Leakage Detection Method Based on Laser Reflection Intensity Characteristics
In this paper,the sensitivity of LiDAR reflection intensity signal to leakage reaction is studied.The point cloud data of concrete is collected by LiDAR,and the leakage detection data set of concrete structure is constructed.In order to realize automatic detection and accurate segmentation of the leakage region,a dynamic graph convolutional neural network model based on additional intensity features is proposed,and an attention aggregation module is added for aggregation of coordinate features and intensity features.This model can realize the learning of difference features between point cloud intensity information.The experimental results show that the average crossover ratio and average accuracy of point cloud segmentation reach 93.5%and 98.0%,respectively.Compared with other similar algorithms,the proposed method shows a better segmentation detection effect.
NETL
NSTL
万方数据
