Detection Method Of Normal Feature Of Exterior Surface Of Cross Through Structure Based On Visual Point Cloud

Taking the full-volume ultrasonic inspection of naval and army equipment as the research background,due to the changeable curvature of the surface of the tested workpiece during the inspection process,it is difficult to find the normal direction of the curved surface to incident ultrasonic waves.the incident position to obtain a good echo signal.This study proposes a free-form surface feature detection method based on visual point cloud technology.The visual point cloud technology is used to complete the normal feature detection of the workpiece,which improves the efficiency of ultrasonic detection and has good practical value.The specific research work is as follows:First,the composition of the normal feature detection system of free-form surface is introduced.Starting from the working principle of the normal feature detection system that composes the free-form surface based on visual point cloud,on the premise of expounding the working mechanism of the detection system,the main technological process of the system is studied and analyzed;on the premise of adapting to the experiment and actual needs Next,the design of the system is carried out,the hardware equipment and software development system that constitute the system are selected,and the basic parameters of the system hardware and the development environment in which the software is located are explained respectively.Secondly,the parametric model of the surface normal feature detection system is constructed.Investigate the camera imaging process,build a 3D geometric model,and mark the point cloud camera with the help of Zhang Zhengyou’s calibration method;perform handeye calibration on industrial robots and point cloud cameras,select the eye-on-hand system,and study the camera coordinate system and robot base coordinate system II The conversion relationship between them is obtained,and the corresponding conversion matrix is obtained.The design experiment verifies the positioning accuracy of the hand-eye system through a large amount of experimental data;the maximum deviation of the positioning results is0.266 mm,and the average value is 0.107 mm.The results are relatively accurate and meet the actual use requirements.Third,the point cloud data is preprocessed,and different filtering methods are used for different types of noise.Multiple sets of comparative experiments are carried out for each filtering method,different thresholds are set,and the effect of filtering processing and the running time of the program are compared.The optimal filtering parameter range is selected,and the noise irrelevant to the target surface in the point cloud data is removed.After the preprocessing is completed,the obtained point cloud data is divided into meshes and normals are calculated.The idea and algorithm principle of mesh division adopted are introduced,and the principle and method of calculating surface normals are introduced.The normal feature detection is done by means of mouse interaction.Finally,the normal feature detection experiment is completed to verify the results of the surface normal feature detection system.The normal feature of the outer surface of the cross-through structure of the artillery cradle is detected on the built experimental platform.After many experiments,the experimental data is analyzed in detail.The average error between the theoretical coordinates and the real coordinates detected by the point cloud normal is the Euclidean The distance is 0.375 mm,and the average error of the normal line is 0.361°.It is concluded that the system accuracy meets the requirements for use.