基于显著性特征的岩质边坡缺失岩块搜索方法

A Search Method for Missing Clasts on Rocky Slopes Based on Saliency Features

边坡上滑动后静止的岩块易再次诱发落石灾害.针对此,提出了一种基于显著性特征的缺失岩块搜索方法,该方法首先使用Background Matting算法对不同时刻的边坡图像进行差异分析,以检测由于岩块缺失所引起的图像变化区域;然后,采用YOLOv5s算法对变化区域中的岩块进行识别,确定缺失岩块的位置;接着,运用U2net网络剔除岩块背景噪声,并运用SURF算法提取岩块显著性特征(纹理、轮廓、面积);随后,使用Sgmnet网络进行特征匹配,根据纹理、轮廓和面积三个方面计算岩块间的相似度;最后,根据相似度大小确定移动岩块间的关联关系.结果表明,基于显著性特征的缺失岩块搜索方法能够准确查找缺失岩块,准确率为85%,弥补了现有边坡落石监测方法的不足.

The stationary rock mass after sliding on the slope is easy to induce rockfall disaster again.To address this,a missing rock block search method based on saliency features is proposed.This method firstly uses Background Matting algorithm to analyze the difference of slope images at different moments in order to detect the image change region caused by missing rock blocks.Then,YOLOv5s algorithm is used to identify the rock blocks in the change region and determine the location of the missing rock blocks.Next,the U2net network is used to remove the background noise of the rock blocks,and the SURF algorithm is used to extract the salient features(texture,contour,area)of the rock blocks.Subsequently,the Sgmnet network is used to match the features,and the similarity between the rock blocks is calculated based on the three aspects of texture,contour and area.Finally,the correlation between the moving rock blocks is determined based on the size of the similarity.The experimental results show that the missing rock block search method based on saliency features can accurately find missing rock blocks with an accuracy rate of 85%,which makes up for the shortcomings of the existing slope rockfall monitoring methods.