Deep Learning Neural Network Performance on NDT Digital X-ray Radiography Images: Analyzing the Impact of Image Quality Parameters-An Experimental Study

In response to the growing inspection demand exerted by process automation in component manufacturing, non-destructive testing (NDT) continues to explore automated approaches that utilize deep-learning algorithms for defect identification, including within digital X-ray radiography images. This necessitates a thorough understanding of the implication of image quality parameters on the performance of these deep-learning models. This study investigated the influence of two image-quality parameters, namely signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), on the performance of a U-net deep-learning semantic segmentation model. Input images were acquired with varying combinations of exposure factors, such as kilovoltage, milli-ampere, and exposure time, which altered the resultant radiographic image quality. The data were sorted into five different datasets according to their measured SNR and CNR values. The deep-learning model was trained five distinct times, utilizing a unique dataset for each training session. Training the model with high CNR values yielded an intersection-over-union (IoU) metric of 0.9594 on test data of the same category but dropped to 0.5875 when tested on lower CNR test data. The result of this study emphasizes the importance of achieving a balance in training dataset according to the investigated quality parameters in order to enhance the performance of deep-learning segmentation models for NDT digital X-ray radiography applications.

Automated summary

In order to meet the growing inspection demand in component manufacturing, non-destructive testing (NDT) explores automated techniques using deep-learning algorithms for defect identification in digital X-ray radiography images. This study investigates the impact of image quality parameters, specifically signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), on the performance of a U-net deep-learning model. Varying combinations of exposure factors were used to acquire input images, and the deep-learning model was trained on datasets categorized by measured SNR and CNR values. Training the model with high CNR values resulted in a higher intersection-over-union (IoU) metric, highlighting the importance of balancing training datasets based on quality parameters for improved performance in NDT digital X-ray radiography applications.