Fusion Paths Research Articles

A Lightweight CER-YOLOv5s Algorithm for Detection of Construction Vehicles at Power Transmission Lines

In the context of power-line scenarios characterized by complex backgrounds and diverse scales and shapes of targets, and addressing issues such as large model parameter sizes, insufficient feature extraction, and the susceptibility to missing small targets in engineering-vehicle detection tasks, a lightweight detection algorithm termed CER-YOLOv5s is firstly proposed. The C3 module was restructured by embedding a lightweight Ghost bottleneck structure and convolutional attention module, enhancing the model’s ability to extract key features while reducing computational costs. Secondly, an E-BiFPN feature pyramid network is proposed, utilizing channel attention mechanisms to effectively suppress background noise and enhance the model’s focus on important regions. Bidirectional connections were introduced to optimize the feature fusion paths, improving the efficiency of multi-scale feature fusion. At the same time, in the feature fusion part, an ERM (enhanced receptive module) was added to expand the receptive field of shallow feature maps through multiple convolution repetitions, enhancing the global information perception capability in relation to small targets. Lastly, a Soft-DIoU-NMS suppression algorithm is proposed to improve the candidate box selection mechanism, addressing the issue of suboptimal detection of occluded targets. The experimental results indicated that compared with the baseline YOLOv5s algorithm, the improved algorithm reduced parameters and computations by 27.8% and 31.9%, respectively. The mean average precision (mAP) increased by 2.9%, reaching 98.3%. This improvement surpasses recent mainstream algorithms and suggests stronger robustness across various scenarios. The algorithm meets the lightweight requirements for embedded devices in power-line scenarios.

RCFNet: Related Cross-level Feature Network with Cascaded Self-distillation for Monocular Depth Estimation

Monocular depth estimation (MDE) is a challenging yet crucial computer vision task, which aims to generate accurate depth maps from a single image. Existing MDE approaches mainly rely on extracting and fusing diverse information from multi-level features to improve prediction accuracy. However, these methods often apply a traditional feature pyramid structure, neglecting a comprehensive exploration of feature fusion paths across multiple levels. Moreover, a single-feature fusion strategy has limited ability to optimize the network. We propose a novel related cross-level feature network (RCFNet) with cascaded self-distillation for monocular depth estimation, including a cross-level feature enhancement (CLFE) module, a hierarchical feature cross refinement (HFCR) module, and a cascaded self-distillation (CSD) module. The CLFE module integrates cross-level features to further exploit the highest-level features, where a channel attention mechanism with hybrid weight operations is deployed to enhance the initial features. The HFCR module adaptively captures strongly correlated complementary information through a window-based multi-head cross-attention mechanism to generate refined features. Meanwhile, the CSD module with a hierarchical feature transformation loss is proposed, which can be viewed as a virtual teacher to progressively extract discriminative features within the network for better gradient flow improvement. Extensive experiments on NYUv2 and KITTI datasets demonstrate that our method outperforms existing SOTA MDE methods in terms of accuracy capacity and robustness.

Weighted parallel decoupled feature pyramid network for object detection

As one of the most representative detectors, feature pyramid network (FPN) has achieved remarkable improvement in object detection. Thanks to its distinctive top-down feature fusion path and multi-scale detection paradigm, FPN has become an essential component in modern detectors and has attracted increasing attention. Nevertheless, since the information of each level feature is not object-specific and the heterogeneity of classification and localization tasks in detection, FPN and its numerous variants endure information redundancy and task conflict. To address these two deficiencies that greatly limit the detection performance, this paper proposes weighted parallel decoupled FPN (WPDFPN). Specifically, selective fusion and elimination (SFE) focusing on internal construction of pyramid is first proposed. Compared to existing feature fusion methods, SFE effectively fuses complementary information between different level features and eliminates redundant information to produce object-specific feature at each pyramid level. Then, a novel weighted parallel FPN (WPFPN) is constructed by investigating the combination manners of feature fusion paths. Different from the sequential bidirectional pyramids, WPFPN performs the top-down and bottom-up paths in parallel and aggregates them using a weighted strategy, adaptively integrating object semantics and details. On this basis, feature decoupling mechanism and decoupled head are both developed to learn task-specific features for classification and localization, thus effectively mitigating the task conflict. Without bells and whistles, WPDFPN achieves 43.6 AP on the MS COCO dataset with ResNet-101 as backbone in Faster R-CNN framework. This achievement suppresses the baseline FPN by 3.1 points and demonstrates competitive performance in comparison with some state-of-the-art detectors. Code is available at https://github.com/HB-X/WPDFPN.

