Research On Feature Extraction And 3D Reconstruction Algorithm Of Panel Furniture Based On Laser Sensor

The spraying operation of furniture is an important part of the whole furniture manufacturing industry.The main production methods at this stage are still carried out by a large number of manuals,which are not only inefficient but also harmful to workers.The introduction of industrial spray robots has liberated labor to some extent.However,the working ability of the painting robot at this stage depends on the level of manual teaching,and it is mainly applied to single variety of large quantities of spraying,and the level of flexibility of production is not high.Parallel spraying operations for multi-variety furniture are more difficult to meet production needs.This paper focuses on the flexible automatic spraying process of small panel furniture,the acquisition and preprocessing of furniture point cloud model,multi-view point cloud data registration,furniture surface feature extraction and rapid 3D reconstruction.First of all,the measurement and collection method of point cloud on furniture surface is determined based on the actual production demand.According to the actual measurement platform,the measurement principle is analyzed and the corresponding coordinate system is established.The relationship between the coordinate systems is deduced and the surface point cloud data of the panel furniture is obtained.For the measured data,based on the characteristics of different types of noise point data,statistical methods are used to eliminate outliers.The weighted covariance matrix is used to solve the point cloud normal information on the furniture surface,and the rolling correction of the normal was carried out based on the neighborhood gaussian mean value.On the basis of the modified normal,the surface small noise points are smoothed with the neighborhood distance,so as to realize the filtering preprocessing of the original measurement data.Secondly,research on point cloud registration to solve the problem of registration and integration of real furniture data collected from multiple perspectives.The method of coarse registration and fine registration is adopted.The initial coarse registration is performed using key points and local FPFH features,and the feature matching of the registration process is optimized by RANSAC method.Based on the coarse registration results,the sampling and iterative process of the traditional ICP is optimized by combining sampling to improve the efficiency of the whole process.The effectiveness of the algorithm is verified by the actual acquisition of data and standard data.Thirdly,combined with the surface characteristics of the panel furniture and the subsequent requirements of the spray trajectory planning,the characteristics of different areas to be sprayed on the furniture surface are extracted.The RANSAC combined with the normal is used to extract the main surface features to be painted on the surface.The coplanar multi-plane and the different pattern features were clustered to obtain independent regions.Section features of different pattern areas are obtained by slice projection method.The edge features of plane and pattern area are extracted by angle threshold and distinguished by edge concavity.Finally,the research on point cloud reconstruction of furniture surface after multi-view registration is studied.The localized least squares is used to smooth the processed furniture point cloud.By downsampling,the number of reconstruction points is reduced and the reconstruction efficiency is improved.The point cloud on the furniture surface was quickly reconstructed by local projection combined with the greedy growth of the area.At the same time,constraints are set to screen the growth points to ensure the quality of the surface triangular mesh after reconstruction.Finally,the appropriate sampling proportion is determined through experiments,and the reconstruction effect is verified.