With the development of information and communication technology,the monolithic microwave integrated circuits are also upgrading constantly.And the increasing requirements in MMIC performance makes the traditional manual design or semi-manual design methods no longer sufficient.So,the MMIC automation design technology will be very important.And the component size extraction algorithm,which plays a key role in establishing MMIC data sets and training agents for MMIC automation design,will be very important too.However,those projects are still in their infancy,there are still a lot of deficiencies in this research field.Therefore,this article proposes a series of image-based size extraction algorithms,which is used to extract the size of MMIC components,to fill the gaps in this research field.The main content of this article is extracting the size of MMIC components from the image.First,we use Otsu threshold segmentation method and Laplacian edge detection method to segment the image.After that,the Hough line detection algorithm is used to correct the image.Then,according to the requirements of the extraction algorithm,a largest inscribed circle search algorithm is designed to search for the largest inscribed circle of the component.After that,the size extraction algorithm is designed.First,the MMIC components are divided into three types: line,rectangle and circle.Then,we propose six size extraction algorithms,including parallel scanning method,contour matching method,area method,outer parallel scanning method,internal parallel scanning method and rotating scanning method,to extract the size of those three structures,and succeeded.Then,several modules are built with these algorithms for each type of component.Finally,this article successfully extracted all 15 types of size parameters of 5 types of MMIC components,including microstrip line,resistor,capacitor,rectangular inductor and circular inductor.And 8 sets of experiments are designed for algorithm verification.The experimental results are good,which proves that the algorithm proposed in this article is effective. |