DUGAN: Infrared and visible image fusion based on dual fusion paths and a U-type discriminator

Existing infrared and visible image fusion techniques based on generative adversarial networks (GAN) generally disregard local and texture detail features, which tend to limit the fusion performance. Therefore, we propose a GAN model based on dual fusion paths and a U-type discriminator, denoted as DUGAN. Specifically, the image and gradient paths are integrated into the generator to fully extract the content and texture detail features from the source images and their corresponding gradient images. This incorporation aids the generator in generating fusion results with rich information by integrating output features of dual fusion paths. In addition, we construct a U-type discriminator to focus on input images’ global and local information, which drives the network to generate fusion results visually consistent with the source images. Furthermore, we integrate attention blocks in the discriminator to improve the representation of salient information. Experimental results demonstrate that DUGAN has better performance in qualitative and quantitative evaluation compared with other state-of-the-art methods. The source code has been released at https://github.com/chang-le-11/DUGAN.

A handwritten ancient text detector based on improved feature pyramid network

Text detection is the primary task for digitization of ancient books. Different from the common scene text detection tasks (ICDAR, TotalText, etc.), the texts in handwritten ancient documents are more densely distributed and generally small objects; at the same time, the layout structure is also more complex, with problems such as mixed arrangement of pictures and texts and high background noise, all of which pose challenges for detection. According to the characteristics of ancient book images, this paper proposes a new fusion structure based on Feature Pyramid Networks, and takes FCOS as the baseline model to form a new detector (named RFCOS). We enhance the detection capability for dense and small text instances by adding bottom-up fusion paths, cross-layer connections and weighted fusion. Meanwhile, the loss of high-level feature maps during fusion is reduced by new upsampling method and lateral connections. We verified the effectiveness of our RFCOS on the HWAD (Handwritten Ancient Books Dataset), a dataset containing samples in four languages - Yi, Chinese, Tibetan and Tangut, and verify the generalization of RFCOS on another public dataset MTHv2. The results show that RFCOS outperformed most of the existing text detectors in terms of precision, recall and F-measure.

A Novel Transformer-Based Adaptive Object Detection Method

To accurately detect multi-scale remote sensing objects in complex backgrounds, we propose a novel transformer-based adaptive object detection method. The backbone network of the method is a dual attention vision transformer network that utilizes spatial window attention and channel group attention to capture feature interactions between different objects in complex scenes. We further design an adaptive path aggregation network. In the designed network, CBAM (Convolutional Block Attention Module) is utilized to suppress background information in the fusion paths of different-level feature maps, and new paths are introduced to fuse same-scale feature maps to increase the feature information of the feature maps. The designed network can provide more effective feature information and improve the feature representation capability. Experiments conducted on the three datasets of RSOD, NWPU VHR-10, and DIOR show that the mAP of our method is 96.9%, 96.6%, and 81.7%, respectively, which outperforms compared object detection methods. The experimental results show that our method can detect remote-sensing objects better.

Tribological Properties of Lubricating Grease Additives Made of Silica and Silicon Carbide Nanomaterials

Nanomaterials have been widely used in many fields due to their special microstructure and quantum size effect. In addition, there are many advantages that other traditional materials do not have. Nanometer silicon dioxide and silicon carbide as inorganic nonmetal high-tech materials have high temperature resistance, no pollution and other properties, and are very suitable for making grease additives. At the same time, it is also possible to improve the tribological properties of grease additives by using unique characteristics such as innocuity and tastelessness. Correspondingly, grease is an additive that cannot be obtained in equipment manufacturing. It is very urgent to continuously study and improve the service performance. Therefore, the purpose of this paper is to explore new routes of grease additives and transmission based on the theory of nano-materials, analyze the development and fusion paths in the application of nano-technology technology, and explore the advantages and disadvantages of application. This article will use the research method of specific analysis of specific problems to make data comparison and draw a conclusion. Through theoretical innovation and exploration, we can find a suitable model to promote the development of industrial additives. The research results show that the amount of lubricating grease used in rolling bearings accounts for 80% of the amount of lubricating grease. Adding silica to lubricating grease can improve the stability of additives by 26% and effectively improve the use effect of lubricating grease. Therefore, combining with the characteristics of the current era, nano-technology is used to fully absorb and reform and actively innovate and improve the technological level. This paper analyzes the different application difficulties and explores the development prospects, thus proposing the realization path of improving chemical coatings. These theories are used to guide innovation and integration, providing valuable experience for the wide application of nanotechnology.

