基于全极化SAR影像的海岛地物分类

There are numerous islands with abundant resources in China.Due to the limited information included in common polarization features and the poor effect of traditional classification methods when there are few samples,nine polarization features are analyzed and classification is carried out using active deep learning.Firstly,multiple features are extracted from an original image.Then,the original features can be extracted by anto\|encoder and the initial classifier is trained and fine-tune the whole model with a small number of labeled samples.Finally,the most uncertain samples are selected to label with active learning algorithm and added to the training samples.Experiment comfirms that active deep learning can effectively improve the classification accuracy with less labeled samples and entropy shannon is a more effective feature to distinguish between seawater,mudflats and beaches.     

基于主动深度学习的高光谱影像分类

针对当前高光谱遥感影像分类人工标注样本费时费力,大量未标注样本未得到有效利用以及主要利用光谱信息而忽视空间信息等问题,提出了一种空-谱信息与主动深度学习相结合的高光谱影像分类方法。首先利用主成分分析对原始影像进行降维,在此基础上提取像素的一正方形小邻域作为该像素的空间信息并结合其原始光谱信息得到空谱特征。然后,通过稀疏自编码器得到原始数据的稀疏特征表达,并通过逐层无监督学习稀疏自编码器构建深度神经网络,输出原始数据的深度特征,将其连接到softmax分类器,利用少量标记样本以监督学习的方式完成模型的精调。最后,利用主动学习算法选择最不确定性样本对其进行标注,并加入至训练样本以提高分类器的分类效果。分别对PaviaU影像和PaviaC影像进行分类实验的结果表明,该方法在少量标记样本情况下,相对于传统方法能有效地提高分类精度。     

基于Cesium的地铁智慧化管理平台设计与实现

随着智慧城市建设的快速发展,地下空间管理业务对三维可视化提出了更高要求。当前,大多数城市地下空间信息存在立体界线不明确、权属不明晰等问题,传统二维GIS技术在地下空间信息可视化方面具有一定的局限性。文章基于开源的Cesium框架,以无锡地铁数据为基础,建设了无锡地铁三维智慧化管理平台,实现了三维展示、宗地确权展示、安防展示等功能模块,为实现地铁站地下空间信息化、精细化管理提供了平台支撑,为同类平台的搭建提供了一定的参考。     

基于全极化SAR影像的海岛地物分类

There are numerous islands with abundant resources in China.Due to the limited information included in common polarization features and the poor effect of traditional classification methods when there are few samples，nine polarization features are analyzed and classification is carried out using active deep learning.Firstly，multiple features are extracted from an original image.Then，the original features can be extracted by anto\|encoder and the initial classifier is trained and fine-tune the whole model with a small number of labeled samples.Finally，the most uncertain samples are selected to label with active learning algorithm and added to the training samples.Experiment comfirms that active deep learning can effectively improve the classification accuracy with less labeled samples and entropy shannon is a more effective feature to distinguish between seawater，mudflats and beaches.