Research On Automatic Detection Technology Of IC Pin Welding Defect Based On Deep Learning

With the development of electronic manufacturing technology,Automated Optical Inspection(AOI)has been widely used for Surface Mount Technology(SMT)on circuit boards.At present,statistical modeling and template matching detection methods are basically adopted in the AOI equipment,which has high reliability and speed.However,statistical modeling needs to be redesigned when the product style is replaced.It is necessary to establish a template by performing statistical learning on the samples that are qualified for manual testing,and then do automatic detection.The method of establishing a template is time consuming and cumbersome,so AOI is not effective in small-volume and multi-variety electronic manufacturing enterprises.In particular,the detection of the solder quality of IC pins has been a technical difficulty in AOI.What's more,as the IC is more integrated,and the increase of IC fine pitch pins,the detection is more difficult.In this paper,a deep learning method is used to establish a deep convolutional network.By studying a large number of different IC pin soldering image samples.the pin soldering defect of different IC is tested,which establishes a general method.For different types of IC,only the pin position need to be judged to perform the welding defect inspection.when it avoids re-building the template after replacing the product model,which improves the production efficiency.The deep learning model can extract the classification features directly from the input image,and obtain high-level abstract features through layer-by-layer convolution.The method can find the common features of the solder joint defect images,and the soldering defect of different types of IC pins can be detected accurately and quickly.Finally,through the test of the IC installation image collected by the patch production line,it is shown that the deep learning method designed in this paper can effectively realize the IC pin soldering defect detection.In addition,the Mark point positioning,IC positioning and image processing problems in the AOI detection process were studied and innovative design was carried out.For the Mark point location problem of PCB,an optimized positioning algorithm based on sub-pixel edge point analysis,curve iterative fitting and minimizing the fitting deviation is designed,which further improves the positioning accuracy of Mark points and designs a contrast experiment.It shows that the algorithm has higher accuracy for the Mark point location of the PCB.In addition,the problem of positional positioning,model identification,and pin division of IC chips on the PCB have been studied.Combined with the characteristics of IC chips,this paper proposes a method for fast positioning of IC chips.The method completes the positioning of the IC chip through a fast Hough transform of the edge points of the IC chip plastic base by the improved Hough transform.Comparing the experimental results,the algorithm can improve the positioning speed while achieving the error range allowed by the industry.The method studied in this paper of IC soldering defect detection based on deep learning has a good application prospect for AOI detection of small batch and multi-model PCB production lines of electronic enterprises,which saves the multiple statistical modeling of multi-model components.