DFPGAN: Dual fusion path generative adversarial network for infrared and visible image fusion

Infrared and visible image fusion is an essential task for multi-sensor image fusion. Generative adversarial networks (GAN) have achieved remarkable performance in the fusion of infrared and visible image. Existing GAN based fusion methods merely using infrared and visible image as input for the fusion, while we found that differential images obtained by subtraction between two image sources could provide contrast information for the fusion. To this end, a novel dual fusion path generative adversarial network (DFPGAN) is proposed in this paper for infrared and visible image fusion. We divided the generator of generative adversarial network into two fusion paths namely infrared–visible path and differential path. The input of infrared–visible path concatenated two image sources to make infrared intensity and texture details keep balance fusion in this path. The input of differential path concatenated differential images obtained by subtraction between two image sources to make contrast information fusion in this path. The features extracted by two fusion paths are concatenated at the end of the generator to generate fused images with contrast effect and balanced information distribution. Meanwhile, we have implemented dual self-attention feature refine module (DSAM) on two fusion paths to refine feature maps in two fusion paths. We adopted switchable normalization layer (SN) substitute for batch normalization layer (BN) in the generator and discriminator to avoid fusion artifact. Furthermore, a mixed content loss is integrated in the generator loss functions to guide the generated image keep balanced information distribution and preserving contrast simultaneously. The adversarial training employed dual adversarial architecture to balance the distribution of infrared intensity and texture details. To verifying the improvement effect of fusion image on target detection, we introduce the Scaled-YOLOv4 target detection framework as evaluation framework, and use the proposed network to fuse RGB images and infrared images for target detection. The results of qualitative and quantitative experiments conducted on public datasets demonstrated the superiority of proposed network over other state-of-the-art methods and could generate fused images with distinctly contrast.

Dynamics of 17F + 58Ni reaction via complete and incomplete fusion processes at above barrier energies

The dynamics of 17F + 58Ni reaction induced via a loosely bound projectile (17F) is examined using the collective clusterization approach of the dynamical cluster decay model (DCM) with respect to the recent experimental data available at beam energies Ebeam=54.1 and 58.5 MeV. The calculations are done for quadrupole deformations of fragments using the optimum orientation approach. In view of the loosely bound nature of 17F, the main focus of the present work is on the comparison of complete and incomplete fusion. It is studied using various components such as fragmentation potential, mass distribution, and barrier modification. Different decay modes (ER, IMF, HMF, and fission) are also compared to determine the complete fusion and incomplete fusion paths. Additionally, the decay paths of the nucleus formed from loosely bound (17F) and tightly bound (16O) projectiles are compared. Furthermore, the role of temperature-dependent pairing strength is analyzed in terms of the binary fragmentation of the compound system formed.

Zn clock models and chains of so(n)2 non-Abelian anyons: symmetries, integrable points and low energy properties

We study two families of quantum models which have been used previously to investigate the effect of topological symmetries in one-dimensional correlated matter. Various striking similarities are observed between certain quantum clock models, spin chains generalizing the Ising model, and chains of non-Abelian anyons constructed from the so(n)2 fusion category for odd n, both subject to periodic boundary conditions. In spite of the differences between these two types of quantum chains, e.g. their Hilbert spaces being spanned by tensor products of local spin states or fusion paths of anyons, the symmetries of the lattice models are shown to be closely related. Furthermore, under a suitable mapping between the parameters describing the interaction between spins and anyons the respective Hamiltonians share part of their energy spectrum (although their degeneracies may differ). This spin-anyon correspondence can be extended by fine-tuning of the coupling constants leading to exactly solvable models. We show that the algebraic structures underlying the integrability of the clock models and the anyon chain are the same. For n = 3,5,7 we perform an extensive finite size study—both numerical and based on the exact solution—of these models to map out their ground state phase diagram and to identify the effective field theories describing their low energy behaviour. We observe that the continuum limit at the integrable points can be described by rational conformal field theories with extended symmetry algebras which can be related to the discrete ones of the lattice models.

