Exploiting sparsity for image-based object surface anomaly detection

Chai, Woon Huei; Ho, Shen-Shyang; Goh, C. J.

doi:10.1109/icassp.2016.7472024

Cited by 2 publications

(20 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is useful for the domain where data of each class can be represented by independent identically distributed (i.i.d) Gaussian vector and the means of data of all classes are much closed to each other; for example, face recognition. However, it is not useful for the domain where the data of each class has significantly different means such as viewpoint 1 classification for anomaly detection [5], [6].…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

“…It has been used in image-based surface anomaly detection where the dictionary atoms are constructed using the images from various viewpoints of an object [5], [6]. As the vectorized images from the same viewpoint with different illumination conditions can be represented by a bouquet model as face images.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

“…Anomaly detection has been used in many fields such as automated visual inspection [5], [6], network intrusion detection, fraud detection, fault detection, system health monitoring, event detection in sensor networks, and detecting ecosystem disturbances, etc. Supervised anomaly detection techniques are used in the problems where the definition of both "normal" and "abnormal" classes are clear.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

“…Supervised anomaly detection techniques are used in the problems where the definition of both "normal" and "abnormal" classes are clear. They can be defined by labelled ("normal" and "abnormal") data sets [5], [6]. A novel dictionary-based sparse reconstruction anomaly detection technique [5], [6], inspired by a face recognition technique, is proposed.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

“…They can be defined by labelled ("normal" and "abnormal") data sets [5], [6]. A novel dictionary-based sparse reconstruction anomaly detection technique [5], [6], inspired by a face recognition technique, is proposed. The approach  identifies anomalies in an input data based on the sparse residuals (errors) obtained during data reconstruction using sparse representation, and  learns the threshold to classify an element of the input data to be either a normal or an anomalous element based on its residual.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Novel techniques for sparse representation problems

Chai¹

Self Cite

View full text Add to dashboard Cite

This thesis contains material from papers published in the following peer-reviewed conferences and publication drafts under preparation for the publications in the following peer-reviewed journals where I was the first and corresponding author. Chapter 3 is prepared as W.H. Chai, and C. Quek. A novel Quasi-Newton Technique for Composite Convex Minimization. Pattern Recognition (Under revision) The contributions of the co-authors are as follows:  I defined the initial project direction  Assoc Prof Quek edited the manuscript drafts.  The manuscript was revised by Assoc Prof Quek.  I generated the idea, derived the theory, designed the study and the algorithm, performed all the experimental works, and analyzed the results at the School of Computer Science and Rolls-Royce@NTU Corporate Lab. Chapter 4 is prepared as W.H. Chai, and C. Quek. Representation Recovery via -norm Minimization with Corrupted Data. Information Sciences (Submitted)The contributions of the co-authors are as follows: Assoc Prof Quek provided the initial project direction and edited the manuscript drafts. The manuscript was revised by Assoc Prof Quek. I generated the idea, derived the theory, designed the study and the algorithm, performed all the experimental works, and analyzed the results at the School of Computer Science and Rolls-Royce@NTU Corporate Lab.Chapter 5 is prepared as W.H. Chai, and C. Quek. Feature Selection for Dictionary Transfer. Pattern Recognition Letters (Under preparation for publication)The contributions of the co-authors are as follows:x  I defined the initial project direction  Assoc Prof Quek edited the manuscript drafts. The manuscript was revised by Assoc Prof Quek. I generated the idea, derived the theory, designed the study and the algorithm, performed all the experimental works, and analyzed the results at the School of Computer Science and Rolls-Royce@NTU Corporate Lab.

show abstract

Section: Chapter 1 Introductionmentioning

confidence: 99%

Section: Chapter 1 Introductionmentioning

confidence: 99%

Section: Chapter 1 Introductionmentioning

confidence: 99%

Section: Chapter 1 Introductionmentioning

confidence: 99%

Section: Chapter 1 Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Novel techniques for sparse representation problems

Chai¹

Self Cite

View full text Add to dashboard Cite

show abstract

Distance-Based Anomaly Detection for Industrial Surfaces Using Triplet Networks

Tayeh

Aburakhia

Myers

et al. 2020

2020 11th IEEE Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON)

View full text Add to dashboard Cite

Surface anomaly detection plays an important quality control role in many manufacturing industries to reduce scrap production. Machine-based visual inspections have been utilized in recent years to conduct this task instead of human experts. In particular, deep learning Convolutional Neural Networks (CNNs) have been at the forefront of these image processing-based solutions due to their predictive accuracy and efficiency. Training a CNN on a classification objective requires a sufficiently large amount of defective data, which is often not available. In this paper, we address that challenge by training the CNN on surface texture patches with a distance-based anomaly detection objective instead. A deep residual-based triplet network model is utilized, and defective training samples are synthesized exclusively from non-defective samples via random erasing techniques to directly learn a similarity metric between the same-class samples and outof-class samples. Evaluation results demonstrate the approach's strength in detecting different types of anomalies, such as bent, broken, or cracked surfaces, for known surfaces that are part of the training data and unseen novel surfaces.

show abstract

Exploiting sparsity for image-based object surface anomaly detection

Cited by 2 publications

References 19 publications

Novel techniques for sparse representation problems

Novel techniques for sparse representation problems

Distance-Based Anomaly Detection for Industrial Surfaces Using Triplet Networks

Contact Info

Product

Resources

About