Dissertation
Knowledge-directed lead bond inspection
Doctor of Philosophy (Ph.D.), Drexel University
Aug 1989
DOI:
https://doi.org/10.17918/00008718
Abstract
This research is on automatic IC lead bond inspection based on computer image processing. The major emphasis in this work is to demonstrate that knowledge-based image processing (Expert Systems) can solve complex image analysis problems. In the course of this project we have developed optimal illumination and edge detection techniques for lead bond inspection. Various combinations of filters with different fluorescent lights were used to examine lead bonds. By careful visual and computer analysis of these images the most suitable method for lead bond inspection illumination was found. A special technique for lead bond inspection by using local statistics for image processing was developed and implemented. This includes edge detection, edge point finding, and picture detail enhancement. The method is very effective in lead bond edge enhancement and extraction, especially for low-contrast images. A knowledge-based image processing system for lead bond inspection was developed. It is based on a lead bond spatial model, which is an application of the Desk Image Model to lead bond inspection. The system is a three level bottom-up knowledge-directed search system that includes six experts. Design information regarding the lead bond is used to search the image for the required data. The lower level experts make most of the decisions, and only a few very tough problem are left for the higher level experts. Strategies and standards for the inspection process are set according the bond's geometric structure, dimensions, and engineering tolerance standards. Modular knowledge-directed image processing methods such as line tracing, curve search, and comer and line template matching are used in this system. The features of this system are local feature focus method; using some features that cause trouble to classical image processing as knowledge; using global knowledge to direct changes of search direction for finding the bond boundary; and from uncertainty to confirmation. The system was tested with 100 lead bond images.
Metrics
26 File views/ downloads
6 Record Views
Details
- Title
- Knowledge-directed lead bond inspection
- Creators
- Jie Chen - Drexel University, Electrical (and Computer) Engineering (1970-2026)
- Contributors
- Oleh John Tretiak (Advisor) - Drexel University, Electrical (and Computer) Engineering (1970-2026)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- xii, 174 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Electrical (and Computer) Engineering (1970-2026); Drexel University
- Other Identifier
- 991021888888004721