Defect tolerance of friction stir welds in DH36 steel

Abstract Friction stir welding of steel is in the early stages of development. The aim to commercialise this process creates a trade-off between welding time, cost and quality of the joint produced. Therefore, it becomes critical to analyse the lower quality bound of steel friction stir welds in conventional square edge butt welding configuration. Work has been undertaken to evaluate the microstructure and fatigue performance of 6 mm thick DH36 steel plates friction stir welded with sub-optimal process conditions, resulting in the development of embedded and surface breaking flaws. The defective weldments were characterised to understand the nature of the flaws and a programme of mechanical testing was undertaken (including fatigue assessment) to determine the relationship between the flaw geometry, location and weld quality. A number of characteristic flaws were identified and seen to interact with the samples' fatigue fracture mechanisms. Samples with wormholes at the weld root produced the lowest fatigue performance. Fracture from incomplete fusion paths at the retreating side of the welds' top surface was seen to correspond to the highest recorded fatigue lives. The work provides an insight into the complex nature of characteristic flaws in steel friction stir welds and their interaction with fatigue behaviour.

PH Alters PEG-Mediated Fusion of Phosphatidylethanolamine-Containing Vesicles

Here, we examine the different mechanisms of poly(ethylene glycol)-mediated fusion of small unilamellar vesicles composed of dioleoylphosphatidylcholine/dioleoylphosphatidylethanolamine (DOPE)/sphingomyelin/cholesterol in a molar ratio of 35:30:15:20 at pH 7.4 versus pH 5. In doing so, we test the hypothesis that fusion of this lipid mixture should be influenced by differences in hydration of DOPE at these two pH values. An examination of the literature reveals that DOPE should be less hydrated at pH 5 (where influenza virus particles fuse with endosome membranes) than at pH 7.4 (where synaptic vesicles or HIV virus particles fuse with plasma membrane). Ensemble kinetic experiments revealed substantial differences in fusion of this plasma membrane mimetic system at these two pH values. The most dramatic difference was the observation of two intermediates at pH 5 but loss of one of these fusion intermediates at pH 7.4. Analysis of data collected at several temperatures also revealed that formation of the initial fusion intermediate (stalk) was favored at pH 7.4 due to increased activation entropy. Our observations support the hypothesis that the different negative intrinsic curvature of DOPE can account for different fusion paths and activation thermodynamics in steps of the fusion process at these two pH values. Finally, the effects of 2 mol % hexadecane on fusion at both pH values seemed to have similar origins for step 1 (promotion of acyl chain or hydrocarbon excursion into interbilayer space) and step 3 (reduction of interstice energy leading to expansion to a critical stalk radius). Different hexadecane effects on activation thermodynamics at these two pH values can also be related to altered DOPE hydration. The results support our kinetic model for fusion and offer insight into the critical role of phosphatidylethanolamine in fusion.

Synapses between Interneurons in the Rat Cerebral Cortex at 2NM Resolution and in Three-Dimensions

We used thin sectioning and freeze-fracture electron microscopy to study the synapses in the rat cerebral cortex at 2nm resolution and in three-dimensions. The aim was to determine whether the opening of vesicles in the pre-synaptic axon was correlated to the distribution of the neurotransmitter receptors in the post-synaptic spine predicted by the long-term potentiation (LTP) of synaptic strength hypothesis. In the axon, “hybrid” filaments assembled from rods protruding from the vesicle and the plasma membrane docked synaptic vesicles to the active zone. The orientation of the “hybrid” filaments with respect to the plasma membrane defined two vesicle-docking paths. When parallel to the plasma membrane, the vesicles fuse “randomly” on the active zone, a pathway that was predicted by the SNARE hypothesis. When perpendicular to the plasma membrane, the vesicles dock at the base of plasma membrane depressions (35-45 nm), which are spaced 75-85 nm center-to-center, arranged in files and contained within gutters (the “deterministic” path). This path was not anticipated by the SNARE hypothesis. Computer simulations indicate that the “deterministic” path creates neurotransmitter concentration domains on the post-synaptic spine that are wider and shallower than those of the “random” path. In both paths, the opening of the fusion pore follows steps anticipated by the fusion-through-hemifusion hypothesis. In the post-synaptic spine, we identified intra-membranous particles in the external (E) face with dimensions that match the homotetrameric GluR2 AMPA-sensitive receptor. The particles are either clustered on the E face or distributed in bands on the “non-synaptic” plasma membrane. We thus conclude that in the pre-synaptic axon vesicle exocytosis follows “random” and “deterministic” fusion paths while in the post-synaptic spine LTP of synaptic strength involves the rearrangement of pre-existing receptors instead of de novo insertion and retrieval.

