Fast And Precise Measurement Technology Of Metal Substrate Size Based On Machine Vision

The precision measurement technology of complex workpiece has always been a research hotspot in related fields.Although the manual measurement method represented by CMM has high measurement accuracy,it is slow in measurement speed and high in labor cost,which can not meet the needs of full inspection in mass production.Workpiece measurement technology based on machine vision,because of its high precision,high speed,non-contact nondestructive measurement and other advantages,has been widely concerned in the industry.Aiming at a kind of copper substrate used for chip positioning,installation and heat dissipation,based on the vision detection technology,the following specific research work has been carried out: the metal substrate has the characteristics of micro convex in the middle,slightly low in four weeks,with different sizes and positions of through holes,blind holes and grooves,which are commonly used as heat dissipation plates on the chip.At present,the factory mainly uses the method of CMM contact measurement to obtain the geometric dimension of the circular hole on the copper substrate.Although it can guarantee the measurement accuracy,it has high cost and long measurement period.The vernier caliper can also obtain the geometric dimension of a single circular hole quickly and directly,but the method is affected by human factors,and it can not guarantee the repeatability of the measurement accuracy.Therefore,the purpose of this paper is to measure the geometric dimensions of the circular holes on the copper substrate by non-contact measurement method based on machine vision according to image processing,expecting to achieve the goal of taking a shorter time,lower cost and ensuring the measurement accuracy within the allowable error range.1.Camera calibration and distortion correction.On the basis of Zhang Zhengyou's calibration method,the first three terms of the distortion coefficient are incorporated into the distortion model,the experimental system is calibrated by the black-and-white checkerboard calibration board,and the distortion coefficient is optimized.The experimental results show that the distortion is caused by the superposition of radial distortion and tangential distortion,and the reprojection error is reduced from 0.22 pixels to 0.14 pixels after correction.2.hole size measurement based on vision.The three-dimensional topography of through hole and blind hole is measured by super depth of field microscope,and the effects of different light sources and lighting methods are compared according to their characteristics,the best imaging methods of blind hole and through hole are obtained respectively,and the size is extracted by traditional image processing algorithm.the reason of large measurement error is analyzed by simulation.3.Polarization imaging measurement technology.A time-sharing polarization imaging method is proposed to calculate the polarization information according to the Stockes vector to solve the problem of high reflection on the surface of the copper substrate and the bottom of the blind hole.In order to further improve the image quality,the weighted average method,PCA method,HIS method and the combination of PCA and HIS are used to fuse the polarization image with the visible image,and the quality of the fused image is quantified in turn from the four dimensions of luminance mean,information entropy,standard deviation and sharpness.Through the above fusion method,the average brightness of the image is increased by 18.7%,the standard deviation is increased by 23.2%,the clarity is increased by27.8%,and the information entropy is increased by 16.5%.The results show that this method is better than the traditional pixel-level image fusion method.4.Research on image processing algorithm.In order to extract the edge of blind hole more accurately and further improve the measurement accuracy,this paper compares the advantages and disadvantages of edge detection operator,and improves canny operator from three aspects: filtering method,gradient calculation and optimal threshold,on the basis of wavelet transform,the method of multi-scale decomposition and reconstruction combined with improved morphology is used to obtain accurate blind hole edge,and the best wavelet base and multi-scale size are determined.The low-frequency component and high-frequency component obtained by multi-scale decomposition are reconstructed,and the multi-scale inverse transform is carried out according to the reconstruction coefficient.Through the improved canny operator,interpolation,this paper and the improved morphology combined with multi-scale decomposition and reconstruction,the edge of the copper substrate image is extracted,and the diameter size of the circular hole is obtained.The diameter measurement errors of through hole and blind hole are less than 0.03 mm and 0.04 mm respectively,which meet the measurement error requirements of 0.05 mm required by enterprises.