Online measurement of fill level of disc pelletizers using RGB-D camera

Abstract

Vision-based technique for fill level measurement in pelletization is widely used in industries and is crucial to ensure high-quality production. However, the existing approaches are 2D-vision-based and majorly designed for rotary drums. Nevertheless, for disc pelletizers, it remains an under-study and challenging problem due to the complex structure of apparatus and granules kinetics in disc. In this work, an online fill level measurement method for disc pelletizers by an RGB-D camera is proposed. Firstly, a background depth data subtraction and automatic granular area identification method are developed to obtain the 3D pointe cloud data of the granular area by machine learning and model-based processes; thereafter, a proportion conversion algorithm is developed to real-time calculate the granular pile volume and fill level. Extensive experiments demonstrate the effectiveness of our approach with an average relative error of 5.42%, which meets the requirements in industries.