Rapid Transformation Estimation Using Deep Learning for Defect Detection

Autor:	Xishi Huang, Jing Ren
Rok vydání:	2021
Předmět:	Standard test image Artificial neural network Computer science business.industry Deep learning ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Image registration Pattern recognition Image (mathematics) Wavelet Transformation (function) Artificial intelligence business Image resolution
Zdroj:	ICAIIS
DOI:	10.1145/3469213.3469219
Popis:	Defect detection is a crucial step in the manufacturing of vehicle parts such as the engine. One major method for defect detection is to use image registration and image difference to identify and segment the defects. The key technology of this approach is to extract the accurate transformation information between the template image and the testing images. In this paper, we propose a novel deep neural network (DNN) method to learn the transformations from the training dataset. Wavelet transformation is introduced to denoise the images and reduce the image size for fast image registration. The results show that the trained DNN models are able to effectively predict the transformation between the template image and the actual test image in real time with high accuracy.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::c1adab54b43b4e4ab4179eb0370e62e8 https://doi.org/10.1145/3469213.3469219 Zobrazit plný text záznamu