最优分割尺度的滑坡信息提取

针对面向对象的遥感影像分析技术这一热点问题,通过研究滑坡的地学特征和影像特征,基于QuickBird多光谱影像,提出了一种先对Quickbird影像进行多尺度分割,然后利用改进的分割质量评价函数来选择滑坡最优分割尺度,最后在最优分割尺度上构建滑坡提取规则集并进行滑坡信息提取的方法。该方法通过对提取的滑坡信息进行精度验证与分析,能使对象内部异质性和对象之间异质性达到综合效果最好。结果表明:滑坡体的正确提取率为75.86%,本文方法具有一定的可靠性和准确性。

Landslide information extraction based on the optimal segmentation scale

Aiming at object-oriented remote sensing image analysis technology,which is a hot issue,this paper,through study the geographical features and image features of landslide and based on multispectral QuickBird image,a multi-scale segmentation method is proposed,which was carried out on the image;second,using improved segmentation quality evaluate function to choose the optimal scale of landslide;then building the set of extracting rules at the optimal scale and extracting landslide.The method,through the validation and analysis of the accuracy of landslide extraction information,can achieve the best effect in heterogeneity within the objects and the heterogeneity between objects to achieve the best effect.The results show that:the correct rate of landslide extraction can reach 75.86％,extracting landslide with this method shows higher reliability and accuracy.