Design Of Image Processing System For Mask Defect Detector

Mask is widely used in lithography processes and plays a vital role in lithography production.The rapid development of the semiconductor industry and lithography technology puts forward higher requirements to the quality of the mask,once the mask having defects will lead to the whole batch of semiconductor devices,resulting in huge economic losses,mask defect detection technology is to ensure the quality of mask,to ensure production efficiency of one of the important means.At present,there is less research on mask defect detection technology,and the relevant detection system and equipment mainly rely on imports,so the mask defect detection technology has important research significance and application value.The work of this paper is mainly based on the mask image acquisition system,combined with machine vision and GPU multi-threaded technology to achieve high-precision and rapid detection of mask defects,the main research content is summarized as follows:1.Investigated the current status and development of mask defect detection technology,introduced the main mask detection methods and principles,introduced and analyzed the technologies and theories used in mask defect detection systems,and proposed a machine vision technology is used as the skeleton,and GPU parallel processing technology is used as a mask defect detection method.Based on this,a defect detection overall system including mask image acquisition and image processing is constructed.2.The image acquisition system of mask defect detection is constructed,the design and formation of mask image acquisition system are described in detail,and the defect prototype is drawn,which provides a good experimental platform and environment for mask defect detection.3.The design and implementation of the mask image processing system is completed,the principle of common image processing algorithm is introduced,the algorithm suitable for this system is selected by comparative verification,a fast image stitching algorithm suitable for this system is improved,and some processing algorithmics is optimized.On this basis,the software implementation of the mask image processing algorithms is carried out to complete the detection of the defects of the mask defects required by the system.4.In order to improve the image processing speed,based on GPU architecture and CUDA platform for the host and device side task allocation,we designed CPU multithreaded and GPU multi-threaded combination way to hide data transmission time.Computationally intensive and less logical algorithm is carried out in CUDA design programming.The overall acceleration effect of the system is tested and statistically calculated,and the speed of test verification defect detection and treatment is significantly improved.