Concrete Surface 3D Point Cloud Preprocessing And Early-age Crack Extraction

With the increase of serving time,the concrete surface is prone to cracking,peeling and other damage.Regular detection of concrete surface is of great significance to evaluate the serving condition of the concrete pavement and the safety of concrete structure.The conventional detection method of concrete pavement and structure surface is manual detection,which has the problem of low efficiency.For the concrete pavement detection,there exist extra problems such as poor safety and traffic impact.The development of 3D laser scanning technology provides an efficient and noncontact method for concrete surface detection.By using the point coordinates obtained by 3D scanning,the high-precision 3D model of concrete surface can be reconstructed.The surface information will be than extracted,which will help analyze the service and damage condition of concrete surface and the maintenance scheme can be generated.The detection efficiency will be ensured as well as the accuracy.The preprocessing of concrete surface 3D point cloud is the basic operation to ensure the accuracy of reconstructed model,which mainly include three steps: denoising,smoothing and sampling.The present point cloud preprocessing method shows some limitations in these three steps.In this study,a preprocessing method aiming at 3D point cloud of concrete surface is proposed,which solves the existing problems in the denoising,smoothing and sampling steps of conventional methods.Firstly,the local density based denoising algorithm is used to deal with the scattered noise of point cloud,which solves the problem of poor adaptability or high complexity of conventional denoising methods.For the point cloud after denoising,a smoothing and sampling continuous algorithm based on the moving least square method is adopted.after distinguishing the road featured area and non-featured area,the problem of data repeated degradation caused by the split of smoothing and sampling steps is solved as well as the problems of over-smoothing or insufficient smoothing.The improved denoising,smoothing and sampling methods have been proved to have good environment adaptability by experiments.At the same time,the data obtained from the three-dimensional scanning has been optimized,forming a highly targeted preprocessing method system for the 3D point cloud of the concrete surface.The 3D point cloud of the concrete surface after preprocessing contains important information such as crack,peeling and flatness.Extracting and quantifying these information is the key step of concrete surface detection.Among them,the crack information,especially the early-age crack,is important to evaluate the durability and bearing capacity of concrete pavement and concrete structure.The existing research shows insufficiency in the effective observation and accurate extraction of concrete surface cracks,especially the early-age crack.Based on the preprocessed 3D point cloud,this study proposes a crack point extraction method based on local density.The extracted crack points are connected with each other according to the nearest neighbor principle to form crack elements.Secondly,by setting the threshold of angle and distance,the crack elements are clustered and expanded into independent crack and crack set.The cracks are quantified one and two dimensionally in terms of independent crack and crack set.Finally,the results of crack quantification based on 3D point cloud data are compared with those of crack viewer to verify the accuracy of crack extraction and quantification method.At the same time,the feasibility of two-dimensional crack quantification is verified by comparing the crack quantification results of different specimens with the actual crack distribution and shape.The results show that the crack extraction and quantification method based on 3D point cloud data provides a strong algorithm foundation for the intelligent observation of concrete surface cracks.