FDnCNN-based image denoising for multi-labfel localization measurement. (February 2020)

• Record Type: Journal Article
• Authors: Li, Lin; Yu, Xiaolei; Jin, Zhixin; Zhao, Zhimin; Zhuang, Xiao; Liu, Zhenlu
• Is Part Of: Measurement. Volume 152(2020)
• Publication Date: 2020-02
• Subjects: Image denoising -- Multi-label localization -- Convolutional neural network -- Visual measurement; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.107367
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: The dynamic measurement system is designed with wide field and anti-interference. Image denoising based on FDnCNN can obtain high-quality images in real time. The mapping from image pixel to world space is realized by shaft alignment method. 3D multi-label location based on visual measurement is proposed. Abstract: Recently, the discriminant learning denoising model has deserved special attention due to its outstanding denoising property. This paper designs the RFID multi-label dynamic localization system based on visual measurement to avoid electromagnetic interference. Firstly, the potential sharp images are estimated via denoising images acquired by dual CCD. Then the labels are matched by template. Finally, Shaft Alignment Method (SAM) models the multi-label 3D coordinates in world space and image pixel. Flexible Feed-forward Denoising Convolutional Neural Network (FDnCNN) is proposed to reach the equilibrium between denoising effect and image particulars without producing artifacts in GPU effectively and flexibly. FDnCNN increases the image quality by at least 0.5 dB than WNNM and DnCNN. The 3D coordinate measurement system is evaluated by error, which proves the feasibility and effectiveness of the positioning system.