RÉSUMÉ
Different from the traditional frequency-mixing technique which employs a contacting transducer, the laser-induced acoustic nonlinear frequency-mixing detection technique utilizes a laser source to instigate crack motion and generate acoustic waves. Thus, apart from the temperature oscillation induced by the pump laser, the "basic temperature" originating from the probe laser can also influence the crack. This additional variable complicates the contact state of the crack, yielding a more diverse range of nonlinear acoustic signal attributes. In light of this, our study enhances the conventional opto-acoustic nonlinear frequency mixing experimental setup by integrating an independent heating laser beam. This modification isolates the impact of the "basic temperature" on crack width while also dialing down the probe laser power to mitigate its thermal effects. To amplify the sensitivity of crack detection, we deliberated on the optimal laser source parameters for this setup. Consequently, our revamped system, paired with fine-tuned parameters, captures nonlinear acoustic signals with an enriched feature set. This investigation can provide support for the non-contact opto-acoustic nonlinear frequency mixing technique in the detection and evaluation of micro-cracks.
RÉSUMÉ
Compared with the progress made on human activity classification, much less success has been achieved on human interaction understanding (HIU). Apart from the latter task is much more challenging, the main causation is that recent approaches learn human interactive relations via shallow graphical representations, which are inadequate to model complicated human interactive-relations. This paper proposes a deep consistency-aware framework aiming at tackling the grouping and labelling inconsistencies in HIU. This framework consists of three components, including a backbone CNN to extract image features, a factor graph network to implicitly learn higher-order consistencies among labelling and grouping variables, and a consistency-aware reasoning module to explicitly enforcing consistencies. The last module is inspired by our key observation that the consistency-aware reasoning bias can be embedded into an energy function or a particular loss function, minimizing which delivers consistent predictions. An efficient mean-field inference algorithm is proposed, such that all modules of our network could be trained in an end-to-end fashion. Experimental results demonstrate that the two proposed consistency-learning modules complement each other, and both make considerable contributions in achieving leading performance on three benchmarks of HIU. The effectiveness of the proposed approach is further validated by experiments on detecting human-object interactions.
Sujet(s)
Algorithmes , Apprentissage , Humains , RéférenciationRÉSUMÉ
Herein, we studied the increasing tendency of photoacoustic (PA) conversion efficiency of the Au/polydimethylsiloxane (PDMS) composite. The thickness of the Au layer was optimized by modeling the PA process based on the Drude-Lorentz model and finite element analysis method, and corresponding results were verified. The results showed that the optimal Au thickness of the Au/PDMS composite was 35 nm. Finally, the Au/PDMS composites were coated onto the surface of aluminum alloys, which improved the thermoelastic laser ultrasonic (LU) signals to near 100 times. Besides, the defect mapping was performed by thermoelastic LU signals with Au/PDMS coating and ablation LU signals without coating; the Pearson correlation coefficient was higher than 0.95. The application in the defect detection in metal could provide guides for nondestructive detection on metals by laser ultrasound.
RÉSUMÉ
The Synthetic Aperture Focusing Technique (SAFT) is an imaging algorithm used in laser ultrasonics (LU) to visualise the appearance of defects. However, ultrasound excited by a pulsed laser has the characteristics of wide bandwidth and multi-mode directivity patterns, leading to common problems in the SAFT process, such as low utilisation of ultrasound information and possible artefacts. To solve these problems, a Multi-mode Time-domain SAFT (MMT-SAFT) algorithm is proposed in this paper. The influence of ultrasonic directivity is discussed according to the imaging depth range, and imaging with multiple LU modes is performed to reduce artefacts. Simulations and experimental results prove the feasibility of the MMT-SAFT algorithm, which not only presents a clearer image of the upper part of defects but also improves image quality compared with time-domain SAFT using a single ultrasonic mode. The proposed technique can provide feasible directions for laser ultrasonic defect imaging.
RÉSUMÉ
Named entity disambiguation (NED) finds the specific meaning of an entity mention in a particular context and links it to a target entity. With the emergence of multimedia, the modalities of content on the Internet have become more diverse, which poses difficulties for traditional NED, and the vast amounts of information make it impossible to manually label every kind of ambiguous data to train a practical NED model. In response to this situation, we present MMGraph, which uses multimodal graph convolution to aggregate visual and contextual language information for accurate entity disambiguation for short texts, and a self-supervised simple triplet network (SimTri) that can learn useful representations in multimodal unlabeled data to enhance the effectiveness of NED models. We evaluated these approaches on a new dataset, MMFi, which contains multimodal supervised data and large amounts of unlabeled data. Our experiments confirm the state-of-the-art performance of MMGraph on two widely used benchmarks and MMFi. SimTri further improves the performance of NED methods. The dataset and code are available at https://github.com/LanceZPF/NNED_MMGraph.
RÉSUMÉ
The Laser Ultrasonic (LU) technique has been widely studied. Detected ultrasonic signals can be further processed using Synthetic Aperture Focusing Techniques (SAFTs), to detect and image internal defects. LU-based SAFT in frequency-domain (F-SAFT) is developed to visualize horizontal hole-type defects in aluminum. Bulk acoustic waves are non-destructively generated by irradiating a laser line-source, and detected using a laser Doppler vibrometer at a point away from the generation. The influence of this non-coincident generation-detection on the equivalent acoustic velocity used in the algorithm is studied via velocity mappings. Because the wide-band generation characteristic of the LU technique, frequency range selections in acoustic wave signals are implemented to increase Signal-to-Noise Ratio (SNR) and reconstruction speed. Results indicate that by using the LU F-SAFT algorithm, and incorporating optimizations such as velocity mapping and frequency range selection, small defects can be visualized in 3D with corrected locations and improved image quality.