and contour,the occlusion of trees,as well as the illumination intensity,the precision of building segmentation is decreased. In order to improve the accuracy of building segmentation,two kinds of end-to-end full convolution networks(FCN) are proposed and realized,then crop layer is added to these two models to solve the visible boundary on predicted patches. Meanwhile IOU index are added into the loss function to improve the segmentation accuracy. These two networks use different scale images as input,and the two output images are fused in a weighted way. The experiment on Inria aerial imagery dataset shows that this method is effective in building segmentation of remote sensing imagery.
引文
[1] 孙家抦.遥感原理与应用[M].第2版.武汉:武汉大学出版社,2009.
[2] 巫兆聪,胡忠文,张谦,等.结合光谱、纹理与形状结构信息的遥感影像分割方法[J].测绘学报,2013,42(1):44-50.
[3] 莫玉.高光谱图像特征学习与分类算法研究[D].西安:西安电子科技大学,2015.
[4] 王巧玉.基于深度学习的高光谱遥感图像分类[D].泉州:华侨大学,2016.
[5] 魏德强.高分辨率遥感影像建筑物提取技术研究[D].郑州:解放军信息工程大学,2013.
[6] 付卓,胡吉平,谭衢霖,等.遥感应用分析中影像分割方法[J].遥感技术与应用,2006,21(5):456-462.
[7] 代具亭,汤心溢,刘鹏.基于深度学习的语义分割网络[J].红外,2018,39(4):33-38.
[8] LONG J,SHELHAMER E,DARRELL T.Fully convolutional networks for semantic segmentation[C]//IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2015:3431-3440.
[9] ZHENG S,JAYASUMANA S,ROMERA-PAREDES B,et al.Conditional random fields as recurrent neural networks[C]//IEEE international conference on computer vision.Santiago,Chile:IEEE,2016:1529-1537.
[10] BADRINARAYANAN V,KENDALL A,CIPOLLA R.SegNet:a deep convolutional encoder-decoder architecture for scene segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017,39(12):2481-2495.
[11] RONNEBERGER O,FISCHER P,BROX T.U-Net:convolutional networks for biomedical image segmentation[C]//International conference on medical image computing and computer-assisted intervention.Cham:Springer,2015:234-241.
[12] CHEN L C,PAPANDREOU G,KOKKINOS I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2018,40(4):834-848.
[13] HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al.Deep residual learning for image recognition[C]//IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2016:770-778.
[14] WANG S,FIDLER S,URTASUN R.Holistic 3D scene understanding from a single geo-tagged image[C]//Computer vision and pattern recognition.Boston,MA,USA:IEEE,2015:3964-3972.
[15] MAGGIORI E,TARABALKA Y,CHARPIAT G,et al.Can semantic labeling methods generalize to any city?the inria aerial image labeling benchmark[C]//IEEE international geoscience and remote sensing symposium.[s.l.]:IEEE,2017:3226-3229.
[16] 刘丹,刘学军,王美珍.一种多尺度CNN的图像语义分割算法[J].遥感信息,2017,32(1):57-64.
[2] 巫兆聪,胡忠文,张谦,等.结合光谱、纹理与形状结构信息的遥感影像分割方法[J].测绘学报,2013,42(1):44-50.
[3] 莫玉.高光谱图像特征学习与分类算法研究[D].西安:西安电子科技大学,2015.
[4] 王巧玉.基于深度学习的高光谱遥感图像分类[D].泉州:华侨大学,2016.
[5] 魏德强.高分辨率遥感影像建筑物提取技术研究[D].郑州:解放军信息工程大学,2013.
[6] 付卓,胡吉平,谭衢霖,等.遥感应用分析中影像分割方法[J].遥感技术与应用,2006,21(5):456-462.
[7] 代具亭,汤心溢,刘鹏.基于深度学习的语义分割网络[J].红外,2018,39(4):33-38.
[8] LONG J,SHELHAMER E,DARRELL T.Fully convolutional networks for semantic segmentation[C]//IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2015:3431-3440.
[9] ZHENG S,JAYASUMANA S,ROMERA-PAREDES B,et al.Conditional random fields as recurrent neural networks[C]//IEEE international conference on computer vision.Santiago,Chile:IEEE,2016:1529-1537.
[10] BADRINARAYANAN V,KENDALL A,CIPOLLA R.SegNet:a deep convolutional encoder-decoder architecture for scene segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2017,39(12):2481-2495.
[11] RONNEBERGER O,FISCHER P,BROX T.U-Net:convolutional networks for biomedical image segmentation[C]//International conference on medical image computing and computer-assisted intervention.Cham:Springer,2015:234-241.
[12] CHEN L C,PAPANDREOU G,KOKKINOS I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2018,40(4):834-848.
[13] HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al.Deep residual learning for image recognition[C]//IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2016:770-778.
[14] WANG S,FIDLER S,URTASUN R.Holistic 3D scene understanding from a single geo-tagged image[C]//Computer vision and pattern recognition.Boston,MA,USA:IEEE,2015:3964-3972.
[15] MAGGIORI E,TARABALKA Y,CHARPIAT G,et al.Can semantic labeling methods generalize to any city?the inria aerial image labeling benchmark[C]//IEEE international geoscience and remote sensing symposium.[s.l.]:IEEE,2017:3226-3229.
[16] 刘丹,刘学军,王美珍.一种多尺度CNN的图像语义分割算法[J].遥感信息,2017,32(1):57-64.