基于多尺度近端特征拼接网络的高光谱图像分类方法  

Hyperspectral image classification method based on multi-scale proximal feature concatenate network

作　　者：高红民[1] 曹雪莹 陈忠昊 花再军[1] 李臣明[1] 陈月 GAO Hongmin;CAO Xueying;CHEN Zhonghao;HUA Zaijun;LI Chenming;CHEN Yue(College of Computer and Information Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学计算机与信息学院,江苏南京211100

出　　处：《通信学报》2021年第2期92-102,共11页Journal on Communications

基　　金：国家自然科学基金资助项目(No.62071168);国家重点研发计划基金资助项目(No.2018YFC1508106);中央高校基本科研业务费资金资助项目(No.B200202183);江苏省研究生科研与实践创新计划基金资助项目(No.SJCX20_0181)。

摘　　要：针对基于传统卷积神经网络模型的高光谱图像分类算法细节表现力不强及网络结构过于复杂的问题,设计了一种基于多尺度近端特征拼接网络的高光谱图像分类方法。通过引入多尺度滤波器和空洞卷积,在保持模型轻量化的同时可以获取更丰富的空间−光谱判别特征,并提出利用卷积神经网络近端特征间的相互联系进一步增强细节表现力。在3个基准高光谱图像数据集上的实验结果表明,所提方法优于其他分类模型。Aiming at the phenomenon that the hyperspectral classification algorithm based on traditional CNN model was not expressive enough in detail and the network structure was too complex,a hyperspectral image classification method based on multi-scale proximal feature concatenate network(MPFCN)was designed.By introducing multi-scale filter and cavity convolution,the model could be kept light and the discriminative features of the space spectrum could be obtained,and the correlation between the proximal features of the CNN was proposed to further enhance the detail expression.Experimental results on three benchmark hyperspectral image data sets show that the proposed method is superior to other classification models.

关 键 词：卷积神经网络 高光谱图像分类 特征拼接 多尺度滤波器 空洞卷积 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]