Quality mapping of potato tuber storage facility using real-time location system, wireless communication and machine vision for improved traceability

Abstract

This thesis proposes a novel post-harvest potato tuber storage traceability system designed to enhance quality monitoring and management. The system leverages a real-time localization system (RTLS), wireless communication and machine vision to identify and map the quality characteristics of potato tubers. RTLS experiments, conducted in both controlled laboratory settings and field environments, utilized a 3D grid structure with up to 60 actual tag locations, each generating 100 predicted positions. To further enhance accuracy, a Kalman filter was employed in conjunction with LS and a Support Vector Regressor, achieving a total Root Mean Square error (RMSE) as low as 0.183 meters in obstacle-free environments. Field tests were conducted during active harvest operations at a commercial potato farm, demonstrating the system’s robustness under real-world conditions. Additionally, a Grading Localization Synchronization System (GLSS) was implemented to predict tuber travel times during bin filling operations: thus, synchronizing localization with machine vision grading. The experiment was repeated for 30 trials, capturing both actual tuber travel times, and predicted times for nine combinations of bin-piler extension and conveyor belt speed. The system integrated real-time monitoring of bin-piler extension and belt speed, achieving synchronization RMSE values as low as 0.356 seconds with a Linear Regression model, compared to 0.585 seconds without. By creating a detailed 3D quality map of the storage facility, integrating tuber quality data with their spatial and temporal placement, and uploading this information to a remote server, the system enables centralized monitoring and analysis. This integrated approach significantly enhances traceability, reduces post-harvest losses, and improves the overall quality management of stored potato tubers.