การรู้จำธนบัตรไทยด้วยเรียลไทม์แอปพลิเคชันบนอุปกรณ์เคลื่อนที่สำหรับผู้พิการทางสายตา

Abstract

This research aims to develop prototype of the real time mobile application for banknote recognition to support visually impaired persons. Based on literature review, the identification of each banknote with different colors were classifying according to color characteristics by calculating the average value or comparing the histogram of the color. Although, the color classification is considered as an easy and fast way to categorize banknotes. However, this method has limitations of detection incorrectness for similar object to the banknotes object. This may result error processing and wrong identification of the object as banknotes. In addition, there is deficiency of visually impaired mobile applications with real-time processing, which are usable to user experiences. Therefore, this research proposes the application of the Haar Like Feature to real-time detection and recognition of banknotes, histogram technique to identify banknote types, and text-to-speech method to read types of banknotes out loud. System prototype was developed as an Android-based mobile application. This is a guide to evaluating application performance based on applied techniques abilities and application usability. The research evaluation was divided into 2 parts: 1) evaluation the effectiveness of the banknote recognition by examining 5 types of Thai banknotes: 20 baht, 50 baht, 100 baht, and 1,000 baht 20 banknotes for each type with total amount of 100 pieces. This is included false banknotes and non- banknote objects with colors similar to the color of each banknote. 2) Assessment the usability of the application for user satisfaction. The results of the research on the effectiveness of banknote recognition is at 88% for accuracy level and the result of user satisfaction is at a good level. This research result indicates that the real time mobile application for banknote recognition to support visually impaired persons by applying Haar-Like Feature detection techniques and histogram comparisons can be used to support visually impaired persons.