High-precision Thread Detection Based On Residual Cloud

In recent years,non-contact 3D technology has been used more and more frequently in industrial automation inspection due to its advantages of convenient deployment,no damage to workpieces,and low cost.As a common workpiece in industrial production,the high-precision measurement of threads has attracted more attention from scholars at home and abroad.However,in most cases,the actual production environment cannot meet the conditions of controllable light and low dust in the experimental environment,and only a single point cloud can be used for parameter measurement due to cost constraints.Therefore,it is extremely practical and necessary to study a high-precision detection method for threads based on a single point cloud in the production environment.This paper comprehensively considers the limitations and influences of the actual production environment on thread measurement,and collects and obtains a single incomplete thread point cloud data from different angles based on a non-contact structured light 3D point cloud scanner.Through research and analysis,a high-precision thread parameter measurement system for residual defect clouds is designed and implemented.The main innovations of this paper are as follows:First,for the reality that the residual cloud has many noises and holes.In this paper,a combined filtering method that combines multiple filters has been used to reduce data noise is obtained through research,and a hole completion method is proposed to repair the residual defect cloud vulnerability.Finally,a point cloud rectification method based on boundary contour is proposed,which can rectify a single point cloud without predicting the thread shape.Second,an automatic contour segmentation evaluation method is proposed for the straightened thread point cloud,which can obtain an ideal thread segmentation contour from the straightened point cloud.Next,this paper proposes a specific clustering method according to the continuity of the point cloud contour after segmentation,which is compatible with various point cloud data under the premise of ensuring the consistency of the clustering results.The above method makes preparations for the research on high-precision detection of thread parameters.Third,analyze the physical properties of each parameter of the thread respectively,and propose corresponding detection methods for thread taper,pitch,foot height,flank angle and diameter at any position.Unlike other point cloud studies that only focus on a specific parameter,these detection methods basically cover all the parameters commonly used in the thread detection process.The error source and universality of the detection system are analyzed and studied.Finally,a data actuarial method based on multi-dimensional and Kalman filtering to reduce random errors is proposed to further improve the detection accuracy.After completing the above work,this paper detects the point cloud data measured in the actual work scene and compares the detection results with the design accuracy and actual accuracy.The comparison results show that the average errors of the parameters measured in multiple experiments meet the actual requirements,and the measurement errors of other parameters except the flank angle are lower than the corresponding two-dimensional image detection methods.The above method provides a new method and overall idea for the high-precision 3D inverse measurement of the single residual defect cloud of various thread parameters.