Automated Mura Defect Detection System on LCD Displays using Random Forest Classifier

Mura on LCD displays are challenging visual defects to detect, due to uneven brightness of the panel. So, automatic detection of mura defects on TVs and monitors is an important application to ensure the quality of LCDs manufactured in electronic product factories. In this work, conducted by three partners, called UFAM/CETELI, Envision (TPV Group) and ICTS, we propose an automated mura defect detection system on LCD displays that are based on similarities of the histograms and a machine learning (ML) algorithm called random forest. For comparison with the proposed method, we use one of the most popular methods of ML called k-Nearest Neighbors (k-NN). The experimental results show a detection rate above 99%, and processing time of 27 milliseconds per image, highlighting as contributions of our proposal, the high accuracy and reduced processing time, which allows the use of our solution in real production processes of LCD monitors.