For a typical surface automated visual inspection (AVI) instrument of planar materials, defect classification is an indispensable part after defect detection, which acts as a crucial precondition for achieving the on-line quality inspection of end products. In the industrial environment of manufacturing flat steels, this task is awfully difficult due to diverse defect appearances, ambiguous intraclass and interclass distances. This paper attempts to present a focused but systematic review of the traditional and emerging automated computer-vision-based defect classification methods by investigating approximately 140 studies on three specific flat steel products of con-casting slabs, hot-rolled steel strips and cold-rolled steel strips. According to the natural image processing procedure of defect recognition, the diverse approaches are grouped into five successive parts: image acquisition, image preprocessing, feature extraction, feature selection and defect classifier. Recent literature has been reviewed from an industrial goal-oriented perspective to provide some guidelines for future studies, as well as to recommend suitable methods for boosting the surface quality inspection level of AVI instruments.
The morphologies of various surface defects on strip steel suffer from oil stain, water drops, steel textures and erratic illumination. It is still challenging to recognize defect boundary precisely from cluttered backgrounds. This paper emphasizes such a fact that skip connections between encoder and decoder are not equally effective, attempts to adaptively allocate the aggregation weights which represent differentiated information entropy values in channel-wise, by importing a stack of cross-attention transformer (CAT) into the encoder-decoder network (EDNet). Besides, a cross-attention refinement module (CARM) is constructed closely after the decoder to further optimize the coarse saliency maps. This newly nominated CAT-EDNet can well address the semantic gap issue among the multi-scale features for its multi-head attention structure. The CAT-EDNet performs best on insuring defect integrity and maintaining defect boundary details when compared with twelve state-of-the-arts, and the detection efficiency is at 28 fps even under the noise interfered scenario.
Automated visual inspection (AVI) instrument of surface defects for hot-rolled steel strips is conventionally installed closely before the terminal crimping machine, where the adjacent upstream process is laminar spray cooling. Waterdrops, spreading more or less over the steel strip surface, often trigger false-alarm, which is a quite common problem in AVI. Stimulated by the idea of image rain removal in visual enhancement field, this paper considers the surface waterdrops, pseudo defects in essence, as a conceptual "rain-like layer". A targeted method, namely progressive recurrent generative adversarial network (PreGAN), is designed for active waterdrop tracking and fine-grained image inpainting. Meanwhile, a steel surface database (2400 raw images with the resolution of 1000×1000) captured from actual hot-rolling line is constructed for the first time for open evaluation of waterdrop removal. The experimental results indicate that images enhanced by the PreGAN perform the most informative and spotless, with 52.2073 peak signal-to-noise ratio (PSNR) and 0.9502 structural similarity index (SSIM), when compared with four prestigious networks. Assisted by the PreGAN, the false alarms are proved to be reduced at least a half during the application tests using four traditional simple detection methods.
Steel strip acts as a fundamental material for the steel industry. Surface defects threaten the steel quality and cause substantial economic and reputation losses. Roll marks, always occurring periodically in a large area, are put on the top of the list of the most serious defects by steel mills. Essentially, the online roll mark detection is a tiny target inspection task in high-resolution images captured under harsh environment. In this paper, a novel method—namely, Smoothing Complete Feature Pyramid Networks (SCFPN)—is proposed for the above focused task. In particular, the concept of complete intersection over union (CIoU) is applied in feature pyramid networks to obtain faster fitting speed and higher prediction accuracy by suppressing the vanishing gradient in training process. Furthermore, label smoothing is employed to promote the generalization ability of model. In view of lack of public surface image database of steel strips, a raw defect database of hot-rolled steel strip surface, CSU_STEEL, is opened for the first time. Experiments on two public databases (DeepPCB and NEU) and one fresh texture database (CSU_STEEL) indicate that our SCFPN yields more competitive results than several prestigious networks—including Faster R-CNN, SSD, YOLOv3, YOLOv4, FPN, DIN, DDN, and CFPN.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.