Applied Soft Computing最新文献

The traditional manual concrete vibration work faces numerous limitations, necessitating efficient automated method to assist in this task. This study proposes a vision-based continuous concrete vibration method for vibrating robots. By enhancing the YOLOv8n model with attention mechanisms, our proposed method demonstrates a high AP of 93.31 % in identifying reinforcing grids, and an FPS of 25.6 on embedded systems. For the first time in concrete vibration tasks, this study utilizes spatial positional information to cluster coordinate data, transforming confidence-sorted data into spatially ordered sequences. Vibrating robot case test shows that the proposed method enhances the vibration speed by 22.18 % and improves the vibration success rate by 11.67 % compared to traditional strategies. Additionally, the on-site experiment conducted at four construction sites demonstrated the robustness of the proposed method. These findings advance automation in concrete vibration work, offering significant implications for the fields of robotics and construction engineering.

Clean air, as a symbol of high-quality air quality, is the most basic requirement for people to maintain health. Moreover, in keeping humans fit, accurate short-term air quality prediction is vital. The decomposition algorithm can better capture the local features and temporal changes of the data. However, it increases the computation time, resource consumption, and complexity of the model. On the other hand, existing forecasting systems overlook instability and uncertainty. To solve the above problems, a deterministic and uncertainty AOA-DBGRU-MDN deep learning systems is proposed, which combines arithmetic optimization algorithm (AOA), double-layer bi-directional GRUs (DBGRU), and mixture density network (MDN). The above systems consider meteorological factors and air pollutants comprehensively. It involves feature selection using maximum information coefficient (MIC), decomposition using complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm, classification, and compression of decomposed components using entropy-Huffman tree compression. Firstly, the information measurement process reduces the number of components significantly. Following the incorporation of multi-factor data, the optimal DBGRU model is then obtained using AOA. Finally, the training errors are fitted using MDN to obtain interval prediction results. The experiments demonstrate that (1) Using the CEEMDAN algorithm can improve the prediction accuracy; (2) Classifying and reconstructing the data based on entropy-Huffman tree compression can not only decrease the model's training volume and improve training efficiency but also boost the model's prediction accuracy; (3) The AOA-DBGRU-MDN system performs probabilistic prediction to obtain an effective and intuitive prediction interval to improve the point prediction of air quality prediction.

Hand gestures are crucial for developing prosthetic and rehabilitation devices, enabling intuitive human–computer interaction (HCI) and improving accessibility for individuals with impairments. Recently, gesture recognition systems based on surface electromyography (sEMG) have been widely employed in various fields, demonstrating remarkable advantages and developments. In this paper, we present a comprehensive survey on sEMG-based hand gesture recognition. We provide an overview of the basic knowledge and background of sEMG signals and the acquisition equipment used. We delve into the applied feature extraction methods and classification models, focusing on recent advances in deep learning techniques. We also identify the datasets of sEMG signals used for hand gesture recognition. Moreover, we highlight recent applications of sEMG-based gesture recognition methods, including HCI, sign language recognition, rehabilitation, prosthesis control, and exoskeletons for augmentation. Additionally, we outline the latest innovative progress in this field, such as the influence of force, user identity detection, and migration effects. We also discuss the current limitations and challenges. Finally, we summarize the main findings and discuss future directions to enhance sEMG-based hand gesture recognition.

Recently, transformer trackers mostly associate multiple reference images with the search area to adapt to the changing appearance of the target. However, they ignore the learned cross-relations between the target and surrounding, leading to difficulties in building coherent contextual models for specific target instances. This paper presents a Temporal Relation Transformer Tracker (TRTT) for robust visual tracking, providing a concise approach to modeling temporal relations by dual target memory learning. Specifically, a temporal relation transformer network generates paired memories based on static and dynamic templates, which are reinforced interactively. The memory contains implicit relation hints that capture the relations between the tracked object and its immediate surroundings. More importantly, to ensure consistency of target instance identities between frames, the relation hints from previous frames are transferred to the current frame for merging temporal contextual attention. Our method also incorporates mechanisms for reusing favorable cross-relations and instance-specific features, thereby overcoming background interference in complex spatio-temporal interactions through a sequential constraint. Furthermore, we design a memory token sparsification method that leverages the key points of the target to eliminate interferences and optimize attention calculations. Extensive experiments demonstrate that our method surpasses advanced trackers on 8 challenging benchmarks while maintaining real-time running speed.

