Automatic detection of dark stain defects in low-light-level image intensifiers based on uniformity correction and...

Dark stain defects are caused by the slow drying of cleaning agents accidentally sprayed on the cathode window during the manufacturing stage of low-light-level (LLL) image intensifiers. Their existence leads to a deterioration in the imaging quality of the device and weakens its target detection capability. The main drawback of the traditional subjective detection method is the introduction of artificial uncertainty, while sensitivity to background uniformity and texture is a weakness of existing objective detection methods. To address the above issues, an automatic detection method for such defects based on uniformity correction and background texture filtering is proposed. First, existing techniques are used to detect and eliminate regular defects in the target image, and a dual uniformity correction strategy based on multi-directional fixed area segmentation and grayscale compensation is utilized to alleviate background unevenness. After obtaining candidate target defects through region detection and region comparison, a texture filtering strategy based on Gaussian differential filtering of spectrograms is adopted to purify the image background. Finally, the frequency-domain bandpass filtering is combined with regional feature analysis to achieve automatic detection of dark stain defects. The experimental results demonstrate that the proposed UC strategy can produce ideal results for the studied case, and the performance of the presented TF method is superior to state-of-the-art relevant techniques. Compared with existing defect detection technologies, the proposed approach has higher detection accuracy, with an average detection accuracy of 94.2% during batch testing. Therefore, this method can be deemed as an effective automatic detection scheme for dark stain defects of LLL image intensifiers.