การจำแนกชนิดยาจากฉลากยาด้วยการรู้จำอักขระด้วยแสง

Abstract

This engineering project aims to study methods for detecting and classifying medications for elderly individuals and those who have difficulty managing multiple medications. It has been found that people around the world use many medications, and medication intake can be observed at every stage of life. As we age, the number of types of medications we take increases, which leads to confusion in medication management. Therefore, this research proposes methods consisting of 2 processes: A process for segmenting and counting the number of packages. And A process for detecting characters from the medicine packages. The performance metrics of the system presented in this project include Intersection over Union (IoU), which resulted in 95.6%, and the accuracy of counting the number of packages, which was 100%. The results of character detection for medication classification from the experiment are as follows: Method 1 using OCR only, Method 2 using OCR combined with Bradley, and Method 3 using OCR combined with Bradley and an additional step for noise removal achieved accuracies of 32.26%, 38.71%, and 41.96%, respectively. Additionally, the accuracy of using confidence values from the three methods for medication name identification resulted in the highest accuracy of 45.16%.