Although face recognition technology has been applied in many scenarios, it still suffers from many types of presentation attacks, so face anti-spoofing (FAS) becomes a hot topic in computer vision. Recently, vision transformer is recognized as the mainstream architecture for FAS, which always relies on auxiliary information, sophisticated tricks and huge model parameters. Considering that face based identity authentication usually takes place on mobile-like devices, therefore how to design an effective and lightweight model is of great significance. Inspired by the powerful global modeling ability of self-attention and the model compression ability of knowledge distillation, a simple yet effective knowledge distillation approach is proposed for FAS under transformer framework. Our primary idea is to leverage the rich knowledge of a teacher network pre-trained on large-scale face data to guide the learning of a lightweight student network. The main contributions of our method are threefold: (1) Feature- and logits-level distillation are combined to transfer the rich knowledge of teacher to student. (2) A head-aware strategy is proposed to deal with the dimension mismatching issue of middle encoder layers between teacher and student networks, in which a novel attention head correlation matrix is introduced. (3) Our method can bridge the performance gap between teacher and student, and the resulting student network is extremely lightweight with only 5 MB parameters. Extensive experiments are conducted on three public face-spoofing datasets, CASIA-FASD, Replay-Attack and OULU-NPU, the results demonstrate that our method can obtain performance on par with or superior to most FAS methods and outperform many knowledge distillation methods. Meanwhile, the distilled student network achieves excellent performance with 17$\times$ fewer parameters and 9$\times$ faster inference time compared to the teacher network. The code will be publicly available at https://github.com/Maricle-zhangjun/HaTFAS.

Multiple exposure image fusion is a technique used to obtain high dynamic range images. Due to its low cost and high efficiency, it has received a lot of attention from researchers in recent years. Currently, most deep learning-based multiple exposure image fusion methods extract features from different exposure images using a single feature extraction method. Some methods simply rely on two different modules to directly extract features. However, this approach inevitably leads to the loss of some feature information during the feature extraction process, thus further affecting the performance of the model. To minimize the loss of feature information as much as possible, we propose an ultra-high-definition (UHD) multiple exposure image fusion method based on multi-scale feature extraction. The method adopts a U-shaped structure to construct the overall network model, which can fully exploit the feature information at different levels. Additionally, we construct a novel hybrid stacking paradigm to combine convolutional neural networks and Transformer modules. This combined module can extract both local texture features and global color features simultaneously. To more efficiently fuse and extract features, we also design a cross-layer feature fusion module, which can adaptively learn the correlation between features at different layers. Numerous quantitative and qualitative results demonstrate that our proposed method performs well in UHD multiple exposure image fusion.

Batch Processing Machines (BPM) and transportation are seldom studied simultaneously in Flexible Flow Shop. In this study, Flexible Flow Shop Scheduling Problem (FFSP) with BPM at the last stage and transportation is considered and an adaptive shuffled frog-leaping algorithm (ASFLA) is proposed to minimize makespan. To produce high-quality solutions, a heuristic is employed to produce initial solution, two groups are formed by using all memeplexes, then an adaptive memeplex search is implemented, in which the number of searches is dynamically determined by the quality of the memeplex, an adaptive group search is also conducted by exchanging memeplexes or supporting of the worse memeplex. A novel population shuffling and the worst memeplex elimination are proposed. A number of computational experiments are executed to test the new strategies and performances of ASFLA. Computational results demonstrate that new strategies are effective and ASFLA is a very competitive algorithm for FFSP with BPM and transportation.

