Supervised Machine Learning Based Surface Inspection by Synthetizing Artificial Defects

The preparation of labeled training data for supervised machine learning methods involves a lot of effort. Regarding surface inspection tasks, this endeavor is often not economically reasonable. In this paper, an artificial defect synthetization algorithm based on a multistep stochastic process is proposed. It adds defects to fault-free surface images, which can be used for supervised machine learning. By this means a deep convolutional neural network has been trained, achieving a detection rate of 94% of occurring real defects on the presented test surface.

[1]  Chen Sun,et al.  Revisiting Unreasonable Effectiveness of Data in Deep Learning Era , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[2]  V. Sugumaran,et al.  Machine learning approach for automated visual inspection of machine components , 2011, Expert Syst. Appl..

[3]  Gerald Berger,et al.  Measurement of the visual perceptibility of sink marks on injection molding parts by a new fast processing model , 2014 .

[4]  Peter Norvig,et al.  The Unreasonable Effectiveness of Data , 2009, IEEE Intelligent Systems.

[5]  Peter Waszkewitz,et al.  Industrial Image Processing: Visual Quality Control in Manufacturing , 1999 .

[6]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.

[7]  David A. Clifton,et al.  A review of novelty detection , 2014, Signal Process..

[8]  Edwin Lughofer,et al.  On improving performance of surface inspection systems by online active learning and flexible classifier updates , 2015, Machine Vision and Applications.

[9]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[10]  Matthias Haselmann,et al.  Anomaly detection on arbitrarily distorted 2D patterns by computation of a virtual golden sample , 2016, 2016 IEEE International Conference on Image Processing (ICIP).

[11]  J. Aisa,et al.  Characterization of In-Mold Decoration Process and Influence of the Fabric Characteristics in This Process , 2011 .

[12]  Martín Abadi,et al.  TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems , 2016, ArXiv.