Electric Current-Induced Liquation Features of Cast Pure Tin

Electric current-induced liquation feature of cast pure ¢-tin under alternating current (AC) was investigated in this study. According to the experimental results, the initial fusion emerges from grain interior. Since the fusion paths traverse the grain boundaries preferentially in a highangle manner, their track inside a grain tends to be curved and thus longer than the grain size. Continued electrification leads fusion paths to form network-like interconnected fusion structure with most triple junctions residing within grain interior. Lateral thickening of the interconnected fusion paths leads to final failure. The result that the network-like fusion structure is rather equiaxed can eliminate the effect of crystal anisotropy on the fusion behavior if it exists. [doi:10.2320/matertrans.M2012162]

DYNAMICAL DIPOLE MODE IN FUSION HEAVY-ION REACTIONS

The prompt γ-ray emission, associated with the dynamical dipole mode decay, was investigated in the 32,36 S + 100,96 Mo and 36,40 Ar + 96,92 Zr fusion-evaporation reactions in the energy range E lab= 6 - 16 MeV / nucleon . The above reaction pairs populate, through entrance channels having different charge asymmetries, a compound nucleus in the 132 Ce mass region at excitation energies of 117, 174 and 284 MeV with identical spin distribution. By studying the differential γ-ray multiplicity spectra of the considered systems, the features of the dynamical dipole mode as a function of the beam energy were extracted while the γ-ray angular distributions were used to prove its pre-equilibrium character. The experimental findings were compared with theoretical predictions performed within a BNV transport model and based on a collective bremsstrahlung analysis of the entrance channel reaction dynamics. As a fast cooling mechanism on the fusion path, the prompt dipole radiation could be of interest for the synthesis of super heavy elements through hot fusion reactions providing a way to cool down the hot fusion paths, so ending up with a larger survival probability. To shed light on this hypothesis, the dynamical dipole mode investigation was extended to a heavier compound nucleus, 192 Pb , by means of the 40 Ca + 152 Sm and 48 Ca + 144 Sm reactions at E lab= 11 and 10.1 MeV/nucleon, respectively. Preliminary results of this measurement, concerning both fusion-evaporation and fission events are presented.

Entrance and Exit Channels for the Heaviest Elements

The potential barriers governing the entrance and α decay channels of the heaviest elements have been determined within a Generalized Liquid Drop Model including the asymmetry, the proximity effects, an adjustment to reproduce the Q value and within the asymmetric two-center shell model and the Strutinsky method. In cold fusion reactions double-hump barriers stand and incomplete fusion may occur. Warm fusion paths lead to one-hump potential barriers but also to very excited systems cooling down by neutron or even α particle evaporation. Partial half-lives for α decay have been calculated and compared with the new available experimental data.

On the formation and alpha decay of superheavy elements

The potential energy governing the entrance and α-decay channels for superheavy elements has been determined within a Generalized Liquid Drop Model including the proximity effects, the asymmetry, an accurate nuclear radius, an adjustment to reproduce the experimental Q value and within the asymmetric two-center shell model and the Strutinsky method. In cold fusion paths double-hump fusion barriers stand and incomplete fusion events may occur. Warm fusion reactions lead to one hump potential barriers and to very excited systems cooling down by neutron or even α-particle evaporation. α-decay half-lives of superheavy elements have been calculated and compared with the new available experimental data.

Comparison between the symmetric fission and fusion paths

The symmetric fission path leading to smooth extended shapes and the fusion or new fission path going through the two tangent sphere configuration are investigated within the liquid-drop model including the nuclear proximity energy. Analytical formulae are given for the various shape-dependent functions which govern the dynamics. The quadrupole moment, the perpendicular moment of inertia and the Coulomb energy are similar in the two paths. In contrast, the neck radius, the rupture point between the fragments, the parallel and effective moments of inertia, the surface energy and the critical angular momentum against fission are quite different in the two valleys. The introduction of the proximity energy strongly lowers the deformation energy in the fusion valley and for the light, medium and very heavy nuclei the barrier heights are nearly equal in the two paths. This flattening of the potential surface by the proximity forces allows to better understand the sudden transition between the one- and two-body configurations.