Defect Detection Of Ceramic Substrate Based On Photometric Stereo And Frequency Domain

Chip multilayer ceramic capacitors are common chip capacitors.As a semiconductor electronic component,they are widely used in electronic equipment such as automobiles and mobile phones.There will be indentations,scratches or lint on the surface of the ceramic substrate formed after lamination in the production process.Before cutting,the surface needs to be optically inspected,and the indentation and lint are detected and marked.Eliminate defective areas during cutting to improve product reliability and yield.At present,it is mainly relying on manual use of flashlights from different angles to find small indentations and flakes.There are problems of low detection efficiency,poor reliability,and high labor costs.It is urgent to design a CCD optical detection system to complete this task.Due to the high detection accuracy requirements and the difficulty in meeting the lighting effects of general light sources,there is no good solution in the industry.In this paper,photometric stereo method and frequency domain detection methods are carried out for the surface defects of ceramic substrates.The main work is as follows:For the defect detection on the surface of ceramic substrates,there are problems such as large field of view,small defects,and the need for specific direction of light to detect defects.This paper proposes a ceramic substrate defect detection method based on the photometric stereo method.Illuminated pictures,obtain the gradient map of the ceramic substrate surface,calculate the average curvature,Gaussian curvature,divergence and curl,and fuse them into a single-channel image,so that the indentation,scratches,dander and other defects on the surface of the ceramic substrate can be presented produce better results.Finally,it traverses all defect areas,filters out the areas that are too small,and judges whether it is an indentation or dander by the average gray value.If the average gray value is greater than the background gray value,it is an indentation,and vice versa.For the shallower defect information,the fused image is converted to frequency,and the low frequency part is removed,so that the finer defect information can be obtained.The fused image is subjected to fast Fourier transform and a self-designed high-pass filter in the frequency domain to filter out low-frequency components such as the background.Simultaneously perform operations such as gray value extraction and closed operation on the fused image and the filtered image,initially screen out the defect area and perform a merge operation,and then use the area and average gray value to screen out indentation and dander defects.The test results show that the automatic detection system for ceramic substrate surface defects designed by the photometric stereo method in this paper,combined with analysis methods such as gradient,curvature,and frequency domain,can accurately detect defects such as indentations and dander,and greatly improve the detection efficiency.The detection rate of defects above 0.05mm2 can reach 100%,and the detection time is within one second.