Classification of industrial surface defects based on neural architecture search
- PDF / 947,085 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 52 Downloads / 198 Views
Classification of industrial surface defects based on neural architecture search Tiejun Yang 1 & Tianshu Zhang 1 & Lin Huang 1 Received: 2 June 2020 / Revised: 28 August 2020 / Accepted: 24 September 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Surface defect classification (SDC) is the visual inspection of the surface of an object to identify appearance defects. Efficient and accurate SDC is i mportant for improving the quality of industrial products. A manually designed convolutional neural network (CNN) is traditionally used for SDC. In this study, a simpler SDC scheme with a higher classification accuracy, named NAS-SDC, is developed based on the neural architecture search (NAS) technique. A max-pooling cell based on NASNet is introduced to reduce the search space and the number of network parameters, thus simplifying the candidate operators for the search. Two network architectures are proposed to stack the search candidates or the best cells. The proposed method can be used to automatically design an efficient CNN model for SDC on a specific dataset. Experimental results show that the proposed method can find the best cells in ~11 h using a single graphics processing unit (GPU) and achieves higher classification accuracies (99.98%, 99.8% and 99.26%) than state-of-the-art methods on the Northeastern University (NEU-CLS), DAGM, and bridge defect datasets. The number of network parameters used in the proposed method is only 0.35 M, and the average test time per sample is approximately 61 ms, thus achieving a balance between performance and speed. Keywords Neural architecture search . Deep convolutional neural networks . Surface defect classification
1 Introduction Surface defects affect the quality of a surface (Fig. 1 shows examples of surface defects on a bridge and a hot-rolled steel strip). In recent years, the use of computer vision technology to automatically detect and classify surface defects has been widely studied [3, 5]. Surface defect * Lin Huang [email protected]
1
Guangxi Key Laboratory of Embedded Technology and Intelligent System, Guilin University of Technology, Guangxi 541004 Guilin, People’s Republic of China
Multimedia Tools and Applications
Fig. 1 Types of surface defects, as marked with red arrows: (a) a bridge crack [18]; (b-c) defects in hot-rolled steel strips [27], specifically, (b) a patch (Pa) and (c) an inclusion (In); and (d) a type of defect appearing in the DAGM dataset [15]
classification (SDC) based on machine learning offers the advantages of high efficiency and accuracy compared with traditional manual inspection. For example, Song and Yan [27] proposed a noise-robust feature descriptor named the adjacent evaluation of completed local binary patterns (AECLBP) and achieved satisfactory classification results using a support vector machine (SVM) model on the Northeastern University (NEU-CLS) defect dataset. Xiao et al. [36] first performed feature extraction based on the gray-level co-occurrence matrix and the Gabor filter and then empl
Data Loading...
