衡艺婷

摘 要：本文以河南省北部洪涝灾害为例，结合洪灾前后多期光学和雷达卫星，分析阈值法和机器模型迁移法对不同传感器影像的水体识别能力，并分析研究区的洪涝灾害情况。重点分析机器学习模型在时间序列影像上的水体迁移识别的技术，讨论不同特征对水体的识别能力，并对分类结果进行精度检验和评价。研究结果表明，基于阈值法的光学影像水体提取受到云阴影的影响，而雷达影像水体提取受到部分道路的干扰。基于机器学习模型迁移的水体识别提取方法能够快速、准确提取系列光学和雷达的洪灾影像。本次洪灾持续时间近两个月，淹没面积最大值达到320km2。基于机器学习模型迁移开展多源光学雷达卫星影像水体快速识别，对于洪涝灾害范围监测具有重要的意义。

关键词：洪涝灾害，水体识别，机器学习，模型迁移，多源遥感

Automatic Extraction and Analysis of Flood Area in Henan Province Based on Multi-Source Remote Sensing Satellite

HENG Yiting

（Liaoning Normal University， Dalian， Liaoning Province， 116000 China）

Abstract： In this paper， it takes the flood disaster in the north of Henan Province as an example， combined with the multi-scene optical and radar satellites before and after the flood， the water extraction ability of the threshold method and the machine model transfer method for different sensor images as well as the flood disaster situation is analyzed. This paper focuses on the technology of water transfer classification based on machine learning model in time series image， discusses the recognition ability of different features to water， and verifies and evaluates the accuracy of classification results. The results show that the water extraction from optical image based on threshold method is affected by cloud shadow， while from radar image is interfered by some roads. The water extraction method based on machine learning model transfer can extract a series of optical and radar flood images quickly and accurately. The flood lasted nearly two months and submerged an area of 320 km2 at its maximum. Rapid water extraction based on multi-source radar satellite image based on machine learning model transfer is of great significance to flood disaster area monitoring.

Key Words： Flood disaster; Water extraction; Machine learning; Model transfer; Multi-source remote sensing

洪涝灾害是世界上主要自然灾害之一，遥感是监测洪灾的重要手段，洪涝灾害的遥感监测技术关键在于水体信息提取和识别。

已有水体提取方法主要基于阈值法和分类法。阈值法构建特征直方圖，或直接使用类间最大方差法，找到区分水体和背景的最佳阈值分割水体范围。王帆等[1]提出了基于红绿和近红外波段组合的改进水体指数NNDWI，该指数能较好地区分水体、云及云的阴影，得到去云的水体范围。而光学影像常受到云雨影响难以提取完整水体，雷达影像则具有穿云透雨的能力。……