Learning Defect Classifiers for Textured Surfaces Using Neural Networks and Statistical Feature Representations

Weimer, D. R.; Thamer, Hendrik; Scholz‐Reiter, Bernd

doi:10.1016/j.procir.2013.05.059

Cited by 40 publications

(15 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The focus is primarily on deep-learning methods that have, in recent years, become the most common approach in the field of computer vision. When applied to the problem of surface-quality control (Chen and Ho 2016;Faghih-Roohi et al 2016;Weimer et al 2013;Kuo et al 2014), deeplearning methods can achieve excellent results and can be adapted to different products. Compared to classical machinevision methods, the deep learning can directly learn features from low-level data, and has higher capacity to represent complex structures, thus completely replacing handengineering of features with automated learning process.…”

Section: Introductionmentioning

confidence: 99%

Segmentation-based deep-learning approach for surface-defect detection

et al. 2019

View full text Add to dashboard Cite

Automated surface-anomaly detection using machine learning has become an interesting and promising area of research, with a very high and direct impact on the application domain of visual inspection. Deep-learning methods have become the most suitable approaches for this task. They allow the inspection system to learn to detect the surface anomaly by simply showing it a number of exemplar images. This paper presents a segmentation-based deep-learning architecture that is designed for the detection and segmentation of surface anomalies and is demonstrated on a specific domain of surface-crack detection. The design of the architecture enables the model to be trained using a small number of samples, which is an important requirement for practical applications. The proposed model is compared with the related deep-learning methods, including the state-ofthe-art commercial software, showing that the proposed approach outperforms the related methods on the specific domain of surface-crack detection. The large number of experiments also shed light on the required precision of the annotation, the number of required training samples and on the required computational cost. Experiments are performed on a newly created dataset based on a real-world quality control case and demonstrates that the proposed approach is able to learn on a small number of defected surfaces, using only approximately 25-30 defective training samples, instead of hundreds or thousands, which is usually the case in deeplearning applications. This makes the deep-learning method practical for use in industry where the number of available defective samples is limited. The dataset is also made publicly available to encourage the development and evaluation of new methods for surface-defect detection.

show abstract

Section: Introductionmentioning

confidence: 99%

Segmentation-based deep-learning approach for surface-defect detection

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Therefore, non-maximum suppression technique is used. Figure 2 displays an overall Defect detection framework [22].…”

Section: Defect Detection Methodsmentioning

confidence: 99%

“…Learning procedure with patch extraction and feature representation[22] Number of patches p with size s is used for learning instead of using the whole image, as shown inFigure 2, In the first stage random patches are generated, in the second stage patch layoutis displayed, where each image patch shows a row vector and at last features are represented. Remember that (x i ,y i ) represents the i th example of learning with 1 and 0 for defected and non-defected pictures respectively.…”

mentioning

confidence: 99%

Deep Convolution Neural Network Approach for Defect Inspection of Textured Surfaces

Garg¹,

Dhiman²

2020

JIEC

View full text Add to dashboard Cite

Defect Inspection of Textured Surfaces is a challenging problem which occurs during manufacturing in many processing phases. With arbitrary length, shape and orientation, these defects occur. Moreover, there are fewer and different photos of defective products. Deep Convolution Neural Network (CNN) has an impressive development in target detection, and better results have been obtained with the implementation of deep CNN design for texture detection. Nonetheless, with the growing detection accuracy of deep CNNs, there are the drawbacks of significantly increasing computational costs and processing resources, which seriously hinders CNN's use in resource-limited environments such as mobile or embedded phones. In this paper, a novel framework is proposed that uses raw image database patch statistics joint with two layers of neural network for surface defect detection. For defect detection, a convolution neural network (CNN) classifier is used. Imaging analysis of training samples using Deep Convolution Neural Network (CNN) is used to find the defect in an image's target area. In point of energy saving, the results of the experiment show that proposed method has numerous advantages in terms of reduction in time and cost. It also shows the high-performance contrast to conventional manual inspection process with less repetition and helps to build the object detection classifier with high generalization potential and high detection accuracy.

show abstract

“…These extracted features are sometimes specific to a particular product and may need to be manually adapted to different products. 16 On the other hand, CNN-based deep learning image analytics hold the potential to be quickly adapted to a new product and do not have limitations for a large-scale inspection, which is necessary for industries to generalize these methods into the production line rapidly. 9 CNNs can directly learn features from complex structures and are capable of being adapted to a new product, 17 thereby holding the potential to replace handcrafted feature-based learning programs with automated processes.…”

Section: Introductionmentioning

confidence: 99%