Population growth in crowded cities and the resulting increase in vehicle use have led to the problem of insufficient parking. When public parking lots and urban growth are not in coordination, vehicles park on the street and close the crosswalks. In the coming years, this problem will become more complicated with the addition of autonomous vehicles (AVs) to urban traffic. This study addresses the research question of how to effectively select AV parking lots in urban areas experiencing population growth and increased vehicle usage. For this aim, a hybrid Multi-Criteria Decision Making (MCDM) methodology, combining SWARA (Step-wise Weight Assessment Ratio Analysis) and TOPSIS (Technique for Order Preference by Similarity) approaches in a Fermatean Fuzzy (FF) environment is proposed. The decision hierarchy based on the SCOR model has been developed to determine and construct the evaluation criteria. Then, a case study analysis has been applied to selected districts in Istanbul, which is Turkiye's most populous and developing city. Operating expenses, safety and security, and land costs are determined as the most important factors. As a result of the detailed fuzzy analysis, which districts should primarily be chosen for AV parking lots in Istanbul is determined and finally, the robustness and validity of the results obtained by the sensitivity analysis being questioned. The study contributes by providing insights into AV parking lot selection, demonstrating the efficacy of the proposed methodology, and highlighting the importance of addressing this issue in urban planning.

Sound event detection (SED) is a core technology in smart home projects that rely on detected sound events to trigger specific actions. SED systems face two major challenges: high labeling costs and complex acoustic environments. To reduce labeling costs, some semi-supervised systems extract both global and local features for classification. However, these methods treat global and local features equally, not accounting for their varying importance when recognizing different types of sound events. Furthermore, to address complex acoustic environments, some studies use multitask learning frameworks to introduce SED-related tasks as auxiliaries to improve detection performance. However, these methods fail to align tasks within the framework, leading to conflicting outputs that may limit system performance. To address these issues, in this paper we propose a “two-to-one” teacher-student learning based semi-supervised SED system. This system employs a gating mechanism to selectively enhance global and local features, improving adaptability to different types of sound events, and incorporates a cross-task alignment module to interact SED with related tasks, reducing the risk of performance degradation caused by conflicting outputs. Experimental results on two datasets demonstrate that our system achieves the best performance in all metrics, with EB-F1 scores of 48.1 % and 64.7 %, representing improvements of 15.3 % and 10.6 % over the baseline ConformerSED system, respectively. Our work offers an effective SED solution for smart home projects by providing a semi-supervised SED system that performs well while reducing labeling costs.

The reasonable ranking of binary pairs that characterize fuzzy information in many fuzzy decision problems is very important. To overcome some defects of the existing score functions for the q-rung orthopair fuzzy numbers (q-ROFNs), a novel score function and ranking criterion are proposed by the q-compression transformation and hesitation factor. The main motivation is to introduce the generalized Zhenyuan (GZ)-integral into the q-ROFN environment, and cleverly transform the aggregation operations into a linear programming problem through the arithmetic operations of q-ROFNs. The main contribution is to solve the aggregation problem of q-rung orthopair fuzzy generalized Zhenyuan integral ordered weighted average (q-ROFGZIOWA) operator through the optimization technique of linear programming, and a new decision making method is established by using the q-ROFGZIOWA operator and ranking criterion. The main innovation is to map all q-ROFNs to the unit triangle in the first quadrant (converted into intuitionistic fuzzy numbers, IFNs) according to the q-compression transformation in geometric significance, and the novel score function and its ranking criterion are proposed by combining hesitation factor, and then the aggregation operation based on generalized Z-integral is converted to an optimization problem in linear programming. Finally, the superiority of the proposed method are verified by comparing the aggregation results of two integral operators through an example, and apply the proposed method to the optimal decision-making of talent recruitment in universities. The proposed method can not only correct some flaws in the ranking of existing q-ROFNs, but also overcomes some defects of existing Choquet integral average (geometric) operators in a q-ROFN environment. These results are of great significance for further research on the widespread application of q-ROFNs.