Manufacturers of printed circuit boards (PCB) need various footprint patterns to produce PCB. With the differences in manufacturing capabilities, technologies, methods, conventional manufacturers' representations, different rules in the interval, size, etc., of components, the manufacturers use human resources to reorganize the parts designed from different customers. However, using human resources would encounter three potential problems: (1) human calculation error, (2) long time spent on integration, and (3) non-uniform design. Without solid design principles, the engineers would draw patterns with a slight bias. In such a case, they need to conduct the re-draw procedure, which causes duplication of drawing the patterns and increases design costs. This project aims to develop an automated system to solve the current circuit board design problems. The automated solution mainly involves two major tasks: (1) document data extraction and analysis (2) components pattern structured analysis so that the format will be unified. After implementing this project for nine months, we have used deep learning technologies such as object detection, text detection, and text recognition to identify the manufacturer and positions of the components from an electronic component report. In addition, the texts (including values and codes) corresponding to important parameters are extracted around the drawing views. Accordingly, in the second year of this project, we aim to establish statistical models of the extracted parameters for each component, including generalizing the related rules between the parameters and summarizing the design methods adopted by various manufacturers. Besides, we will also conduct comprehensive discussions and experiments on the proposed method. This project aims to find the most suitable document automation process by combining professional expertise from academia and industry. We wish the results of this project can make a remarkable contribution to the electronic industry's technical development.
|Effective start/end date||1/08/22 → 31/05/23|
- component drawing
- maximum likelihood estimation
- object detection
- statistical modeling
- text recognition
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.