The crack initiation stress threshold is widely used in excavation industries as rock spalling strength when designing deep underground structures to avoid unwanted brittle failures. While various ...strain-based methods have been developed for the estimation of this critical design parameter, such methods are destructive and often requires subjective interpretations of the stress–strain curves, particularly in rocks with pre-existing microcracks or high porosity. This study explore the applicability of non-destructive ultrasonic through-transmission methods for determining rock damage levels by assessing the changes in transmitted signal characteristics during loading. The change in velocity, amplitude, dominant frequency, and root-mean-square voltage are investigated with four different rock types including marble, sandstone, granite, and basalt under various stress levels. Results suggest the rate of signal variations can be reliably used to estimate crack closure and crack initiation stress levels across the tested rocks before failure. Comparison of the results between the conventional techniques and the new proposed methods based on ultrasonic monitoring are further discussed.
•Ultimate strength of nailed samples decreased by 40 % after 34 weeks of brown rot exposure.•Ultimate strength of nailed samples decreased by 50–60 % after 34 weeks of white rot exposure.•Connection ...stiffness experienced no statistically significant changes.•The loss of ultimate strength was aligned with the loss of reported wood compression.
The effects of wetting and exposure to two decay fungi were evaluated on nailed slash pine sapwood timber assemblies. Sterilization had no significant effect on connection properties, but exposure to a brown (Fomitopsis ostreiformis) or a white rot fungus (Pycnoporus coccineus) was associated with 40–60 % losses in ultimate strength over 34 weeks. Attempts to more closely assess changes in wood around the fasteners using localized Janka hardness or longitudinal compression were confounded by density differences between earlywood and latewood. The results highlight the effects of decay on connector behaviour and suggest the need for more precise methods for characterizing the environment around the fastener.
3D face recognition (FR) has become increasingly widespread due to the illumination invariance and pose robustness of 3D face data. Most existing 3D FR methods can only achieve excellent performance ...on complete and smooth faces. However, low-quality 3D FR with missing facial features still suffers from insufficient discriminative feature extraction for visible face regions. In this paper, we propose a dual-stream multi-scale fusion network (DSNet) for low-quality 3D FR. First, in the first stream, we design a new multi-scale local and global feature fusion network, which consists of an enhanced shallow feature extraction module, an enhanced deep feature extraction module, and a layered multi-scale feature correlation fusion module, aiming to obtain more discriminative details and category information of the facial visible region, reducing the interference of similar features and the redundancy of the same features. Second, we also introduced a capsule network as the second stream to enhance the expression of 3D facial spatial position information, thereby further improving the performance of low-quality 3D FR with missing facial features. We conduct extensive experiments on low-quality datasets (Lock3DFace, KinectFaceDB, and IIIT-D) and cross-quality datasets synthesized by Bosphorus. These results show that our proposed DSNet can achieve state-of-the-art recognition performance and exhibit excellent performance on low-quality 3D faces with missing facial features.
The Moon-based Ground Penetrating Radar (GPR) carried by the Yutu rover performed in-situ radar measurements to explore extraterrestrial objects, which provides an unprecedented opportunity to study ...the shallow subsurface structure of the Moon and its internal resources. Exploiting lunar resources might be one of the solutions to the Earth's energy shortage in the future. In this paper, firstly, the thickness distribution of the lunar regolith at the Chang'E-3 landing site is derived using the high-frequency Yutu radar observation data. Secondly, the surface concentration of helium-3 is determined based on the surface TiO<inline-formula><tex-math notation="LaTeX">_{2}</tex-math></inline-formula> content of the lunar regolith. Finally, the reservoir of helium-3 resources in the lunar regolith is estimated. Our result suggests that the helium-3 reservoir along the Yutu rover traveling route from the navigation points N105 to N208 (<inline-formula><tex-math notation="LaTeX">\sim</tex-math></inline-formula>445 m<inline-formula><tex-math notation="LaTeX">^{2}</tex-math></inline-formula>) is <inline-formula><tex-math notation="LaTeX">\sim</tex-math></inline-formula>37-51 g, and its helium-3 content per unit area is <inline-formula><tex-math notation="LaTeX">\sim</tex-math></inline-formula>0.083-0.114 g/m<inline-formula><tex-math notation="LaTeX">{^{2}}</tex-math></inline-formula>, which is at least 5 times higher than that of the global average. Currently, the nuclear fusion experiment is facing a severe shortage of tritium fuel. We discuss the possibility of replacing it with lunar helium-3 as the fuel for nuclear fusion. Meanwhile, we also suggest that the Chang'E-3 landing area can be a potential site selection for the exploitation of the lunar helium-3 in the future. Our results will provide a valuable reference to evaluate the economics and feasibility of mining in-situ helium-3 resources on the Moon.
