Vision-based Identification Extraction for Freight Containers

doi:10.1109/ICWAPR.2015.7295925

機構典藏 > College of Engineering > Graduate Institute & Department of Computer Science and Information Engineering > Proceeding > Item 987654321/120093

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/120093

Title:	Vision-based Identification Extraction for Freight Containers
Authors:	Fahn, Chin-Shyurng;Su, Bo-Yuan;Wu, Meng-Luen
Keywords:	Identification extraction;Freight container;Optical character recognition;Computer vision;ISO/TC104
Date:	2015-10-12
Issue Date:	2021-03-06 12:11:44 (UTC+8)
Abstract:	The car license plates have regular formats and styles, and can be recognized by computer vision systems. However, some information that exclude from the alphanumerical characters of the car license plates are printed directly on a car body. Unlike car license plates, freight container's identifications have different colors and alignments. Some car bodies are not flat, such as trucks and freight containers, which increases the difficulty for character recognition. The character intervals in the identifications are varied from each other, and other information such as advertisement and logos are also marked on the car body, which make it hard to recognize characters too. Based on computer vision techniques, a fast and efficient method to extract the identification from freight containers is presented in this paper. The identification formats and styles conform to the standard of ISO/TC104. The information fliat does not pertain to part of the identification is eliminated during the processing. The most possible groups remaining in the image are extracted for identifications. Finally, an optical character recognition algorithm, Tesseract, is applied to recognize character images and then converted into texts. Thirty-five back surface images of freight containers are used to verify the effectiveness of the proposed method. The resolution of the images is set to 600 × 500. From the experimental results, we are encouraged that the accurate rate of identification extraction is over 90.37% and the average execution time is 12.91 seconds achieved by our proposed method.
Relation:	2015 International Conference on Wavelet Analysis and Pattern Recognition (ICWAPR)
DOI:	10.1109/ICWAPR.2015.7295925
Appears in Collections:	[Graduate Institute & Department of Computer Science and Information Engineering] Proceeding

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	43	View/Open

Loading...