基于高分遥感影像的多尺度纹理特征变化检测技术  

Multi-scale Texture Feature Change Detection Technology Based on High-Resolution Remote Sensing Images

作　　者：刘自增 张德政[2] 姚建华 严瑾 黄涛 Liu Zi-zeng;Zhang De-zheng;Yao Jian-hua;Yan Jin;Huang Tao(Remote Sensing Survey Institute of Ningxia Hui Autonomous Region,Yinchuan 750000,Ningxia Hui Autonomous Region,China;University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]宁夏回族自治区遥感调查院,宁夏银川750000 [2]北京科技大学,北京100083

出　　处：《科学与信息化》2022年第4期81-83,共3页Technology and Information

基　　金：宁夏回族自治区重点研发计划项目,项目名称:基于高分遥感的空间规划智能监测关键技术研究与示范应用,项目编号:2019BFG02009。

摘　　要：近年来,基于深度学习的图像处理方法取得了巨大的成功,越来越多的研究都尝试将深度学习算法应用于遥感影像的场景中。利用精度高、鲁棒性强的深度学习方法,可以为实现基于遥感卫星影像的区域自动检测识别工作提供技术保障。在这一背景下,本研究根据宁夏地区实际和项目需求制定了地物解译的方法,使用高精度的深度学习算法来实现地物的快速解译检测,并基于项目场景中样本场景复杂、标注困难的情况尝试了对高分遥感影像的多尺度纹理特征变化检测技术试验和检测算法多次改进。In recent years,image processing methods based on deep learning have achieved great success,and more and more researches have tried to apply deep learning algorithms to remote sensing image scenes.The application of deep learning methods with high accuracy and robustness can provide technical support for the realization of automatic detection and identification of regions based on remote sensing satellite images.In this context,this research formulates a method for interpreting ground objects according to the actual situation and project needs in Ningxia,and uses high-precision deep learning algorithms to achieve rapid interpretation and detection of ground objects.Based on the complex and difficult annotation of sample scenes in project scenes,the multi-scale texture feature change detection technology test and detection algorithm of high-resolution remote sensing images are tried for several times.

关 键 词：多尺度图像 遥感影像 纹理特征 地物分类 检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]