Surface defect saliency of magnetic tile

• Published in: The Visual computer Vol. 36; no. 1; pp. 85 - 96
• Main Authors: Huang, Yibin, Qiu, Congying, Yuan, Kui
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.01.2020; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Computer Graphics; Computer Science; Computer vision; Datasets; Defects; Image contrast; Image processing; Image Processing and Computer Vision; Neural networks; Original Article; Salience; Surface defects; Tiles; Surface defect; Convolutional network; Saliency detection
• ISSN: 0178-2789; 1432-2315
• DOI: 10.1007/s00371-018-1588-5

Computer vision builds a connection between image processing and industrials, bringing modern perception to the automated manufacture of magnetic tiles. In this article, we propose a real-time model called MCuePush U-Net, specifically designed for saliency detection of surface defect. Our model consists of three main components: MCue, U-Net and Push network. MCue generates three-channel resized inputs, including one MCue saliency image and two raw images; U-Net learns the most informative regions, and essentially it is a deep hierarchical structured convolutional network; Push network defines the specific location of predicted surface defects with bounding boxes, constructed by two fully connected layers and one output layer. We show that the model exceeds the state of the art in saliency detection of magnetic tiles, in which it both effectively and explicitly maps multiple surface defects from low-contrast images. The proposed model significantly reduces time cost of machinery from 0.5 s per image to 0.07 s and enhances detection accuracy for image-based defect examinations.