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

Automatic Extraction Algorithm of Seismic Landslide Information Based on After-Calamity High-Resolution Remote Sensing Image

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。

Aiming at the high efficiency requirement of extraction seismic landslide, we propose an automatic extraction algorithm of seismic landslide based on after-calamity high-resolution remote sensing image. The algorithm utilizes the spectrum, shape, and texture features of seismic landslide of the remote sensing image to remove the disturbed features based on the analytic hierarchy process, and realizes the automatic extraction of the seismic landslide. All the feature parameters are computed automatically by the improved Otsu algorithm. In a test using high-resolution aerial remote sensing data acquired by ADS40 image after earthquake Wenehuan in 2008, the experiments demonstrate that more than 70% landslides are correctly detected by the proposed method. Furthermore, the area accuracy is more than 80% and the extraction time is less than one minute for the ADS40 image with 10000 rows and 10000 columns data. Compared with traditional visual interpretation of human-computer interaction, the proposed algorithm has high degree of automation and landslide extraction speed, and the landslide recognition accuracy can meet the earthquake disaster emergency requirements.