灾后高分辨率遥感影像的地震型滑坡信息自动提取算法研究

发布时间：2018-05-18 05:03

本文选题：图像处理 + 高分辨率　；参考：《激光与光电子学进展》2017年11期

【摘要】：针对地震滑坡灾害应急响应的高时效性要求,提出了一种基于灾后高分辨率遥感影像的地震滑坡体自动提取算法。该算法综合利用了高分辨率遥感影像的光谱、形状和纹理等特征,基于多特征阈值分层次逐步剔除干扰地物,实现了地震滑坡体的自动提取。涉及到的特征参量阈值均采用改进的Otsu算法自动确定。在利用2008年汶川地震后ADS40航空遥感影像自动提取滑坡的实验中,所提算法的滑坡个数正确提取率超过70%,面积正确提取率超过80%。对于10000row×10000column的ADS40影像,算法执行时间低于1min。相较于传统的人机交互目视解译方法,该算法的自动化程度高、滑坡提取速度快,滑坡识别精度可以满足地震灾害应急要求。
[Abstract]:In view of the high timeliness requirement of the emergency response of earthquake landslides, an automatic extraction algorithm of seismic landslide body based on high resolution remote sensing image is proposed. The algorithm combines the characteristics of high resolution remote sensing images, such as spectrum, shape and texture, and gradually eliminates the disturbed ground objects based on the multi feature threshold value, and realizes the earthquake. Automatic extraction of landslide body. The threshold of characteristic parameters involved is automatically determined by improved Otsu algorithm. In the experiment of automatic extraction of landslides by ADS40 aerial remote sensing images after Wenchuan earthquake in 2008, the correct extraction rate of the number of landslides of the proposed algorithm is more than 70%, and the correct extraction rate of the area is more than 80%. for 10000row x 10000column ADS40 Image, the execution time of the algorithm is lower than the traditional method of human-computer interactive visual interpretation compared with 1min.. The algorithm has high automation, fast speed of landslide extraction, and the accuracy of landslide recognition can meet the requirements of earthquake disaster emergency.
【作者单位】：中国科学院遥感与数字地球研究所数字地球重点实验室;中国科学院大学;
【基金】：国家863计划(2015AA7026087) 高分辨率对地观测系统重大专项(民用部分)(30-Y20A37-9003-15/17) 中国地质调查局项目(12120113089200) 国家科技支撑计划(2015BAB05B05-02) 中国科学院遥感与数字地球研究所所长青年基金(Y6SJ1100CX) 国家自然科学基金(41501489)
【分类号】：TP751

【相似文献】