Computer Vision and Deep Learning Enabled Real-Time Liquid Level Detection and Measurement in Transparent Containers

Abstract

With the advent of the Industry 4.0, the capability to automatically detect and measure the liquid levels and volumes in the transparent containers (i.e. test tubes in a laboratory) in real-time is a crucial part in the whole perception system of an autonomous robot/system. And by the use of the state-of-the-art Computer vision and deep learning technology we can easily help solve this problem. This task of liquid level detection and measurement in transparent containers is a crucial task in various industries, including pharmaceuticals, manufacturing, and chemical processing especially in automated oil testing laboratories, where autonomous systems require precise perception capabilities. This paper explores the potential of computer vision and deep learning to accurately measure the liquid level and volume in transparent containers. Our approach is implemented by training a deep learning model, especially YOLOv8 (You Only Look Once version 8) model trained to recognize features like container type, liquid surface and height. After identifying these key features, our proposed algorithm uses them to correct the errors in the liquid height due to perspective distortion and get the liquid level followed by liquid volume accurately in real-time in a given transparent container. Our approach achieved an accuracy of 98.57 percent for transparent liquid and 98.12 percent for colored liquid, demonstrating the system's potential for practical applications particularly in various automated laboratory testing industries.