基于卷积神经网络模型的遥感图像分类

摘要

The remote sensing image classification was studied.In consideration of the problems of feature extraction difficulty and low classification accuracy of the shallow structure classification model of support vector machine, a convolutional neural network model was designed for remote sensing image classification.The model comprises the input layer, convolution layer, full connection layer and output layer, and uses the SoftMax classifier for classification.The LandsatTM5 remote sensing image of Fujin city in June 6, 2010 was used as the data source to perform the classification experiment.The experimental results show that the proposed model employs several convolutional and pooling layers to extract the nonlinear and invariant features from the remote sensing image.These features are useful for image classification and target detection.The classification accuracy of the model was 92.57% when it was used in this image.Compared to the support vector machine classifier, the classification accuracy of this model was improved by 5%.Therefore, this model has a greater advantage in remote sensing image classification.%研究了遥感图像的分类,针对遥感图像的支持向量机(SVM)等浅层结构分类模型特征提取困难、分类精度不理想等问题,设计了一种卷积神经网络(CNN)模型,该模型包含输入层、卷积层、全连接层以及输出层,采用SoftMax分类器进行分类.选取2010年6月6日LandsatTM5富锦市遥感图像为数据源进行了分类实验,实验表明该模型采用多层卷积池化层能够有效地提取非线性、不变的地物特征,有利于图像分类和目标检测.针对所选取的影像,该模型分类精度达到94.57%,比支持向量机分类精度提高了5%,在遥感图像分类中具有更大的优势.