The large amount of point cloud data, unbalanced distribution and uneven density of urban road scenes make it difficult for the current point cloud classification methods to meet the requirements of high-precision classification. To deal with the problem of insufficient local feature extraction by PointNet++ networks, a local feature aggregation module is designed based on the attention mechanism, which adequately captures local information by dynamically merging neighboring point features. Considering that the existing classification models cannot take into account contextual information, which leads to limited classification performance in complex scenes, a dual-attention module and a context-aware module are constructed to extract contextual information from several dimensions to further enhance the feature representation capability. The experimental results show that the new method has higher accuracy and stronger generalization performance (overall accuracy reaches 98.70% and 96.84% in Oakland and Paris publicly available datasets) under large point cloud datasets, and is more suitable for large-scale point cloud classification.

It is the current mode of air traffic management that flight crews fly according to the instructions from air traffic controllers. With the increase of the number of flights, in order to effectively improve the efficiency of air traffic operation and reduce the workload of controllers, the development of intelligent air traffic management based on track prediction has become a new topic. Aiming at the research of flight track prediction technology, a two-stage flight track prediction method is put forward innovatively in this paper, which includes classification and then prediction. Firstly, K-means is used to cluster and classify the flight track based on the data of a certain airport. Next, Adam-LSTM deep learning model is constructed for each type of approach track, and high quality track prediction is realized. The results show that, compared with the traditional prediction model, the track prediction effect is greatly improved. The research results can provide technical support for intelligent air traffic management and abnormal track recognition.

The grid-connection of distributed generations and the gradually expanding scale of distribution networks make the traditional fault location methods more difficult. A multi-strategy improved Hybrid Whale Optimization Algorithm Gray Wolf Optimization (H-WOA-GWO) combined with region correction fault location method is proposed to address this problem. Firstly, the WOA encircling contraction and spiral updating mechanisms are integrated into GWO to construct a hybrid algorithm to effectively improve the convergence speed; then the nonlinear convergence factor, improved leader wolf position and adaptive hunting weights are applied to enhance the search adaptability, global development capability and shorten the iteration time. Different location models are established to choose in constructing the objective function based on the evaluation function value method, and the region correction strategy is proposed by analyzing the potential information of the pseudo-optimal solution. After simulation verification, under triple faults: the correct rate of hybrid algorithm is 11 percentage points centa higher than that of single algorithm, and the iteration time can be saved by 0.326 7 s; the correct rate and solution time after combining with the region correction strategy are improved by 17 percentage points and 74.88%, respectively, compared with that of pure hybrid algorithm. It shows that the proposed algorithm and correction strategy can quickly and accurately recognize multiple and multi-deformed node faults with efficient solution speed and stability.

In order to reduce the randomness and volatility of wind speed and further improve the prediction accuracy, a combined prediction model combining with quadratic mode decomposition, which can improve dung beetle optimizer and bidirectional gated recurrent unit was proposed. Firstly, aiming at the problems such as local optimization and poor global search ability in Dung Beetle optimizer (DBO), the improved algorithm (LDBO) is improved by introducing Latin hypercube sampling, tangential flight and other strategies, and the improved algorithm (LDBO) is used for parameter optimization of BiGRU. Secondly, the quadratic modal decomposition is used to reduce the complexity of the original data and provide stable sequence data for subsequent modeling. Then, BiGRU is used to predict each modal component obtained after the quadratic mode decomposition, and the prediction results of each modal component are superimposed as the final prediction results. Finally, the proposed QMD-LDBO-BiGRU prediction model is compared with other four mainstream prediction models (CNN-LSTM, TCN-RVM, ELM-Adaboost, BiTCN-SVM). Experimental results show that the evaluation index R² of the QMD-LDBO-BiGRU model reaches 98.086%, which is increased by 21.396, 19.525, 11.474, 5.457 percentage points compared with the comparison model, respectively, which verifies the effectiveness and applicability of the proposed model and provides a certain reference for further improving the accuracy of wind speed prediction.

