Tree Defect Feature Reconstruction Based On TRU Detection Image Segmentation And Point Cloud Processing

In the modern field of nondestructive testing,when analyzing and observing the data fed back by various testing methods,human beings increasingly hope to see the same vision as their own Three dimensional representation,It promotes the continuous development of 3D application technologies.The technical difficulty is to realize the mapping of 2D data to The three dimensional space,among which the acquisition and reconstruction of point cloud data with 3D spatial information is one of the key technologies.In the application of forestry engineering automation,monitoring and diagnosis of tree growth status,quality pre-inspection of wood processing pr,non-destructive testing technology of trees and three-dimensional reconstruction technology of tree morphology are the optimal methods for efficient utilization,and organic combination point cloud data processing has become an important research content in this field.Trunk surface morphology characteristics of the growth of digital expression,trunk grows inside the empty area of corrosion defects detection and recognition,is the use of all kinds of test medium for detection of feedback signal,feedback signal expressed as graphic and image,or expressed as statistical data,and then empty decay ratio inversion calculated or digital simulation.Therefore,3D reconstruction of tree(NDT)data mainly involves three aspects.Firstly,nondestructive testing signal recognition and inversion,image segmentation after inversion imaging,that is,how to invert the detected signal into two-dimensional data and extract image boundary features.Second,the 3D scanning technology in reverse engineering is used to obtain the point cloud of tree appearance feature,and the 3D reconstruction of tree appearance feature surface is carried out by the processing technology of point cloud data.Thirdly,the data analysis of air decay area detection is integrated with point cloud processing technology.In the detection process,the growth morphology features are mapped to 3D point cloud to realize automatic reverse modeling.The subject has carried out the following relevant researches on the above issues:(1)on the radar wave nondestructive testing(TRU)technique to obtain trees standing timber internal decay empty feedback image image segmentation,threshold summation method is put forward implementation decay empty area feature boundary recognition and extraction,complete calibration decay empty area center,according to the corrosion trend boundary contour convex hull construction of voxel method refactoring decay empty area 3D model,The three-dimensional point cloud reconstruction of void decay area is realized by using the triangular partition method,which not only realizes the mapping of two-dimensional detection data of void decay to three-dimensional space.Improve the data analysis ability of nondestructive testing of living wood radar wave.(2)The point cloud data of wood surface topography was accurately obtained by using the double cycle weighted collision avoidance splicing scanning based on the information compensation method of the mark points;The optimal side length enveloping box sampling filtering method of point cloud data is proposed,and the FPFH method of distance weight is proposed to realize the consistency feature estimation correction of point cloud data.Programming using photo library functions to improve the sampling uniformity SAC-IA coarse registration process and the essence of the ICP registration algorithm,namely dynamic point to the center of gravity of the cloud,effectively solve the boundary point cloud recognition problem of local optimal solution and registration,and use the streamlined boundary point cloud data with the method of feature extraction and matching to achieve 3D reconstruction trees morphology characteristics,Improve the point cloud data processing ability of tree growth morphology inside and outside.(3)The comparison test of TRU detection feedback data 3D reconstruction for multiple sample groups was designed and completed,and the comparison test of point cloud data simplification and registration was completed.The point cloud simplification ratio was 34.52%when the feature retention was optimal.The algorithm error and registration deviation were comprehensively analyzed.The maximum deviation of registration fitting is about 34% to 40%lower than the horizontal data of relevant studies.The calculation and application of void rot grade of 3D void rot reconstruction model are completed.The research results can be applied to other non-destructive testing data analysis fields,promote the transformation from two-dimensional feedback data form to three-dimensional space in the field of non-destructive testing,and promote the improvement of the application technology of pattern recognition technology and point cloud processing technology in the field of tree morphology reconstruction.