基于纹理特征的深度学习云和云阴影检测  

Deep learning based on texture features for cloud and cloud shadow detection

作　　者：张昊[1] 焦瑞莉[1] 乔聪聪 霍娟[2] 宗雪梅[2] ZHANG Hao;JIAO Rui-li;QIAO Cong-cong+;HUO Juan;ZONG Xue-mei(School of Information and Communication Engineering,Beijing Information Science and Technology University,Beijing 100101,China;Key Laboratory of Middle Atmosphere and Global Environment Observation,Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029,China)

机构地区：[1]北京信息科技大学信息与通信工程学院,北京100101 [2]中国科学院大气物理研究所中层大气和全球环境探测重点实验室,北京100029

出　　处：《计算机工程与设计》2024年第5期1580-1587,共8页Computer Engineering and Design

基　　金：国家自然科学基金重点基金项目(42030107);国家自然科学基金面上基金项目(42175150)。

摘　　要：针对云和云阴影检测过程中存在边界不准确以及易与地表混淆等问题,构建一种融合纹理特征模块的卷积神经网络模型对Landsat 8遥感图像进行云和云阴影检测。引入基于统计特性的纹理特征模块进行纹理特征的提取和学习,在训练过程采用焦点损失函数削弱样本不均衡带来的影响。实验结果表明,该模型细化了云和云阴影的边界等纹理细节,减少了云和云阴影的误检和漏检现象,提高了云和云阴影的检测精度。Aiming at the problems of inaccurate boundary and easiness to be confused with the surface in the process of cloud and cloud shadow detection,a convolutional neural network model incorporating texture feature module was constructed to detect clouds and cloud shadows in Landsat 8 remote sensing images.A texture feature module based on statistical properties was introduced for texture feature extraction and learning,and focal loss function was adopted in the training process to weaken the inf-luence of sample imbalance.Experimental results show that the proposed model refines the texture details such as boundaries of clouds and cloud shadows,reduces the false detection and missed detection of clouds and cloud shadows,and improves the detection accuracy of clouds and cloud shadows.

关 键 词：云检测 云阴影检测 统计特性 纹理特征 卷积神经网络 遥感图像 焦点损失函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]