An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features

Abstract

A complete defect detection task aims to achieve the specific class and precise location of each defect in an image, which makes it still challenging for applying this task in practice. The defect detection is a composite task of classification and location, leading to related methods is often hard to take into account the accuracy of both. The implementation of defect detection depends on a special detection data set that contains expensive manual annotations. In this paper, we proposed a novel defect detection system based on deep learning and focused on a practical industrial application: steel plate defect inspection. In order to achieve strong classification ability, this system employs a baseline convolution neural network (CNN) to generate feature maps at each stage, and then the proposed multilevel feature fusion network (MFN) combines multiple hierarchical features into one feature, which can include more location details of defects. Based on these multilevel features, a region proposal network (RPN) is adopted to generate regions of interest (ROIs). For each ROI, a detector, consisting of a classifier and a bounding box regressor, produces the final detection results. Finally, we set up a defect detection data set NEU-DET for training and evaluating our method. On the NEU-DET, our method achieves 74.8/82.3 mAP with baseline networks ResNet34/50 by using 300 proposals. In addition, by using only 50 proposals, our method can detect at 20 ft/s on a single GPU and reach 92% of the above performance, hence the potential for real-time detection.

Research Progress of Automated Visual Surface Defect Detection for Industrial Metal Planar Materials

“…Most of the strip surface defect detection algorithms outlined in this section are tested on exposed offline databases. For example, in the literature [ 127 ], Song et al applied an end-to-end defect detection network (DDN), which combined ResNet and RPN for accurate defect detection and location and tested them on the database NEU defect detection dataset (NEU-DET). The experiment indicates that the mean average precision of DDN for defect detection task reaches 82.3.…”

A Light-Weight Deep-Learning Model with Multi-Scale Features for Steel Surface Defect Classification

“…The adjacent resolution is further processed by the global context attention technique. A new defect classification system based on deep learning was proposed by [ 20 ], where multiple features of deep learning are fused into one feature. Then, the regions of interest are produced by a region-based network.…”

“…Deep learning models, mostly in the form of Convolutional Neural Networks (CNNs), have been widely used for image classification [ 4 ], object detection [ 20 ], and text understanding [ 23 ]. CNNs are well known due to the advantages like high accuracy, high efficiency (with few layers), and ease-to-use (end-to-end).…”

A Deep-Learning-based 3D Defect Quantitative Inspection System in CC Products Surface

“…The MV merits make it more competitive in harsh environment application like CC manufacturing field [12,13]. On the other hand, MV-based 3D optical metrology has gradually demonstrated superiority, such as [14][15][16] stereoscopy triangulation (mm), interferometry (nm), con-focal vertical scanning, and fringe projection (um). ArcelorMittal Corp. developed a conoscopic holography rangefinders system tested in ACERALIA Crop.…”