Integrated Image Processing Framework to Determine Nutrient Quantities from Guideline Daily Amount (GDA) Label

Guideline Daily Amounts (GDA) is a labeling scheme designed to provide a better understanding of nutrition, as compared to traditional nutrition labels. Ready-To-Eat foods containing high sugar, fat and salt were first selected for GDA labeling. The Thai GDA label features key nutritional information such as energy, sugar, fat and sodium. This paper presents an integrated image processing framework that extracts numerical nutrition information from the Thai GDA label images. The proposed technique utilizes a combination of global thresholding, image histogram analysis and template matching to classify numerical nutrition information from the label. The method is divided into five main processes: image acquisition and transformation, label region extraction, nutrient recognition, number extraction and number classification. The image dataset of packaged food products, captured from a mobile device, containing the total of 162 numbers is used in the study. The experimental results show that the classification accuracy rate of nutrition information can achieve 76.54% when the optimal number of templates is reached. Moreover, the effect of the number of templates on the classification performance is evaluated. It is shown that the classification accuracy is enhanced when the number of templates is increased. However, the accuracy gains diminish and may result in lower accuracy when the number of templates is increased beyond an optimal threshold.