顾及时空谱规律的城市水体时序变化检测方法  

Time Series Change Detection Method of Urban Water Bodies Using Spatial-temporal-spectral Regulations

作　　者：宋艳若 庄启智 陈辉 朱大明[1] 程亮[2,3,4,5] SONG Yanruo;ZHUANG Qizhi;CHEN Hui;ZHU Daming;CHENG Liang(School of Land and Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;School of Geography and Oceanographic Science,Nanjing University,Nanjing 210023,China;Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology,Nanjing University,Nanjing 210023,China;Collaborative Innovation Center of South China Sea Studies,Nanjing University,Nanjing 210023,China;Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources,Nanjing University,Nanjing 210023,China)

机构地区：[1]昆明理工大学国土资源工程学院,昆明650093 [2]南京大学地理与海洋科学学院,南京210023 [3]南京大学江苏省地理信息技术重点实验室,南京210023 [4]南京大学中国南海研究协同创新中心,南京210023 [5]南京大学自然资源部国土卫星遥感应用重点实验室,南京210023

出　　处：《遥感信息》2022年第4期66-72,共7页Remote Sensing Information

基　　金：国家重点研发计划项目(2017YFB0504205)。

摘　　要：遥感变化检测技术为动态监测城市水体变化和演化趋势提供了有效手段。针对传统遥感时序变化检测未能充分考虑像元间空间邻近特征等问题,提出一种顾及影像时空谱特征的城市水体时序变化检测方法。首先,依据影像光谱特性构建时间序列,对比傅里叶变换、小波变换、中值滤波、Savitzky-Golay滤波的重建效果,并评价优选;然后,依据变化像元时间维的特性,采用动态时间规整算法,衡量重建时间序列与样本时间序列相似度,确定疑似变化子序列;最后,结合空间邻域规则和归一化差值建筑用地指数进行空间相关性分析控制误报率,获取最终检测结果。利用Sentinel-2影像,在钱塘江研究区内开展实验,结果表明,该方法总体精度达到99%,虚检率降低了20%~50%,为城市水体变化检测提供了一种可行思路。Remote sensing change detection technology provides an effective means for dynamically monitoring the change and evolution trend of urban water bodies.In view of the problem that traditional remote sensing time series change detection fails to consider the spatial proximity features between pixels,this paper proposes a method for urban water bodies which considers the spatial-temporal-spectral regulations of remote sensing time series change detection.First,construct a remote time series based on the spectral characteristics of the image,compare the time series reconstruction effect of Fourier transform,wavelet transform,median filtering,and S-G filtering methods,evaluate and choose the best and choose the best.Then,DTW algorithm is used to calculate the time similarity between the reconstructed timing sequence and the sample timing sequence,and obtain the time series of suspected changing pixels.Finally,the NDBI index and the spatial neighborhood rules are fused to perform spatial correlation analysis,and control false positive rate to obtain the final change detection results.In this paper,we conduct experiments with the Sentinel-2 images on the study area of Qiantang river.The result shows that the overall accuracy of this method is 99%,and the false detection rate is reduced by 20%~50%,providing a feasible idea for urban water body change detection.

关 键 词：城市水体 变化检测 时间序列 时空谱规律 Sentinel-2数据 

分 类 号：P237[天文地球—摄影测量与遥感]