摘要
针对高分辨率遥感影像信息丰富,地物变化复杂,导致变化检测结果精度较低问题。提出融合变化向量分析(CVA)与深度学习的高分辨率遥感影像变化检测方法。首先,采用CVA变化检测方法提取遥感影像上变化区域、非变化区域以及不确定区域;然后,利用改进的小型U型网络模型(Unet)进行遥感影像变化检测区域提取;最后,利用影像空间信息对提取的变化区域进行后处理,以减少漏检、虚检以及“椒盐噪声”影响。实验结果表明,该方法比仅使用小型Unet网络或CVA算法可更准确地检测出遥感影像中的变化地物。
High resolution remote sensing images are rich in information of complex changes,which is easy to get low accuracy of change detection results.A change detection method of integrating change vector analysis(CVA)and deep learning for high-resolution remote sensing image was proposed.Firstly,CVA change detection method was applied to extract the changing region,non changing region and uncertain region of image;Then,the improved small Unet network model was explored to extract the change detection area of remote sensing image;Finally,the extracted change region was processed by using image spatial information to reduce the influence of missed detection,false detection and"salt and pepper noise".The experimental results showed that this method could detect the changing features in remote sensing images more accurately than using Unet network or CVA only.
作者
张友桐
ZHANG Youtong(Guangdong United Gemdale Real Estate Evaluation and Surveying and Design Company Limited,Shaoguan Guangdong 512000,China)
出处
《北京测绘》
2022年第8期1079-1083,共5页
Beijing Surveying and Mapping
关键词
变化向量分析
小型Unet网络
空间信息约束
变化检测
change vector analysis(CVA)
small Unet network
spatial information constraints
change detection
