摘要
针对近海用海项目动态监测中复杂环境如潮汐、船舶等干扰造成的变化检测精度不足,提出了一种面向多时相遥感影像的高精度变化检测方法。该方法先融合CLAHE与高斯滤波以增强边缘、降低噪声;再结合NDWI与迭代阈值分割实现多时相协同边界提取;最后通过像素-对象级联合分析剔除伪变化,确定真实用海项目变化区域。在横滨港口Sentinel-2数据集测试中,低幅度变化场景IoU与F1较CVA方法分别提升15.91%和11.62%,高幅度场景提升11.14%和7.81%,显著优于主流方法。综上所述,该方法在准确性、适应性方面表现优异,可为智能化海域管理提供可靠技术支撑。
A time-series remote sensing image-based change detection method is put forward.Aimed at the problem of low accuracy to tide and ship interferences in complex offshore settings during offshore sea using project monitoring.Firstly,merging CLAHE and Gaussian filtering enhances edges and suppresses noise.Then,it integrates NDWI with iterative threshold segmentation to acquire multi-temporal collaborative boundary extraction.Finally,it removes pseudo-changes to identify real project change areas via pixel-object level joint analysis.In the Yokohama Port Sentinel-2 dataset test,compared to the CVA method,this approach yielded a 15.91%IoU increase and an 11.62%F1 boost for low-magnitude change scenarios,along with 11.14%and 7.81%improvements respectively for high-magnitude cases.In conclusion,this method excels in accuracy and adaptability,which can reliable technical support for intelligent marine area management.
作者
王振华
秦玮泷
张熙
宋刚成
隋家军
WANG Zhenhua;QIN Weilong;ZHANG Xi;SONG Gangcheng;SUI Jiajun(College of Information Technology,Shanghai Ocean University,Shanghai 201306,China;Shandong provincial institute of land space data and remote sensing technology,Jinan 250000,China;Shandong Yuanhong Survey,Planning and Design Co.,LTD,Jinan 250000,China;Shandong Jincheng Land and Real Estate Appraisal and Surveying Co.,LTD,Jinan 250000,China)
出处
《海洋测绘》
北大核心
2025年第6期68-72,共5页
Hydrographic Surveying and Charting
基金
上海市教委人工智能促进科研范式改革赋能学科跃升计划(A1-3405-25-000307)
关键词
遥感变化检测
机器学习
影像增强
迭代式阈值分割
多时相影像
remote sensing change detection
machine learning
image enhancement
iterative threshold segmentation
multi-temporal images
