摘要
随着遥感技术的快速发展,基于遥感影像的变化检测被广泛应用于国土资源管理、地物变化、生态监测等众多领域,特征提取是变化检测中的关键,如何获取最优的特征提取策略是变化检测的难点。针对该问题,对目前常用的纹理特征、颜色特征、光谱特征、形状特征以及卷积神经网络(CNN)特征对高分辨遥感影像场景变化检测性能的影响进行对比分析,基于MtS-WH标准数据集的实验结果表明,CNN特征的变化检测性能最高,颜色特征适合于农田植被区域变化检测、纹理特征和形状特征对建筑物区域变化检测性能较高,而光谱特征适合地表区域的变化检测。
With the rapid development of remote sensing technology,change detection based on remote sensing images is widely used in many fields,such as land and resource management,ground object change,ecological monitoring,etc.Feature extraction is a key link in change detection,and how to obtain the optimal feature extraction strategy is a difficult point in current research.To solve this problem,the influence of commonly used texture features,color features,spectral features,shape features and convolutional neural network(CNN)features on the performance of high-resolution remote sensing image scene change detection is analyzed.The experimental results based on the MtS-WH standard data set show that the overall performance of CNN feature change detection is the highest,and the color feature has good performance on farmland and other vegetation change detection,texture features and shape features have better detection performance for building changes,while spectral features have better detection performance for ground changes.
作者
陆庆虾
LU Qingxia(Foshan Surveying and Mapping Geographic Information Research Institute,Foshan 528000,China)
出处
《测绘与空间地理信息》
2024年第10期79-82,共4页
Geomatics & Spatial Information Technology
关键词
高分辨遥感影像
变化检测
纹理特征
特征提取
卷积神经网络
high-resolution remote sensing images
change detection
texture feature
feature extraction
convolutional neural network
