Defect Detection Of Diode Packaging Bracket Based On Machine Vision

At present,China's LED industry is in the stage of rapid development,diode packaging bracket as the core components of diode devices,the diode service life,performance and reliability have an important impact,so packaging bracket before the factory needs to be strictly tested.However,in the process of continuous upgrading of packaging stent production process,there are many problems such as immature technology,high defect rate,and most of the domestic stent manufacturers are still manual sampling,low detection efficiency also limits the further development of packaging stent industry.Therefore,it is necessary to develop an economical,efficient and reliable defect detection method.This article will machine vision detection technology applied to the diode encapsulation stents to defect detection,on the basis of the existing machine vision inspection technology,further several kinds of defects of diode encapsulation support(such as silver surface contamination,scratches and defects)for a particular study,a design flaw detection algorithm,implements the stent defects automatic detection and evaluation,and the feasibility and practicability of the algorithms is verified by experiment.In this paper,the surface defect detection of diode packaging bracket is studied as follows:(1)Image acquisition and image quality optimization.According to the manufacturing process of packaging support and the formation mechanism of surface defects,the classification standards and testing requirements of packaging support are clarified,and the research object of defect detection is further determined.By adjusting the height and Angle of the light source,the problem that the metal surface reflects light and is difficult to be imprinted is solved.Aiming at the problems of noise,brightness,darkness and Angle deviation in the process of image acquisition,this paper proposes a pre-processing algorithm that combines image denoising and image enhancement.Firstly,the defect image is de-noised and smoothed by adaptive median filter,then the image Angle is adjusted by imrotate function,and finally the image is processed by histogram equalization and Laplacian image sharpening to eliminate the edge blur and adjust the image brightness.Finally,the defect image with high contrast and clarity is obtained.(2)An algorithm for defect detection of specific scaffolds based on image features.Aiming at the problem that the image detection area is too large to locate the detection target,this paper designs an interactive ROI region algorithm,which reduces the workload of detection.On the basis of ROI region,this paper designs a morphological segmentation algorithm for pollution defects,which combines open operation with closed operation to segment the pollution defect region.The gray value difference detection algorithm is designed for the scratch defect,and the similarity matching is completed on the basis of the connected domain coordinates.A color image threshold segmentation method was designed for damaged defects.The threshold segmentation was completed by marking the ROI region.The experimental results showed that the average detection accuracy of the three defect algorithms reached 97.3% and the average time consumption was 3.22 s.(3)A variety of scaffold defect detection algorithms based on adaptive templates.According to the cycle characteristics of stent surface module arrangement,this paper design the adaptive template algorithm,first of all,through the analysis of the autocorrelation function and spectrum estimation,extraction of stent surface module of periodic signal,and then based on the cycle characteristics to obtain support unit building blocks,finally using bilinear interpolation method to generate support benchmark template matching,complete template matching and segment the defect area and the experimental results show that the algorithm of the average detection rate was 93.3%,the average consumption of 1.52 s,compared with a specific defect detection algorithm,detection accuracy rate was reduced by 4%,but the efficiency is improved by 52.7%.The comprehensive shape feature classification was used to carry out the defect identification experiment on 90 defect samples.The average recognition rate of the defect area obtained by the multi-defect detection algorithm was 89.9%.Based on this,it can be concluded that although both algorithms can meet the requirements of defect detection of EMC scaffolds,the results of multiple defect detection algorithms are better.The research work in this paper has certain reference value for defect detection of diode packaging bracket based on machine vision.