| Title | Keywords | ||
|---|---|---|---|
| Author | Authorship | ||
| Corresponding Author | Funds | ||
| DOI | Column | ||
| Summary | |||
| Timeframe | - | ||
Found 760 relevant information
Article
Lightweight multi⁃modal brain tumor MRI image segmentation based on Mamba
Hou Xiangning12; Huang Xiaobin12; Xu Caocao12; Yao Jun12The existing 3D U⁃Net model has high computational complexity and poor accuracy when processing 3D medical image data, so it is difficult to meet the needs of clinical real⁃time diagnosis. Therefore, a lightweight technology architecture based on 3D U⁃Net combined with Mamba is proposed to reduce the computational complexity and improve the segmentation accuracy of the traditional deep learning models in multi⁃modal brain tumor MRI image segmentation. Mamba′s linear complexity self ⁃ attention mechanism is used to optimize the 3D U⁃Net model to reduce the computational complexity and improve the inference speed and segmentation accuracy. The specific method is to introduce LWM module and attention bridge module on the basis of 3D U⁃Net model. Parallel Mamba architecture is adopted in the LWM module, which reduces the number of model parameters by reducing the number of input channels, greatly reduces the computational complexity and memory consumption while extracting global information, and realizes efficient segmentation in the case of fewer parameters. In addition, the LWM further improves the segmentation performance of fine⁃grained tumor cores by introducing SE module. The experiments on the dataset BraTS2020 show that the proposed method significantly improves the accuracy of tumor segmentation, and greatly reduces the training and inference duration, so it shows great practical application potential.
UAV swarm task allocation based on coalition game and simulated annealing
Bi Yijun; Wang Xie; Jiang TianIn view of the high computational complexity of task allocation in heterogeneous unmanned aerial vehicle (UAV) swarms and the lack of hard task priority constraints, this study proposes a coalition⁃game⁃based two⁃stage allocation algorithm. The hard priority constraints are introduced while establishing a mathematical model of UAV capabilities and task requirements, and a two⁃stage greedy pre⁃assignment+improved simulated annealing (PC_ISA) algorithm is designed with the framework of coalition game. Stage1 employs a local greedy dynamic algorithm to shrink the infeasible domain, while Stage2 applies the simulated annealing (SA) algorithm with adaptive temperature decay to minimize the coalition cost function. Simulations show that, with 12 UAVs being allocated for three tasks, PC_ISA achieves the best overall performance; after introducing weather disturbance factors, it exhibits the highest stability; and when the numbers of UAVs and tasks are further increased, it delivers the gentlest cost growth. By integrating the ISA algorithm with coalition game and embedding hard task priority constraints, the proposed method provides an effective solution to the task allocation in heterogeneous UAV swarms.
Design of multi⁃channel data acquisition system based on FPGA+ARM
Liu Jiajia12; Qi Xue12; Fan Lei12; Tan Qiulin12In view of the excessive size and insufficient collection accuracy of the traditional data acquisition systems, this paper designs a multi⁃channel data acquisition system based on the FPGA+ARM architecture. The system is tailored to meet the precise measurement requirements for parameters such as pressure and temperature in the field of aerospace. It comprises a power supply module, a channel switching module, a data acquisition module, and a data transmission module. Among them, the data acquisition module utilizes the SOM⁃TLT3F as its core controller, and integrates a signal conditioning circuit, a 24⁃bit analog⁃to⁃digital conversion chip AD4630, and eight multiplexing chips ADG1607 to enable synchronous acquisition of data of 64 channels. The FPGA (field⁃programmable gate array) achieves SPI communication with the ARM by interrupt control, and the processed data is transmitted to the upper computer via the UDP protocol. Experiments demonstrate that the system combines the characteristics of high collection accuracy with miniaturization, enabling stable operation in harsh environments such as high temperature and high pressure. The multi⁃channel switching design ensures efficient acquisition of multiple data streams, with a sampling relative accuracy not more than 0.23%. Featuring miniaturization and high collection accuracy, the proposed system provides a viable solution for precision testing in the field of aerospace.
