SmartScan | Algotechniq

Objective

The goal of this project is to create a large-scale metallic surface defect detection system that is more robust and accurate in real-world industrial environments. This is achieved by using datasets that include defect types collected from real industry situations, and by proposing an end-to-end defect detection network that combines inferences from several deep learning and data augmentation techniques.

Illustrating typical defects on a metallic surface.

Prior Work

Previously, defect detection for metallic surfaces relied on traditional image processing algorithms that utilized hand-crafted features, such as local binary patterns (LBP) and histogram of oriented gradients (Cumbajin et al., 2023). These methods, while effective in controlled environments, performed poorly in varying industrial applications due to their sensitivity to environmental factors like lighting and background clutter. Deep learning methods were also used (Arikan et al., 2019)., but they were often limited by the scale and diversity of the datasets available for training.

Benefits

Industries that require high-quality control for metallic products will benefit from the new approach, particularly those involved in manufacturing processes where surface defects can significantly impact product quality. The improved robustness and accuracy of the defect detection system will help in reducing manual inspection costs and improving the overall efficiency of the production line.

Enabling Technologies

The key enabling technologies of the new approach include:

Faster R-CNN: High accuracy using a two-stage process with region proposals, but slower than the older Single Shot MultiBox Detector (SSD).
YOLO: Fast and suitable for real-time detection, but less accurate for small objects and dense scenes.
RetinaNet: Balances speed and accuracy, with improved detection of small objects using focal loss.
EfficientDet: Optimized for both speed and accuracy, scalable for various resource constraints.
Data augmentation methods, which are used to expand the training dataset and improve the model’s robustness to various defect shapes and sizes.

Each offers different trade-offs between speed, accuracy, and complexity.

Quantifying Results

The success of our approach can be quantified using performance metrics such as Recall, Average Precision (AP), and mean Average Precision (mAP) across different defect categories. The approach’s effectiveness is demonstrated by its superior performance on these metrics compared to previous methods, as well as its ability to detect defects in real-time with a fast processing speed per image.

References

2023

A Systematic Review on Deep Learning with CNNs Applied to Surface Defect Detection

Esteban Cumbajin, Nuno Rodrigues, Paulo Costa, and 7 more authors

Journal of Imaging, Oct 2023

Abs HTML

Surface defect detection with machine learning has become an important tool in industries and a large field of study for researchers or workers in recent years. It is necessary to have a simplified source of information that helps us to better focus on one type of surface. In this systematic review, we present a classification for surface defect detection based on convolutional neural networks (CNNs) focused on surface types. Findings: Out of 253 records identified, 59 primary studies were eligible. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we analyzed the structures of each study and the concepts related to defects and their types on surfaces. The presented review is mainly focused on finding a classification for the types of surfaces most used in industry (metal, building, ceramic, wood, and special). We delve into the specifics of each surface category, offering illustrative examples of their applications within both industrial and laboratory settings. Furthermore, we propose a new taxonomy of machine learning based on the obtained results and collected information. We summarized the studies and extracted the main characteristics such as type of surface, problem types, timeline, type of network, techniques, and datasets. Among the most relevant results of our analysis, we found that the metallic surface is the most used, as it is the one found in 62.71% of the studies, and the most prevalent problem type is classification, accounting for 49.15% of the total. Furthermore, we observe that transfer learning was employed in 83.05% of the studies, while data augmentation was utilized in 59.32%. Our findings also provide insights into the cameras most frequently employed, along with the strategies adopted to address illumination challenges present in certain articles and the approach to creating datasets for real-world applications. The main results presented in this review allow for a quick and efficient search of information for researchers and professionals interested in improving the results of their defect detection projects. Finally, we analyzed the trends that could open new fields of study for future research in the area of surface defect detection.

2019

Surface defect classification in real-time using convolutional neural networks

Selim Arikan, Kiran Varanasi, and Didier Stricker

arXiv preprint arXiv:1904.04671, Oct 2019

HTML PDF