Research And Implementation Of Semiconductor Wafer Process Monitoring Method Based On Machine Visio

Wafer is a very important part in semiconductor manufacturing.The production quality of wafer directly affects the performance of chip in the future.With the development of semiconductor processing technology and manufacturing process,quality detection technology is also improving.In the past,wafer inspection was either observed and judged by quality control inspectors with the naked eye,or tested by using expensive laboratory equipment.Now machine vision detection effectively fills the gap between human eye detection and complex laboratory equipment detection.Noncontact computer vision equipment can transform the texture and features that are difficult to find on the wafer surface into visual images,and then analyze and process the images through various advanced machine vision algorithms.At the same time,image processing technology and visual algorithm are also constantly improving and changing,and the algorithm suitable for various complex scenes and extremely fast operation speed is also gradually improving.Therefore,based on machine vision,this paper can effectively solve a series of problems in wafer production and greatly improve the automation and intelligence of enterprise production.Firstly,based on the demand analysis of wafer surface particle counting.An improved watershed algorithm and two segmentation methods of skeleton contour curvature analysis are designed to separate the adhesion regions and complete the counting.And a method for analyzing,segmenting and counting the skeleton contour curvature of strip like particles.The test shows that the method has high accuracy.Secondly,based on the analysis of wafer edge mark monitoring requirements in the wafer boat box,a semantic segmentation algorithm is designed to extract the wafer region from the wafer boat box background,and then find the target through shape based template matching.The designed algorithm can effectively solve the practical problems in the production process.Then,through the research on the relationship between detection methods and visual algorithms in the above scenes,from the perspective of visual system requirements,a visual detection system framework is designed,and the main components of the framework are introduced.The functional modules can be flexibly combined for different detection requirements,in which the algorithm unit and algorithm connection reflect the innovation of the system.Finally,based on the combination and application of the system detection framework designed in this paper and the two design detection algorithms,the wafer surface particle counting system and the wafer edge detection system in the wafer boat box are developed.The specific implementation method of the combination of algorithm and system is described in detail.