摘要
为了提高卫星遥感图像的识别与分类效果,提出一种基于卷积神经网络的卫星遥感图像识别与分类方法。该方法通过导向滤波去雾和旋转图像数据提高了模型的泛化能力,同时采用了双全连接层网络结构增强了模型数据表达能力。实验证明,该方法在卫星遥感图像的识别与分类上优于传统图像识别方法和一般卷积神经网络模型。
In order to improve the effect of remote satellite sensing images recognition and classification,a kind of satellite remote sensing images recognition and classification method is put forward based on convolutional neural network. It uses guided filter to dehaze and rotates the image to improve the generalization ability of the model,and uses double connection layer network structure to enhance the model data expression ability. The experiments and comparisons show that this method is much batter than traditional image recognition method and general convolutional neural network model on satellite remote sensing images recognition and classification.
出处
《信息技术》
2017年第11期83-86,共4页
Information Technology
关键词
卷积神经网络
深度学习
遥感图像
识别分类
convolutional neural network
deep learning
remote sensing images
recognition and classification