A method for simulating flood dynamics using Physics-Informed Neural Networks with Attention Mechanism (PINNsFormer) is proposed to address the shortcomings of Physics-Informed Neural Networks (PINNs) in handling time-dependent problems. The PINNsFormer model is applied to solve the Saint-Venant equations. The model effectively captures spatiotemporal dependencies, thus improving accuracy and generalization. Experimental results show that this method performs excellently in simulating flood propagation and capturing water surface profile details. Compared with PINNs and neural network models FLS and QRes, which handle time features, PINNsFormer demonstrates higher stability and accuracy. Numerical experiments on a horizontal plane and a uniform adverse slope indicate that the PINNsFormer method achieves the lowest loss values and test errors, reaching an accuracy of 10^-4 magnitude, accurately reproducing the shape of flood inundation boundaries.

Due to the characteristics of small inter-class differences and large intra-class distinctions in fine-grained image, many studies have utilized Vision Transformer to mine critical region features to improve the accuracy of fine-grained image classification. However, there still exists two major problems: firstly, background regions are also considered when the network mines critical classification cues, bringing additional noise interference to the model. Secondly, there is a lack of spatial connection between the local embeddings feature of the input images, and the model lacks the ability of object structure cognition, leading to inaccurate extracted category features. To address these problem, this paper proposes two modules: adaptive spatial mutual attention module and feature pair integrated discrimination module, which first learns the mutual attention weights of different embedding layers to select better discriminative regions through mutual attention spatial adaptive module, and adaptively learns the neighbor relationship of different regions through graph convolutional network. Then the feature pair integration discrimination module is utilized to treat the cue interactions between image pairs and reduce the confusion between fine-grained images. The final prediction results are derived under the token feature enhancement strategy. The proposed method achieves an accuracy of 92.5%, 93.3% and 91.8% on three benchmark datasets, namely, CUB-200-2011, Stanford Dogs and NABirds, which are better than many other existing state-of-the-art methods.

Aiming at the problems of poor accuracy of traditional apple leaf disease classification methods, an apple leaf classification algorithm CALDNet based on improved ConvNeXt is proposed. 3223 Network is designed to adjust the structure of the model, while jump connection and position coding are introduced to enhance the model’s ability to capture the space and to improve the stability of the training process, and Spatial Pyramid Pooling (SPP) is used to capture spatial features on different scales and enhance the model’s ability to adapt to large and small lesions; on the basis of ConvNeXtblock, G-ConvNeXtblock is designed to improve the depth convolution, and a Gabor filter is introduced as a convolution kernel to better capture texture information in the image. In order to improve the model’s ability to recognize a small range of apple leaf disease recognition ability, an enhanced channel and attention mechanism (enhanced CBAM) is designed. In the experiments, seven common leaf diseases (black-star disease, black-rot, brown-spot, mosaic disease, healthy, rust, gray-spot) are chosed as the main research subjects, and the experimental results by using the improved algorithm and other mainstream algorithms are compared. The experimental results show that the CALDNet model recognizes the leaf disease model with the precision rate, recall rate, and F₁ value of 97.58%, 97.54%, and 97.54%, respectively, compared with the original ConvNeXt model, which increased by 4.63,4.56 and 4.60 percentage points, solving the problems of poor precision of traditional apple leaf disease classification.

To address the issues of poor inter-modal interaction, insufficient consideration of temporal order, and varying importance of modalities in multimodal sentiment analysis, a temporal multimodal sentiment analysis framework based on a cross-modal augmentation network (TCAN-SA) is proposed in this paper. Firstly, the inter-modal interaction module enhances the information exchange between modalities. Secondly, a bidirectional temporal convolutional network (BiTCN) layer is introduced to capture the temporal characteristics of the modal information. Finally, a multimodal gating module is employed to balance the varying importance among modalities. Experimental results demonstrate that the framework performs well on two public datasets, CMU-MOSI and CMU-MOSEI, and outperforms other existing models.

With the in-depth application of blockchain technology in industries such as finance and healthcare, the challenges it faces are becoming increasingly prominent. Among them, the optimization and development of consensus mechanisms are the most crucial. Currently, the two consensus protocols are mainly adopted by consortium chains, PBFT and Raft, and both have certain limitations. The communication volume of the PBFT consensus mechanism increases exponentially with the increase in the number of nodes in the blockchain network, leading to a decrease in efficiency. Although the Raft consensus mechanism has been optimized in terms of efficiency, its ability to resist Byzantine attacks is weak. To address these issues, a consensus mechanism MRBFT based on Raft that resists Byzantine attacks is proposed. Firstly, a reputation value mechanism is introduced in the Raft election process. By electing nodes with higher reputation values, the reliability of the elected nodes is enhanced. At the same time, a certain proportion of Monitor nodes are elected along with the Leader. Secondly, during the consensus process, the Monitor nodes supervise the behavior of the other nodes to enhance the algorithm’s ability to resist Byzantine attacks. At the same time, the reputation values of the nodes are updated to ensure that the most trustworthy nodes are elected in each round, improving the security of the algorithm. Experimental results show that, under the same conditions of resisting Byzantine attacks as PBFT, as the number of nodes increases, MRBFT has lower communication volume, with communication overhead being 44% of PBFT and throughput being 1.5 times that of PBFT. Compared with similar algorithms, in scenarios with similar security, it has more obvious optimization effects in terms of throughput and consensus delay.

