27 July 2023 in IEEE Transactions on Circuits and Systems for Video Technology
Recently, deep learning has been widely employed to improve the quality of low-light videos. However, most existing low-light video enhancement methods fail to effectively explore temporal dependence, and the enhanced videos may suffer from severe noise, loss of detailed texture, and temporal inconsistency. In this paper, we propose a novel SNR-prior Guided Trajectory-aware Transformer (SGTT) to enable effective video representation learning for low-light video enhancement. Specifically, signal-to-noise ratio prior and cosine similarity are introduced to build the trajectory-aware dual-attention for exploiting the dependence of long-range spatio-temporal information, which searches for sharper and highly correlated patches within the same trajectory to assist in enhancing the target frames. Moreover, to adaptively fuse spatio-temporal information of support frames propagated bidirectionally, an attention-guided spatio-temporal feature aggregation module is proposed to perceive and enhance the specific high-quality features. The evaluation of both dynamic and static videos shows the effectiveness of our network, which significantly outperforms the state-of-the-art methods.
1 October 2021 in Minerals Engineering
25 March 2024 in Expert Systems with Applications
1 September 2021 in Insight - Non-Destructive Testing and Condition Monitoring
This study proposes a weld defect inspection method based on a combination of machine vision and weak magnetic technology to inspect the quality of weld formation comprehensively. In accordance with the principle of laser triangulation, surface information about the weldment is obtained, the weld area is extracted using mutation characteristics of the weld edge and an algorithm for identifying defects with abnormal average height in the weld surface is proposed. Subsequently, a welding seam inspection process is developed and implemented, which is composed of a camera, a structured light sensor, a magnetic sensor and a motion control system. Inspection results from an austenitic stainless steel weldment show that the method combining machine vision and magnetism can identify defect locations accurately. Comprehensive analysis of the test results can effectively classify surface and internal defects, estimate the equivalent sizes of defects and evaluate the quality of weld formation in multiple dimensions.