Study On Wafer Identifier Recognition System Under Complex Background

With the aggravation of Sino US trade war,China's disadvantage in the field of semiconductor manufacturing has become increasingly prominent.Among them,the recognition of wafer identifier is accompanied by the whole semiconductor wafer manufacturing process,which is mainly used for wafer processing information tracking.However,there are few researches on wafer identifier recognition system in China.Moreover,in the field of semiconductor industry,foreign wafer readers are basically used,which is expensive,which greatly increases the development cost of related semiconductor equipment.Therefore,it is of great significance to develop a domestic wafer identifier recognition system.Due to the influence of wafer processing technology and lighting system of image acquisition device,the collected wafer identifier image has complex background interference such as occlusion,confusion,character micro deformation and nonlinear illumination.In order to read the wafer identifier in this complex background,this project is committed to develop a machine vision device which integrates the recognition algorithm of wafer identifier,combined light source system and human-computer interaction control system.On the one hand,starting from the analysis of the causes of complex background interference,the combined lighting system of three-color coaxial light source and three-color strip light source is designed,in order to weaken or eliminate the influence of complex background interference on wafer identifier as far as possible.The influence of background interference of wafer identifier is analyzed from four aspects of light source type,spectrum(luminous color),illumination direction(luminous angle)and illumination brightness.The effectiveness of combined light source lighting system is verified by experiments and theories through manual dimming system.On the other hand,the real-time recognition algorithm of wafer identifier which is insensitive to complex background interference such as occlusion,confusion,character micro deformation and nonlinear illumination is studied in order to read the wafer identifier correctly and quickly.Due to the standardization of character types in the industrial field,the shape template matching based on the extended gradient direction is used to identify the wafer identifier.The data sets of Arial italics template and semi orthographic template without background interference are made to train the algorithm model.The time required for the algorithm model training is analyzed.It is proved that the algorithm is fast in adapting to the training of new font model.The test sets of Arial italics and several semi fonts are made to verify the effectiveness of the algorithm.It shows the recognition effect of the algorithm model trained under various background interference.The wafer identifier recognition system in this study is compared with WID 120,which is the most recognized chip reader in the industry.The experimental results show that the combined lighting system designed in this project can adapt to different wafer substrate materials,effectively distinguish the gray value difference between target foreground and interference background,and weaken or eliminate the influence of complex background interference on wafer identifier recognition.The shape matching algorithm based on the extended gradient direction can quickly and accurately identify the wafer identifier disturbed by occlusion,confusion and micro deformation of characters.