A method to read numbers in Thai nutrition facts label by using SVM
Conference proceedings article
Authors/Editors
Strategic Research Themes
No matching items found.
Publication Details
Author list: Ekkasan S., Charoenpong T., Chamnongthai K., Charoensiriwath S.
Publisher: Hindawi
Publication year: 2018
Start page: 1
End page: 4
Number of pages: 4
ISBN: 9781538626153
ISSN: 0146-9428
eISSN: 1745-4557
Languages: English-Great Britain (EN-GB)
Abstract
Due to a problem of current research occurring when reading the nutrition facts label with Thai font. In this paper, we proposed a method to read the amount of the nutrition numbers on nutrition facts label with Thai font by using integrated image processing technique. This method consists of three steps: label extraction, number segmentation, and number classification. Nutrition facts label image is used as input of algorithm. First, the label is extracted by using a local adaptive threshold. Four sides of a label box are detected from histogram in horizontal and vertical axis. Second, numbers are segmented based on blob region analysis. Blob of number can be defined according to blob sequence in a line. A number in blob of number is then separated. Third, a number is divided into 5ื7 regions. A number vector defined from 35 regions is used to classify the number by the support vector machine. To test the performance of the proposed method, number of the amount of the nutrition is extracted from twenty Thai labels. In such labels, there are 180 numbers. A number is classified into number 0-9. By using support vector machine for classification, the accuracy is 81.11%. The experimental result shows the satisfactory results. This is the first method for reading number in Thai nutrition facts label. ฉ 2018 IEEE.
Keywords
number calssification, Nutrition facts label, the amount of nutrition
Contact Information