| In light of the swift advancements within the realm of TFT-LCD technology,the identification of display imperfections is paramount to maintaining a high caliber of production.This treatise delves into the utilization of machine vision for TFT-LCD defect detection,specifically focusing on discerning abnormal display and chromatic aberration anomalies in intricate imagery when the TFT-LCD is illuminated.The primary areas of investigation and innovation consist of:(1)The formulation and establishment of an experimental procedure centered on machine vision for the detection of aberrant TFT-LCD display defects,thereby facilitating the assemblage and examination of anomalous display data.To overcome the challenges posed by limited defect exemplars,disparate classifications,and protracted collection periods in real-world manufacturing processes,two varieties of TFT-LCD display defect specimen datasets were simulated and generated.Initially,complex image defect datasets were synthesized by incorporating varying degrees of distortion.Subsequently,a dataset of chromatic aberration defects was fabricated,based on the principles of color blending,by adding incremental degrees of color deviation.This approach effectively underpins algorithmic research and evaluative testing.(2)To tackle the complexity of defect detection in intricate images engendered by TFT-LCD illumination,we put forth a defect identification technique predicated on spatial feature matching and color rectification.This method employs a Butterworth lowpass filter to eradicate periodic texture disruptions within TFT-LCD arrays.Thereafter,the integration of the SURF and MSAC algorithms facilitates the pairing and optimization of feature points between the detection template image and the scrutinized image.By calculating the projection transformation matrix,a spatially aligned image is procured,followed by the implementation of adaptive histogram matching for color coordination to mitigate the impact of color backgrounds.Finally,adaptive local threshold segmentation is employed to eliminate differential information pertaining to geometric and color backgrounds,thus yielding the defect detection image.This technique boasts a remarkable accuracy of 99.43%,with an average processing time of under one second,rendering it suitable for practical applications.(3)We introduce a colorimetric feature-centric method for the detection of chromatic aberration defects within TFT-LCDs.This approach utilizes a CCD camera and color analyzer to acquire TFT-LCD images and tristimulus values,confirming that the TFTLCD adheres to channel independence and color constancy criteria.The colorimetric feature is applied to the TFT-LCD,and a GOG model for chromatic aberration defects is formulated.The image under inspection is input into the GOG model,yielding the corresponding tristimulus values.These values are transposed into a uniform CIE-Lab color space,and the color discrepancy is computed employing the CIEDE2000 color difference equation.The accuracy of chromatic aberration defect detection in this study is an impressive 95%,with an average processing time of under 107 ms.This method offers a viable means of detecting chromatic aberration defects in display screens using CCD technology. |
