Thesis
Machine learning approach to barcode detection and stamp identification
Washington State University
Master of Science (MS), Washington State University
2010
Handle:
https://hdl.handle.net/2376/101183
Abstract
We seek to radically improve processing of public records using open source software and machine learning algorithms. In this thesis, we focus on a core aspect of document recognition--identifying and isolating recording stamps that are added to each document by the county recording office. We aim to identify barcodes within a document image using machine learning. We then apply a probabilistic voting scheme to determine which barcode is part of the recording stamp. The vote indicates: (i) whether a recording stamp is present; and (ii) the most likely bounding box for the stamp. Once the recording stamp is identified, we extract the associated document code that directly identifies its type. In this thesis, we show the results of our experiments on the barcode detection and the stamp identification. We find our classifier achieved an average of 99.665% accuracy for detecting the stamp barcode in eight counties. Our method achieves 99.978% average accuracy in detecting stamps among eight counties in our test corpus. Finally, we show how useful information can be extracted from the recording stamp after it has been properly located.
Metrics
40 File views/ downloads
50 Record Views
Details
- Title
- Machine learning approach to barcode detection and stamp identification
- Creators
- Bhadresh Patel
- Contributors
- Scott Andrew Wallace (Degree Supervisor)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University; [Pullman, Washington] :
- Identifiers
- 99900525200901842
- Language
- English
- Resource Type
- Thesis