Molten image fusion and enhancement based on image decomposition in frequency domain
- PDF / 2,439,954 Bytes
- 9 Pages / 595.276 x 790.866 pts Page_size
- 5 Downloads / 255 Views
ORIGINAL PAPER
Molten image fusion and enhancement based on image decomposition in frequency domain Linli Xu1 · Jinru Hang1 · Jing Han1 · Tian Wang2 · Lianfa Bai1 Received: 7 September 2019 / Revised: 8 February 2020 / Accepted: 13 July 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract In this paper, we propose a novel molten image enhancement and fusion method based on image decomposition in frequency domain. The algorithm combines the guided filter to maintain the original edge and details and to make it show a more permeable visual effect. Firstly, the high-quality molten imaging band is obtained by analyzing the characteristic spectra of welding materials and arcs. We choose the bands with more spectra feature of materials and weaker arc interference as our optical path channels and collect color images at the same moment to obtain more molten information. Then, after a series of image preprocessing, we combine the details extraction strategy and the guided filter together to yield a novel fusion algorithm, which can make the fusion result to have rich information, clearer edge and higher contrast. Finally, the experimental results show that the proposed method has obvious advantages over some existing methods from the objective and the subjective view. Keywords Molten pool vision · Image decomposition in frequency domain · Image fusion · Dual optical path
1 Introduction With the welding robot widely used in the welding process, welding quality is no longer to be judged by welder’s eyes. Since the camera can obtain the surface information of the molten pool and timely feedback the welding results through online monitoring system, it is of great significance to study the online molten image for the welding quality evaluation, Linli Xu and Jinru Hang have contributed equally to this paper.
B
Lianfa Bai [email protected] Linli Xu [email protected] Jinru Hang [email protected] Jing Han [email protected] Tian Wang [email protected]
1
Jiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, China
2
School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China
monitoring and prediction. In order to obtain the discontinuous information, a lot of work has been done. For example, the edge of the image is enhanced to obtain the contour of the molten pool [1], a tracking method for weld crack detection based on fusion image with different characteristics is proposed [2], and the real-time tracking of weld seam is carried out by image processing [3]. All these methods above are based on the information collection and image processing for the weld. Thereby, it is a prerequisite for all kinds of weld monitoring to obtain high-quality weld images in real time. At present, passive vision sensing always uses composite filter system instead of monocular camera to obtain image with weak arc interference and strong radiation of the weld. Xu et al. [4] propose a low-priced pass
Data Loading...
