A Method for Measuring Liquid Volume in Transparent Glass Vials Utilizing Image Processing and Feature Recognition

The noncontact detection method of liquid volume in glass vials is essential for various industries, particularly pharmaceuticals. This study presents a novel method that integrates image processing and feature recognition. Optical path simulations were conducted using COMSOL Multiphysics software to analyze light propagation through the inner and outer boundaries, meniscus, and bottom boundary of transparent glass vials. We found that constraining the upper light considerably improves the capture accuracy of the bottom information. Based on these findings, we proposed an integration-based methodology to calculate liquid volume, which thoroughly investigates the precise position of each feature in the image, accurately describes the liquid boundary in the glass vials, and offers an integral method for liquid volume calculation. Experimental findings revealed that the proposed approach consistently exhibits high stability and reliability through different exposure times and shoot distances, attaining an average detection accuracy of over 99% for small glass vials.