Research On Automatic Recognition Of Building Materials Based On 3D Laser Point Cloud

Under the background of the development of information technology in the construction industry,the automatic recognition technology of building materials has received the attention of the industry and academia for its role in promoting the construction and management informatization,and its application value has been further highlighted after the promotion and application of Building Information Modelling(BIM)technology.At present,the image-based automatic recognition method of building materials has occupied the mainstream.However,the accuracy of this method is easily affected by the environment lighting.In contrast,the method of automatic recognition of building materials based on 3D laser point cloud is not easy to be affected by the environment lighting,and 3D laser scanner can collect image data and laser point cloud data at the same time,with more information types.However,these methods are either based on the experimental environment test of control conditions,or only recognize one or two kinds of building materials in the real environment,which is difficult to meet the actual engineering needs of automatic identification of a variety of building materials in the real environment.Based on this,the purpose of this study is to develop a method of automatic recognition of a variety of building materials based on3 D laser point cloud.Firstly,this paper analyzes the current research situation,expounds the practical significance of building materials automatic recognition technology for engineering construction and management,determines the research methods,and expounds the basic concepts.Then through the theoretical analysis and verification test method,the method of automatic recognition of various building materials based on 3D laser point cloud is constructed from three aspects: data type,automatic recognition feature and automatic recognition algorithm.At the same time,the mathematical model of each recognition feature value calculated from the original 3D laser point cloud data is established.Then,this research takes the recognition accuracy as the evaluation standard of recognition model performance,collects 10 kinds of common building materials data from the real environment to test the method,and simulates the performance of the method in the actual application scene to the greatest extent.Finally,the test results are compared and analyzed from many angles,which proves the validity and feasibility of the method proposed in this study.The followings are the research conclusions.Firstly,for automatic recognizing multiple building materials,the recognition accuracy and stability achieved by using the material reflectance as the recognition feature is better than the laser reflected intensity of the material;meanwhile,the material reflectance can be calculated based on the 3D laser scan data.Secondly,using the surface roughness of the material as the recognition feature can improve the accuracy of building material recognition.Thirdly,the proposed approach of automatic recognition of building materials based on 3D laser point cloud achieves an accuracy of 96.7%;meanwhile,when the one-class classification algorithm is used for special cases,the average accuracy of the proposed approach is 86.9%.The accuracy results prove that the proposed approach has good robustness and feasibility,and it has high potential in engineering practical application...