Automatic recognition method for liquid level gauge reading based on machine vision

In the traditional industrial field, the reading recognition of various liquid level gauges mainly relies on manual reading and recording, which presents issues such as low accuracy and inefficiency. This paper proposes an automatic recognition method for liquid level gauge reading based on machine vision, utilizing machine learning models to accurately and in real-time recognize the readings of liquid level gauges. To achieve a high recognition accuracy, various types of image features were extracted, and classification algorithms such as Logistic Regression, SVM, and Random Forest were employed. Different image features and models were trained and tested for comparison. The test results show that the algorithm achieved a classification accuracy of 98.05% on the dataset, with a reading error of less than 0.5%. Moreover, in real-world industrial testing, the system can accurately track the continuously fluctuating liquid level and recognize the current reading. This method has high accuracy and stability, meeting the requirements for real-time and accurate automatic recognition of liquid level gauge readings in industrial environments.