Research On Algorithms For Measuring Geometrical Features Of Hole And Shaft Parts

Hole and shaft parts play important roles in the machinery industry.Their geometric features such as size and appearance is vital to functioning of the machine.The traditional measurement methods have problems such as low measurement accuracy and high cost of high-precision instruments.This thesis proposes algorithms for measuring the geometric characteristics of hole and shaft parts,and designs measurement schemes to improve efficiency and accuracy.Firstly,geometric characteristics of hole and shaft parts are analyzed.Two non-contact methods are proposed,including machine vision for the diameter of hole parts and laser ranging for the roundness and cylindricity of shaft parts to measure different elements.Then based on the measurement of hole parts,an algorithm for selecting pixel-level coarse-positioning target area first and then sub-pixel-level fine-positioning contour extraction is proposed.A method of Gradient-based Hough Circle Transform(GHCT)for coarsepositioning of contours,neighboring pixels expanding to mark the Region Of Interest(ROI),Zernike moments edge-positioning is proposed.The results reach the accuracy requirements.Then based on the measurement of shaft parts,an Improved Artificial Bee Colony algorithm(IABC)to evaluate the roundness is proposed.And combined with the idea of decomposition method,XGBoost regression algorithm(XGBR),which is based on multidimensional features,takes roundness,straightness and taper as the basic features to evaluate the cylindricity is proposed.The results reach the accuracy requirements.Finally,an experiment platform is built,and the results from the coordinate measuring machine are used to verify the algorithms.The performance of the algorithms proposed in this thesis are evaluated in three aspects,including numerical deviation,accuracy rate and measurement efficiency,and requirements are all met,and better than existed algorithms.