When deep learning applied, mini-LED high dynamic range quality assured

Mini-LED backlight technology has emerged as a leading solution to enhance the performance of high dynamic range (HDR) liquid crystal displays (LCDs), particularly for automotive applications. The introduction of Mini-LEDs offers enhanced brightness, contrast, and longevity; however, it also presents significant challenges, including the detection of dead pixels and precise LED placement due to their small scale. Traditional inspection systems, both manual and automated, struggle to meet the high-precision requirements necessary for compact Mini-LED layouts, underscoring the need for more sophisticated manufacturing solutions.

To address these challenges, researchers at Xiamen University led by Professor Weijie Guo have developed an innovative approach using a high-resolution network (Hrnet) enhanced with a mixed dilated convolution and dense upsampling convolution (MDC-DUC) module, and a residual global context attention (RGCA) module. This model dramatically outperforms conventional methods, achieving a mean intersection over union (Miou) of 86.91%. By integrating advanced deep learning techniques, the team has significantly improved precision in detecting the quality of vehicular Mini-LED backlights, providing a robust framework for automated optical inspection (AOI) systems. This advancement not only enhances the efficiency of production lines but also ensures higher quality control standards in the manufacturing of Mini-LED backlights.

The work entitled “Vehicular Mini-LED backlight display inspection based on residual global context mechanism” was published on Frontiers of Optoelectronics (published on Oct. 29, 2024).