In order to study the bifurcation and dynamical behavior of traveling wave solutions of the time-space fractional Sasa-Satsuma equation, the fractional-order complex transformation of the time-space fractional Sasa-Satsuma equation is performed to transform them into an equivalent ordinary differential system,the corresponding plane dynamic system is derived,and the corresponding phase diagram is obtained by discussing the different values of the parameters of the plane dynamic system. According to the bifurcation of the system, the exact expressions of various traveling wave solutions for the time-space fractional Sasa-Satsuma equations with different trajectories are obtained.

In the step stress partially accelerated life test (SSPALT) with Type-I progressive hybrid censoring (Type-I PHC) scheme, the statistical inference of a dependent two component system with masked data are studied. When the component life follows exponential distribution, the tampered random variable (TRV) model and copula function are used to construct the dependent competing risks model. Based on the maximum likelihood theory, the maximum likelihood of unknown parameters is established, and the corresponding confidence interval estimation of the parameters is provided through asymptotic likelihood theory and bootstrap method. Finally, the feasibility of the proposed estimation method is verified by Monte-Carlo method. The results show that the method is feasible and effective for reliability analysis of dependent masked data systems.

In view of the possible collision between the leader agent and the follower agents, combined with the characteristics of practical consensus and identical consensus, the set consensus of the leader-following multi-agent system is proposed and studied in this paper. Firstly, a bounded set without zero is selected and the concept of set consensus is given. Then, by designing an appropriate segmented pinning control protocol and establishing an error system, the set consensus problem of the multi-agent system is transformed into a set stability problem of the error system, and a sufficient condition for the multi-agent system to achieve the set consensus under the weighted directed network topology is obtained. Finally, the numerical simulation verifies the correctness of the theoretical results.

To investigate the effect of phenolic resin addition on the microstructure and electrochemical properties of carbon materials, and to obtain a relatively economical method of preparing carbon materials for supercapacitors, phenolic resin/bamboo based composite activated carbon was prepared by KOH activation using bamboo waste as raw material and phenolic resin as modifier. The results showed that the microstructure of the composite activated carbon was affected by the addition of phenolic resin. The specific surface area and total pore volume of the composite activated carbon with the resin addition of 15%~35% increased. The average pore diameter decreased, providing more adsorption sites for storage charge. The specific surface area and total pore volume of the composite activated carbon were increased by 28.5% and 29.2%, respectively, and the average pore size was reduced to 1.78 nm compared with that of the bamboo-based activated carbon when 35% of phenolic resin was added; the specific capacitance at a current density of 0.5 A·g^-1 reached 263.85 F·g^-1, which was 53.3% higher than bamboo activated carbon. The results indicated that the electrochemical properties of activated carbon prepared by adding phenolic resin were obviously improved.

A fluorescence intensity and polarization method for simultaneous detection of pesticide acetamiprid was established based on the fluorescence quenching effect and mass amplification effect of graphene oxide (GO). When nucleic acid signal probe which labeled with fluorescent molecule FAM attached to the GO surface, a lower fluorescence intensity and a higher polarization signal was detected. After acetamiprid binds to its aptamer, the released complementary probe was hybridized with the signal probe to move it away from graphene, and then the enhanced fluorescence intensity and the reduced polarization signal are obtained. The relationship between the fluorescence intensity or polarization signal and different concentrations of acetamiprid was examined, and the conditions such as the target recognition time, GO concentration and reaction time were optimized. Under the optimized experimental conditions, the fluorescence polarization detection method (LOD=5 nmol/L) had a lower detection limit than the fluorescence intensity method (LOD=50 nmol/L), while the fluorescence intensity method was more stable. This method showed good selectivity and achieved spiked recovery analysis of acetamiprid in celery leaves and lake water.

