基于深度协同稀疏编码网络的海洋浮筏SAR图像目标识别

发布时间：2018-03-21 01:00

本文选题：合成孔径雷达　切入点：深度学习　出处：《自动化学报》2016年04期 　论文类型：期刊论文

【摘要】：浮筏养殖广泛存在于我国近海海域,可见光遥感图像无法完全准确地获取养殖目标,而基于主动成像的合成孔径雷达(Synthetic aperture radar,SAR)遥感图像能够得到养殖目标,因此采用SAR图像进行海洋浮筏养殖目标识别.然而,海洋遥感SAR图像包含大量相干斑噪声,并且SAR图像特征单一,使得目标识别难度较大.为解决这些问题,提出一种深度协同稀疏编码网络(Deep collaborative sparse coding network,DCSCN)进行海洋浮筏识别.本文方法对预处理后的图像先提取纹理特征和轮廓特征,再进行超像素分割并将同一个超像素块特征组输入该网络进行协同表示,最后得到有效特征并分类识别.通过人工SAR图像和北戴河海域浮筏养殖SAR图像的实验验证所提模型的有效性.该网络不仅具有优异的特征表示能力,能够获得更适合分类器的特征,而且通过近邻协同约束,有效抑制相干斑噪声影响,所以提高了SAR图像目标识别精度.
[Abstract]:Floating raft culture is widely used in offshore waters of China. The target of culture can not be obtained by visible light remote sensing image, but the target can be obtained by active imaging of synthetic aperture radar synthetic aperture radar radar (SAR) remote sensing image. Therefore, the SAR image is used to recognize the marine floating raft culture target. However, the marine remote sensing SAR image contains a lot of speckle noise, and the feature of SAR image is single, which makes it difficult to recognize the target. In order to solve these problems, A deep cooperative sparse coding network (Deep collaborative sparse coding Network) is proposed for ocean raft recognition. In this paper, texture features and contour features are extracted from preprocessed images. Then the hyperpixel segmentation is carried out and the same super-pixel block feature group is input into the network for cooperative representation. Finally, effective features are obtained and classified and recognized. The effectiveness of the proposed model is verified by experiments of artificial SAR images and SAR images of floating raft cultivation in Beidaihe Sea area. The features of the classifier can be obtained more suitable for the classifier, and the speckle noise can be suppressed effectively by using the nearest neighbor co-constraint, so the target recognition accuracy of SAR image is improved.
【作者单位】：大连理工大学电子信息与电气工程学部;国家海洋环境监测中心;
【基金】：国家自然科学基金(61273307,61301130) 中国博士后面上基金(2014M551082) 北戴河邻近海域典型生态灾害与污染监控海洋公益专项(201305003) 海域使用动态监测和污染监测研究专项资助~~
【分类号】：TN957.52

【相似文献】