A Novel System to Monitoring Food Waste in School Canteens Based on RGB-D Images

With the increasing severity of hunger issues, calculating and analyzing food waste has become one of the current hot research topics. However, the mainstream research methods for food waste in university canteens, such as the 24-hour recall method, questionnaire surveys, and weighing records, are not sufficiently intuitive and effective. These research methods have certain problems like non-representative statistics or insufficient data, leading to a certain deviation between the collected data and the actual food waste situation. Therefore, we chose to analyze the types and quantities of food waste using RGB-D image data. We collected a large amount of food waste image data using depth cameras installed in school canteens, then we utilized ISAT tool for data annotation and organization. Employing a multimodal semantic segmentation model, we conducted analysis of the obtained image depth data. We design an automated method for measuring food waste with a precision rate 91.455%. The results have shown that this method is reliable for monitoring food waste in school canteens, which can provide a data foundation for further in-depth analysis and research.