Key Technologies In Vision Defect Inspection Of High Density Flexible Integrated Circuit Substrate

Flexible Integrated Circuit Substrates(FICS)are the carries to connect electronic components.With the miniaturization of electronic equipments,FICS trends to microminiaturization,while the line width of FICS has been reaching to the level of micron.The quality detection of FICS is a key part during the manufacturing process;however,the basic detection is carried out by using artificial visual inspection at present.The results of artificial visual inspection are greatly influenced by the mental state of detectors and so are uncertainty.Moreover,artificial visual inspection is going to be more and more difficult and inefficient when the key parameters of FICS become smaller.Therefore,the techniques of using image processing technique to defect FICS image's detection automatically become the main effective way when FICSs' parameters are going microminiaturization.In this thesis,a high precision image detection platform is developed and several key problems in automatic visual inspection are focused on.The main aspects of the research are as follows:(1)To automatic detect of high density flexible substrates,a high precision image acquisition platform and an efficient software system framework which makes the multiple function run efficiently are developed.(2)To quickly locate the current detect position on a FICS,an image locating method combining deep convolution network with speeded-up robust features registering is proposed.The method can quickly and accurately locate the FICS's imge to the position in standard image.It avoids placing the flexible substrate and identifying MARK points by artificial and is helpful to achieve a high level of automation.(3)A high accuracy edge detetion algorithm is designed based on muti-scale space.First,anchors of image's edges in high scale are extracted;second,according to the extracted anchors,high precision edges are then searched for under the fusing image with multi-scale gradient intensities.(4)A progressive fitting line method is developed based on high presion edges.It not only can fit lines,get straight angles,the beginning and ending coordinates,but also can obtain the uneven regional locations on lines,which is convenience of measuring width and detecting open and short circuits.(5)Combining with image morphology processing and weighted mean idea,a method is proposed to detect circular aperture in real time.By using morphology processing,round holes can be quickly segmented,then combining with weighted mean idea,the aperture center coordinates and radius of the round holes are precisely calculated.Moreover,an evaluation method to measuring the roundness of a hole is also proposed.The research about the designs,methods and key technologies of detection system for high-density ultra-thin flexible substrate manufacturing process provides new knowledge in electronic industry.