Machine Vision-based LED Display Area Defects Detection Technology

Due to such problems as manufacturing process, heat dissipation and aging, LEDdisplay shows defects, thus influencing the effect of the display. Since many kinds ofdefects exist in LED display imaging, there is no unified standard of defect detection.At present, the defect detection of LED display is accomplished through the subjectiveevaluation of experienced technicians. It is difficult to guarantee the efficiency andaccuracy. So it seems particularly important to test the quality of display imagingaccurately and evaluate the quality of the display.An analysis is made of the causes of defects in terms of the structure of LED lampsand displays as well as attenuation characteristic, with the display defects classifiedaccording to the shape and size. Through such analysis, I draw an conclusion that planardefect is a key problem for detection. Based on its characteristics, the application of thedisplay defect detection based on machine vision technology is determined. The finalpurpose of detecting defects is achieved through the separation of defects by polynomialsurface fitting based on least-squares, combined with corresponding denoising, filteringand morphological methods. Then the quantitative classification is done for the degreeof defects, and an experiment is designed to determine the accuracy of testing methods.The results show that when the LED display background is good, it is possible todetect the display defect effectively. It is meaningful for the quantification of defects,the determination of detection accuracy as well as the evaluation of the display qualityof an LED display.