Research On Fast Correction And Character Region Positioning Method Of Semiconductor Chip Image

Semiconductor chip manufacturing technology is currently considered as the most advanced technology,which plays a vital role in the rapid development of industry in the 21 st century and is widely used in national economic construction and people's daily social life.With the rapid development of the electronics market,the demand for semiconductor chips continues to increase,and higher and higher requirements are placed on semiconductor chips.There are various inevitably problems occuring,such as chip tilt and poor chip character position,which directly affect the quality of character printing,in the process of appearance inspection of semiconductor chip packaging.Although there are many researches and applications of image processing at present,there are few researches on image processing technology related to semiconductor chip surface inspection.Therefore,this thesis mainly studied the key image processing technology in the process of semiconductor chip package visual inspection,and established the automatic chip image acquisition and inspection system.This thesis specifically includes the following:(1)Design of automatic chip image acquisition and inspection system.In order to achieve continuous and high-precision image acquisition of semiconductor chips,firstly,the hardware of the image acquisition system and a suitable lighting scheme is designed.Secondly,according to the technical requirements such as the shooting range of the chip,the calculation of the key parameters of the industrial camera and lens is performed,and the corresponding hardware is thus calculated.Then,according to the size of the tray,the platform structure of automatic chip image acquisition and inspection system is designed.Finally,based on the design of each individual software function modules of the system,the automatic semiconductor chip image acquisition and inspection system with high-quality and high-efficiency is established.(2)Research on the segmentation method of multiple semiconductor chip images.Aiming at the problem that a single image acquired by the automatic image acquisition and inspection system usually contains multiple chips,a method for segmenting multiple semiconductor chip images is proposed,in which the polygon approximation algorithm is improved,combined with the contour moment to locate the contour centroid,to realize the complete process of dividing multiple chips into several single chip images in one image.The efficiency of chip inspection then can be effectively improved.(3)Research on fast correction method of semiconductor chip image.Aiming at the problem of the image tilt of the chip during the visual inspection of package defects,this thesis proposes a fast correction method for semiconductor chip images.Firstly,Harris corner detection algorithm is improved,combined with Canny edge detection and the removal of irrelevant contours,and the corner vertices of the target contour are extracted.Secondly,the least square method is used to fit a straight line to the vertex of the longest side.Finally,the image is corrected based on the result of the straight line fitting.The experimental results reveal that the accuracy of the obtained tilt angle and the operating efficiency by the proposed method are better than the traditional Hough transform,the minimum external moment method and the Fourier transform correction method.(4)Research on the positioning method of the character region of semiconductor chip.Aiming at the defects such as character tilt and character position error in laser printing of smallsize chips,a method for positioning the character region of the chip image is proposed.Firstly,Harris corner detection is used to obtain the image corner distribution map.Secondly,the convex hull detection algorithm is improved,based which the corners of the non-character area can be removed,and the convex hull of the outermost corners can be obtained.Finally,the minimum enclosing rectangle of the convex hull is fitted,and the character region is positioned.The experimental results reveal that,compared with the morphological filtering positioning method,the positioning method based on edge characteristics,the positioning method based on character's texture features,and the positioning method based on convex hull and the minimum external moment,the chip character region positioning is more accurate by the proposed method,and the calculation amount can be better reduced,the calculation efficiency can be better improved.(5)Experimental comparison and analysis.Firstly,based on the camera,lens,PLC,computer and three-axis platform,the MFC program and the control interface are designed.The automatic chip image acquisition and inspection system is established,which includes image acquisition,image segmentation,image correction,and image character region positioning.Secondly,according to the requirements of a large semiconductor company for chip lead surface/plastic cover inspection,the main technical indicators of chip surface inspection are listed.The detailed analysis of the experimental results of the chip image acquisition and inspection is carried out,based on the realization of automatic chip image acquisition,multiple chip image segmentation,fast correction and character region positioning.The experimental results meet the technical indicators of the chip package appearance quality inspection.