LCD Graphic Element Location Method Combining NL-means and Local Threshold Segmentation

Accurate positioning of screen primitives is crucial in the machine vision-based automatic detection of intelligent water meter LCD screens. Detecting edge details of the LCD screen using the A component of the LAB color space improves the accuracy of LCD screen area positioning. This approach reduces background interference from Gaussian noise, non-uniform lighting, specular reflection, and local highlights. This study proposes a technique that combines NL-means and Sauvola local threshold segmentation methods to locate LCD screen areas and graphics elements. The experimental results indicate that this technique satisfies the defect detection criteria for smart water meter LCD screens set by the enterprise. The LCD screen detection tool extracted a smart water meter LCD screen image with 98.4% accuracy in LCD screen element positioning. Compared to the median filter method, this represents a significant improvement, and the combination with the maximum between-class variance method further increases accuracy by 2.7%.