Research On Two-dimensional Measurement Method Of Low Contrast Object Based On Machine Vision

In order to meet the requirements of the "Made in China 2025" modern industry,the quality requirements for parts and components are getting higher and higher,and its high-precision and high-efficiency testing has become an engineering problem that needs to be solved urgently.Machine vision is widely used for dimensional measurement of components due to its stability,reliability,high precision,and low power consumption.However,in industrial component inspection,the contrast of the inspection target is usually low,making the measurement of the size of the low-contrast object a bottleneck for machine vision inspection.Therefore,under the premise of reasonable light source system design,this thesis studies the algorithm of two-dimensional image size measurement of low-contrast objects,which overcomes the difficulty of measuring the size of parts in low-contrast objects in machine vision.Firstly,the hardware of the image acquisition system was selected,the hard,ware platform of the whole system was built,and the illumination source was designed.Illumination light path is the key factor affecting image quality.Through ZEMAX simulation and experimental verification,the effects of different light source types,colors and illumination angles on the imaging quality of low-contrast objects are analyzed.On this basis,a scheme of low angle illumination with white LED ring light was determined.Then,an image measurenment algorithm for low-contrast objects is designed.According to the background noise generation,a bilateral filtering algorithm that takes into account both spatial information and range information is used,which takes into account noise removal and edge information retention.Aiming at the problcem of low contrast of objects,the Laplacian enhancement operator,Retincx enhancement algorithm and ACE enhancement algorithm are compared and analyzed.The conclusion that ACE algorithm is suitable for image enhancement under low contrast is obtained.On this basis,the Canny algorithm is first used to extract the contour edges,and then the morphological processing and the connected domain mark small area elements are used to make the edge contour complete and the edges occupy a pixel width,and then use the sub-pixel edge detection algorithm for precise positioning.At the same time,in the analysis of sub-pixel edge detection algorithm,the partial sub-pixel coordinates obtained by polynomial interpolation method which are compared with the coordinates obtained by Sobel algorithm.The results show that the coordinate precision obtained by polynomial interpolation method is higher.Finally,the low-contrast workpiece size is measured from an experimental perspective and the algorithm is verified.The CCD industrial camera used is calibrated with a standard calibration plate to obtain a calibration coefficient,and the actrual size of the mold is measured by a digital universal tool microscope,and the measured size is compared with the actual size,and the measurement error of the method is comprehensively analyzed.Three different shapes of the mold were dimensioned using the machine vision-based low-contrast object two-dimensional measurement system designed in this paper.The gauge block with a size of 20 mm was measured and the measurement result was(20.052±0.252)mm.Therefore,the measurement system of this paper can obtain higher measurement accuracy.