摘要
由地震诱发的滑坡灾害具有数量多、分布面积广、危害大的特点。传统的高精度的专家手动解译不能满足大范围发生的滑坡灾后快速评估的要求,基于计算机的滑坡识别技术具有快速、准确的优势。随着灾后影像数据分辨率的不断提高,面向对象技术比基于像素的技术更具优势。但是在滑坡对象分割、分类中对影像光谱、纹理信息的过度依赖造成了分割不够客观、分类算法复杂等问题。基于震后Worldview 2m分辨率的多光谱数据,提出了一种考虑地形因子的滑坡分割、分类算法。这种算法基于简单的规则集并充分考虑了滑坡发生的机理特点进行分割、分类,具有快速、简便的特点,而且能达到很高的分类精度(98.14%)。这种方法较以往仅考虑影像光谱、纹理信息的面向对象滑坡识别技术具有很大的提高,可以在实际灾害发生时用于大面积、大数量的山地滑坡灾害的快速损失评估。
Earthquake can induce huge amount of landslides causing great human and property losses. Although manual interpretation of landslides can achieve high accuracy, landslide identification based on computer algorithm has the advantage of fast speed and objective with accurate landslide boundaries and is more practical to map land- slides induced by catastrophic earthquakes. With the advent of very high resolution satellite remote sensing, object- oriented image analysis is more and more popular than pixel-based methods. At present, image segmentations in landslide identification methods only rely on spectral and textural information. Such kind of image segmentation has problems in finding suitable segment scales and quick classify algorithms. By considering terrain factors that always limit landslide activities, this work proposed a new method to segment landslide objects more accurately. Based on this segmentation, a simple algorithm based on elevation difference within a single object was used to extract land- slides. The proposed method is easy to use, requires little computing capacity and can achieve very high accuracy, and therefore can be used to extract large numbers of landslides quickly.
出处
《自然灾害学报》
CSCD
北大核心
2015年第4期1-6,共6页
Journal of Natural Disasters
基金
国家重点基础研究发展计划(973)项目(2012CB955403
2012CB955404)
国家自然科学基金创新研究群体项目(413221001)
关键词
面向对象
基于地形因子的分割
分类方法
简易规则集
object-oriented
terrain factors-based segmentation and classification method
simple rule-sets