Visual and audio signals are concurrent and complementary types of modality in some video actions. A single visual modality limits the performance of video action recognition due to the similar ...appearance with subtle movements, such as tapping guitar and playing guitar. However, it is challenging to fuse audio-visual modalities since the heterogeneity gap caused by inconsistent distribution and representation of multi-modal data. In this paper, we propose a local-to-global multi-modal interaction network (LGMI-Net) that integrates RGB, optical flow with sound information. First, for the local multi-modal interaction, we propose a novel inter-modal channel recalibration (IMCR) block to learn a joint representation from different input modalities by recalibrating the channel information distribution of one modality according to another modality. Besides, we also propose a novel RGB modality aggregation (RMA) block to obtain the more robust appearance features by mixing optical flow and sound information. Second, for the global multi-modal interaction, we propose three distinct encoder modes: unitary, parallel and triplet encoders, to capture the global multi-modal representation. The unitary encoder has the lowest computational complexity. The parallel encoder utilizes RGB-guided attention to improve accuracy while maintaining the lightweight. The triplet encoder aggregates the self-attentions of different modalities and achieves the best recognition performance. We implement extensive experiments on three public datasets: UCF101 subset, Kinetics-Sounds and EPIC-Kitchens-55. The results demonstrate the effectiveness of audio-visual complementation. Compared with the state-of-the-art multi-modal methods with sound (i.e. MM-ViT, AdaMML and G-Blend), the proposed LGMI-Net achieves superior accuracies of 96.05%, 88.37% and 50.3% with the 4.49×, 1.40× and 6.02× lower giga floating point operations (GFLOPs) respectively.
Human action recognition (HAR) refers to the process in which computers analyze and process video data to obtain the categories of action presented in the video. It has a wide range of applications, ...such as video surveillance, human–computer interaction, and autonomous driving. The spatio-temporal features required for video analysis are typically extracted from the RGB pixels, which entail significant computational complexity and make it challenging for real-time action recognition. However, in the compressed domain, sparse representations, such as motion vectors, quantization parameters, transform coefficients, and residuals provide the comparable scene semantic information with reduced complexity. This paper provides an overview of the research efforts in action recognition based on the compressed domain. It includes a comprehensive review of both traditional and deep learning-based methods published between 2000 and 2023, focusing on compression standards and compression parameters. Specifically, we first summarize the compression standards and compressed algorithms used for compressing videos. Furthermore, we classify compressed-domain action recognition methods into traditional and deep learning-based approaches. Thirdly, we introduce public datasets and evaluation metrics, analyze the characteristics of these methods and compare their performance. Finally, we highlight challenges and suggest future research directions.
The Pearl River Delta (PRD) region in South China is abundant of geothermal water resources whereas geochemical evolution of the geothermal water is still poorly understood. In this study, geothermal ...water is sampled to investigate the provenance of CO2 and sulphate, residence time, recharge elevation, circulation depth, and water-interaction processes by analyzing chemistry of major ions and isotope geochemistry. The range of δ13C and content of bicarbonate reveal multiple sources of soil CO2, carbonate rocks, and small amounts of mantle CO2. The δ34S ratio and sulphate concentration are attributed to evaporites dissolution, input of atmospheric sulphur, seawater intrusion, and oxidation of pyrite. The residence time of up to 21.01 ka for geothermal water likely suggests the earliest precipitation recharge during the Late Pleistocene. Mountains or/and hills at an altitude of 300–760 m in north PRD region are currently the major recharge areas for the geothermal water which can circulate at the depths of up to 3000–4800 m along regional structures. The isotopic and hydrochemical analysis confirms that extensive water interaction with silicate rock, carbonate rock and evaporite, evaporation-crystallization process, and cation exchange are the key mechanisms that control the geochemical evolution of geothermal water in the PRD geothermal system.
Display omitted
•This study presents the first detailed C-S isotope systematics in the PRD region.•The meteoric recharge can date back to 21 ka ago during the Late Pleistocene.•Small amounts of mantle CO2 is being entrained by the PRD geothermal system.
Image captioning refers to automatic generation of descriptive texts according to the visual content of images. It is a technique integrating multiple disciplines including the computer vision (CV), ...natural language processing (NLP) and artificial intelligence. In recent years, substantial research efforts have been devoted to generate image caption with impressive progress. To summarize the recent advances in image captioning, we present a comprehensive review on image captioning, covering both traditional methods and recent deep learning-based techniques. Specifically, we first briefly review the early traditional works based on the retrieval and template. Then deep learning-based image captioning researches are focused, which is categorized into the encoder-decoder framework, attention mechanism and training strategies on the basis of model structures and training manners for a detailed introduction. After that, we summarize the publicly available datasets, evaluation metrics and those proposed for specific requirements, and then compare the state of the art methods on the MS COCO dataset. Finally, we provide some discussions on open challenges and future research directions.
The existing frequency-based action recognition methods achieve impressive performance in improving efficiency. However, they ignore the low-frequency texture and edge clues, leading to accuracy ...degradation. To address this problem, we propose a novel frequency enhancement (FE) block for efficient compressed video action recognition, including a temporal-channel two-heads attention (TCTHA) module and a frequency overlapping group convolution (FOGC) module. First, the TCTHA module emphasizes the inter-frame temporal context and the inner-frame informative frequency semantics by attention. Then, the FOGC module groups channels in different frequency bands with overlap, to extract low-frequency texture and edge clues, while maintaining the interaction of groups. We integrate the FE block into 2D-CNNs with frequency I-frame input, termed FENet, focusing on the pivotal low-frequency spatio-temporal semantics for action recognition. Experiments on HMDB-51, UCF-101, Kinetics-400, and Kinetics-700 verify that our FENet achieves comparable accuracy compared with RGB-based methods with high efficiency.
PDF