Suspended particulate matter (SPM) is an important component of natural water body and can significantly influence the photolytic behavior of water pollutants. A comprehensive understanding about the photochemical behavior of water pollutants in natural waters requires consideration of the presence of SPM. In this study, montmorillonite-humic acids (MMT-HAs) composite particle was prepared to simulate SPM in natural waters, and their effects on the photolysis of tetracycline (TC) were investigated. The results demonstrated that the presence of MMT-HAs composite particle in water significantly enhanced the photolysis of TC, with the photolytic kinetics following a pseudo-first-order model. Electron spin resonance spectra and free radical quenching experiments indicated that the photoactive components (MMT and humic acids) in the composite particle induced the generation of reactive oxygen species under light exposure, contributing to the enhanced photolysis of TC. Comparative analysis of the photolysis of TC in the MMT and MMT-HAs particle systems and adsorption experiments further revealed that the promoted photolysis of TC was also related to the interfacial interaction between MMT-HAs composite particles and TC molecules. The formation of surface complex between amino groups of TC and the negatively charged sites on MMT surface facilitated light absorption or electron transfer, thereby accelerating the photolysis of TC. The promotion of TC photolysis by MMT-HAs composite particles was the result of the combined action of these two mechanisms. Photoproduct analysis indicated that the hydroxyl radical addition to the aromatic ring of TC, as well as demethylation, deamination and dehydration in the side chains were the main degradation pathways of TC in the composite particle systems. The findings can provide valuable insights into the photolytic behavior of water pollutants in natural waters.

As important biomass platform molecules, isopropyl levulinate and γ-valerolactone have promising applications in the production of chemicals, liquid fuels and polymerization. The acidity-tunable P-Zr/H-Beta bifunctional catalyst for the selective conversion of furfural to γ-valerolactone and isopropyl levulinate were prepared. The catalyst could effectively catalytic production of γ-valerolactone and isopropyl levulinate from furfural via transfer hydrogenation, ring opening and cyclization reactions with isopropanol as the hydrogen donor. The yields of γ-valerolactone and isopropyl levulinate were 41.6% and 23.7%, respectively, with a total yield of 65.3% at 150 ℃. The reaction activity of transfer hydrogenation mainly depended on the acid-base sites of catalyst, which was determined by the molar ratio of P to Zr in P-Zr/H-Beta. In addition, the effects of reaction parameters such as time and temperature on the yield of γ-valerolactone were explored; and the structure of P-Zr/H-Beta was further characterized in detail.

Exploring the characteristics and spatial distribution pattern of the existing ancient tree resources is of great significance for the conservation of ancient trees and the maintenance of global biodiversity. By combining literature and field survey data,this study firstly analyzed the species composition, distribution area types, structural characteristics, growth status characteristics, and spatial distribution pattern, then explored the composition and currents status of spatial pattern of ancient tree resources in the Lijiang River Basin. The results showed that there are a total of 4 360 ancient trees distributed in the Lijiang River Basin, belonging to 124 species in 82 genera of 40 families, of which 3 species are national first-level key protected plant wild and 8 species are national second-level key protected plant Wild. The composition of ancient trees is mainly composed of tertiary ancient trees, accounting for 84.33% of the total number of trees. The overall growth of the ancient trees was in good condition, and the normal growth of the ancient trees accounted for 94.89% of the total number of trees, and only 3 dead ancient trees were found. The distribution types of families and genera are mainly the Pantropic and the Tropical Asia (India-Malaysia), which is consistent with the geographical location of the Li River Basin. Correlation analysis showed that there was a significant positive correlation between the three variables of tree height, breast circumference and crown width of ancient trees (P<0.01). There are significant differences in the number and diversity index of ancient trees among counties in the Lijiang River Basin. In general, the distribution of ancient trees in the central part of the Lijiang River Basin is less, while the distribution of ancient trees around the basin is more.

