Research On Measurement Method Of Crankshaft Geometric Parameters Based On Machine Vision

Crankshaft is one of the key parts of automobile engine and its quality and performance are of great significance to ensure the normal and reliable operation of engine.However,the crankshaft manufacturing and processing process is complicated,in the actual production needs to accurately measure and control the error of multiple sizes and geometric shapes,so the research of crankshaft geometric parameters detection technology is very important.At present,the crankshaft measurement in China is generally low efficiency,poor accuracy and other problems,this paper takes the crankshaft as the measurement object,based on machine vision image processing key technologies and geometric parameter measurement and evaluation methods to carry out an in-depth study,to achieve the rapid,non-contact measurement of the geometric parameters of the crankshaft.Firstly,a machine vision measurement system composed of lighting module,image acquisition module,motion control module and computer is designed.The optical and control hardware are selected,and the corresponding selection methods are described.The motion control module drives the rotation of the crankshaft and the 3D movement of the camera to realize the functions of focusing test and rotating shooting.The modular design system processes and analyzes the software and writes the geometric measurement and evaluation algorithm for image processing.Secondly,the key algorithms of autofocus and image processing are explored.By comparing the existing evaluation functions and improving the traditional focusing methods,a two-step focusing algorithm based on sharpness evaluation function is proposed.The checkerboard calibration method was used to calibrate the camera,correct the perspective projection and lens distortion of the target image,and calculate the pixel equivalent.The Zernike-moment subpixel detection algorithm was improved,Sobel operator edge was taken as the baseline,and the three-gray edge transition model was used to improve the measurement accuracy.The Otsu method was used to analyze the gray value change trend in the transition region to determine the optimal threshold,and experiments proved that the subpixel detection algorithm in this paper had higher accuracy.Thirdly,the imaging model is analyzed and the geometric error measurement and evaluation algorithm is studied.The measurement model of crankshaft geometric parameters is established and the visual measurement scheme of axle diameter based on geometric analysis is proposed.The improved PSO algorithm was used to determine the center of the circle in the minimum inclusion region,which improved the accuracy and efficiency of roundness error evaluation.The upper and lower boundary of the inclusion region of straightness error is determined by geometric iterative search method,and the error evaluation efficiency is better than that of traditional genetic algorithm and particle swarm optimization algorithm.Cylindricity error was evaluated based on least square method,and the global optimal solution of objective function was obtained by genetic algorithm.Finally,a visual measuring platform was built to evaluate the crankshaft and the results were compared with the coordinate measuring machine.The accuracy and repeatability of the visual measuring system were analyzed.The errors of axial diameter,roundness,straightness and cylindricity of visual measurement are 1.8 μm,4.48 μm,4.74 μm and 3.81 μm relative to the absolute errors of CMM.The experimental results show that the crankshaft geometric parameter measurement method based on machine vision is feasible in practice.