Crack extraction using concurrent Delaunay triangle mesh method
In response to the difficult problem of wall crack detection,which is one of the important tasks in building restoration,previous crack extraction techniques based on point cloud edge detection were severely affected by variable thresholds and crack morphology.This paper proposes a wall crack detection method that combines the geometric and two-dimensional distribution characteristics of crack point clouds,and combines the Delaunay triangular mesh of shared vertices with adjacent abnormal points for secondary judgment.Firstly,point cloud data dimensionality reduction is achieved based on plane fitting and 3D coordinate transformation;Then,the Delaunay triangle mesh quality features are used to exclude the grid at the crack location,and combined with the geometric and distribution characteristics of the point cloud,a secondary judgment of the inner and outer abnormal points is achieved;Finally,precise screening of crack areas is achieved through density clustering,and the edge points of cracks are restored to the three-dimensional space to extract the geometric features of cracks.Experimental verification and analysis were conducted using laser point cloud data on building walls.The results showed that the recall and accuracy of crack detection on the measured walls reached 100％.Compared with the manually extracted results,the maximum relative deviation of the geometric features of cracks was-9.7％.This method can provide technical support for large-scale wall damage detection in buildings.
remote sensinglaser point cloudscrack extractionDelaunay gridpoint cloud feature extraction
NETL
NSTL
万方数据