In order to understand the effects of different stand retention densities on the soil characteristics of Cunninghamia lanceolate plantations, the soil water content, stoichiometric ratio, enzyme activity and understory vegetation diversity of Chinese fir plantations under different stand retention densities were studied in the experimental forests with different stand retention densities in Yangkou State Owned Forest Farm in Fujian Province. The results showed that: 1) There was no significant effect on the species richness index and Shannon-Wiener index in the understory vegetation diversity of Chinese fir plantations under different retention densities (P>0.05), but it had a great impact on Simpson index and Pielou evenness index. 2) There were significant differences in soil chemical properties among different stand retention densities (P<0.05), as stand retention density increased, total carbon content of soil increased first and then decreased. Soil total nitrogen decreased first and then increased. 3) The retention density of different stands had a significant effect on soil enzyme activity (P<0.05), among which,with the increase of stand retention density, the activities of polyphenol oxidase and acid phosphatase decreased in soil, while the activities of soil urease and soil sucrase increased. 4) Different stand retention densities had significant effects (P<0.05) on soil microbial dominant bacterial communities, and low stand density treatment could improve the structural diversity of soil microbial bacterial communities to some extent. In the correlation analysis between soil bacterial communities and physicochemical properties, TP, effective phosphorus content (AP) and soil dominant bacterial group (Actinobacteria) were highly significantly correlated (P<0.01). TC and TN were highly significantly correlated (P<0.01) with the dominant bacterial group (Zoopagomycota). Overall, low stand retention density is the best choice for planting Chinese fir plantations.

In this study, the Heat Shock Protein 70 (HSP70) gene of cedar was investigated by cloning, bioinformatics and expression analysis methods to provide a scientific basis for the in-depth understanding of the function of the HSP70 gene of cedar, as well as the genetic improvement and sustainable cultivation of cedar. The annual seedlings of the excellent Chinese fir asexual line “Yang 061” were selected as test materials, and the ClHSP70 gene was cloned by reverse transcription polymerase chain reaction (RT-PCR). The physicochemical properties, transmembrane helical domain, signal peptide, secondary structure and tertiary structure of ClHSP70 protein were predicted and analysed using Expasy software and other online software; the subcellular localization of the protein was predicted using Cell-PLoc 2.0 online software; Mega 11 software was used to construct a phylogenetic tree; the ClHSP70 gene was cloned and constructed into pCAMBIA35s-EGFP vector for analysis of ClHSP70 gene. EGFP vector to analyse the subcellular localisation of the ClHSP70 protein; and its expression level was analysed using Quantitative Real-time PCR (QRP). The cloned ClHSP70 gene encoded 670 amino acids, and the molecular formula of ClHSP70 protein was C₃₂₉₇H₅₃₀₆N₉₀₄O₁₀₀₈S₂₆, which was unstable, did not contain signal peptide and transmembrane region, and was predicted to be localised in the cytoplasm. Phylogenetic analysis showed that Chinese fir ClHSP70 was more closely related to hazelnut (Corylus avellana). The results of subcellular localisation experiments showed that ClHSP70 protein was localised in the nucleus. qRT-PCR expression analysis showed that the relative expression of ClHSP70 gene was the highest in the leaves, and the relative expression of ClHSP70 gene reached the peak after 6 h of high temperature, and the expression of ClHSP70 gene reached the maximum after 12 h of drought treatment, and the expression of ClHSP70 gene was up-regulated by the high temperature and drought stress induced. The ClHSP70 gene was up-regulated by high temperature and drought stress. The successful cloning and sequence analysis of the ClHSP70 gene in cedar revealed its expression in different tissues of cedar and its response to high temperature and drought, and provided an important theoretical basis for the breeding of cedar resistance.

Liujiang River is the second largest tributary of the Xijiang mainstream of the Pearl River water system and plays an important role in human production and life. The Shannon Wiener Diversity Index (H′), Biological Pollution Index (BPI), Sensitivity Index (BMWP), Biological Index (BI), and Benthic-Index of Biotic Integrity (B-IBI) were used to evaluate the health status of the Liujiang River. The results showed that when using the Macroinvertebrate Integrity Index (B-IBI), Biological Index (BI), and Biological Pollution Index (BPI) for evaluation, 97%, 93% and 97% of sampling points showed that the river was in a healthy or sub healthy state. However, in the Shannon Wiener diversity index and sensitivity index BMWP evaluation, only 59% and 34% of the sampling points showed that the river was in a healthy and sub healthy state. Based on the results of the five river health status evaluation indices, the sampling points H12, H13, H14, H20, H22, H23, H24, and H26 near Liubei District of Rong’an County and Liuzhou City showed poor river health status.