Main Article Content
NDT techniques play a pivotal role in ensuring the structural integrity and safety of various components in industries such as aerospace, automotive, and civil engineering. Among the wide range of defects that NDT aims to identify, cracks are one of the most common and critical types. Traditional crack detection methods have been extensively studied and employed over the years due to their reliability and cost-effectiveness. Most other methods of NDT need visual inspection because an operator will typically need to search for flaws. As computer vision continues to advance, its applications across a range of sectors, including object recognition, image segmentation, and image classification have grown significantly. One important area of research within computer vision is image denoising and crack detection, particularly in NDT. NDT is a technique used to evaluate and inspect materials or components without causing damage to the object being tested.
Traditionally, crack detection in NDT has been performed using image processing techniques such as thresholding, mathematical morphology, and edge detection(Lau et al. 2020). The improvement of image quality has become a crucial area of focus in various fields, including medical imaging, industrial applications, construction safety, and medical diagnostics and NDT. Recent advancements in machine learning (ML) and AI have revolutionized research programs in numerous domains, offering new opportunities for optimization and enhancement(Patnaik, Babu, and Bhave 2021). The field of medical imaging has witnessed significant advancements in the application of AI Deep Learning algorithms for automatic defect detection, image enhancement, and denoising(Luo, Zeng, and Chen 2022). These developments have made it easier to extract essential information from documents, websites, and images using Optical Character Recognition techniques.(Patnaik, Babu, and Bhave 2021)
Introduction:Cracks in materials and structures pose significant risks, leading to failures, reduced performance, and even catastrophic events. Thus, it is essential to find cracks early on in order to guarantee the dependability and safety of different components. Traditional crack detection techniques, such as visual inspection or conventional NDT methods, have been widely used for many years. However, these techniques frequently have issues with sensitivity, efficiency, and accuracy. Image denoising, restoration, edge detection, and enhancement are fundamental problems in image processing, computer vision, and AI(Y. Sun et al. 2006). For a variety of applications, image quality is crucial such as NDT, automatic control, robotics, imaging, target tracking, and telecommunications. The application of AI technology has garnered increasing attention in the past few years to improve the imaging quality of NDT Images, medical images, industrial applications, and other fields where image quality is crucial (Luo et al., 2022). Recent studies have shown that intelligent algorithms have a significant effect on denoising MRI images and enhancing the imaging quality of NDT and medical images. In the sector of NDT and medical imaging, the quality of images is of utmost importance for accurate diagnosis and treatment.
Objectives:The goal of this literature review is to investigate how deep learning and artificial intelligence (AI) can be used to optimize algorithm parameters for denoising, image enhancement, and automatic defect detection. The investigation of deep learning techniques' uses for image enhancements and automatic defect detection in Non-destructive testing (NDT) is the main goal of this literature review paper.
Conclusions:Industrial product quality is a crucial component of product development, and research on defect-detection technology is extremely vital in terms of practical use for ensuring product quality. The research state of product crack-detection technology and image denoising in intricate industrial processes is thoroughly reviewed in this study. We have thoroughly reviewed the experimental findings of crack-detection methods and compared and analysed deep learning Image denoising and crack-detection techniques with traditional Image denoising and crack-detection methods.
Main Research gaps, which includes lack of preprocessing of image under inspection before defect detection and speed efficiency along with accuracy of Deep learning models for image denoising and crack detection are considered for call of action of our research and development. Optimal parameters for image denoising and parameter tuning for YOLOv8 Deep learning algorithm is primary key areas identified for research as per overall literature review.