Cooperative game based aggregation algorithm for federated learning
Liu Ying12; Li Yong13; Wen Ming2; He Zhenzhen1A cooperative game based federated learning aggregation algorithm is proposed to address the model training challenges due to client data heterogeneity and potentially malicious clients in federated learning. The method integrates personalized federated meta ⁃ learning with cooperative game Shapley value optimization strategy, which aims to improve the performance and robustness of the global model and reduce the communication and computation costs. Firstly, by collecting client soft labels, the maximum entropy judgment method is used to select the clients with high contribution to the global model. Secondly, a fast estimation strategy based on the TMC⁃Shapley value is designed to efficiently estimate the marginal contribution of clients by finite times sampling, so as to avoid exponential computational complexity. Finally, weighted aggregation is performed based on client Shapley values and data distribution characteristics. Experiments show that the proposed method performs well on the real dataset classification task, significantly improves the accuracy and reduces the computational cost in comparison with the baseline method. Its advantages are more prominent in the scenario with a large number of clients and significant data heterogeneity.
Research on distracted driving detection method based on lightweight networks
Zhong Yalu12; Kong Yanqi12; Lin Chen12; Zhang Hong12; Yang Fugui3; Xie Zhi12A lightweight detection network named HDSL⁃YOLO is designed to avoid the hardware resource constraints of in⁃vehicle systems and improve the deployability and accuracy rate of distracted driving detection models. On the basis of the YOLOv8n, the model is improved by incorporating multiple optimization strategies, including the following four key aspects: firstly, the HGNetV2 lightweight backbone network is introduced to effectively reduce parameter size while significantly improving detection speed and operational efficiency; secondly, the dynamic upsampling module (Dysample) is integrated to enhance feature representation, particularly in multi⁃scale object extraction; additionally, the SimAM (simple attention module) is incorporated to further strengthen the model′s perception and recognition capabilities for small objects; and finally, a lightweight detection head is adopted to further streamline parameters. Experiments demonstrate that the mAP@0.5 and mAP@0.5:0.95 of the HDSL⁃YOLO algorithm is improved by 92.4% and 55.4%, respectively. In comparison with the original YOLOv8n algorithm, the improved algorithm realizes not only an improved detection accuracy rate, but also a further lightweighting, so it achieves dual optimization. Deploying the HDSL⁃YOLO on the embedded platform Jetson Nano confirms shorter response times, which validates the effectiveness of the proposed improvements.
AUKF⁃based robot visual servo model predictive control
Jia Yang; Zhang Jianye; Wu ZizhaoIn view of the poor image Jacobian matrix estimation accuracy, robot system constraints, and camera field⁃of⁃view constraints in robot visual servo, this paper proposes a robot visual servo control method combining adaptive unscented Kalman filtering (AUKF) and interior point model predictive control (IP⁃MPC). Firstly, the traditional UKF algorithm is not accurate enough for noise estimation, so Sage ⁃ Husa filtering is introduced to estimate the process noise covariance matrix online to improve the online estimation accuracy of image Jacobian matrix. Secondly, for the constraints of robot system and camera field⁃ of⁃view, the model predictive controller is designed by the interior point method, and the constraint problem is transformed into the problem of minimizing quadratic programming to realize the tracking control of robot visual servo system. The experiments show that the convergence speed of the proposed method is improved by 38.1%, its average error of image feature points is reduced by 37.4%, and the end⁃effector speed fluctuation of the robot is decreased, which demonstrate that the proposed method has significant improvements in visual servo control accuracy, and system response speed and stability in comparison with the traditional UKF algorithm.
BO⁃XGBoost⁃based microwave recognition and early warning technology for bend traffic
Yan Chengfeng; Xiong Lun; Lu YongxiongMachine learning and microwave recognition technology can offer smarter solutions for bend warning systems. A low⁃cost 10.52 GHz microwave radar is used to detect moving targets on the road. By analyzing the radar echo characteristics of vehicles, two ⁃ wheelers, and pedestrians, 12 feature parameters are defined and extracted from the signal′s time ⁃ frequency diagram, frequency⁃amplitude plot, and instantaneous Doppler frequency⁃time curve, so as to construct feature vectors. A three⁃class dataset is created with pedestrians, two⁃wheelers, and vehicles as the target categories, and the SMOTE (synthetic minority oversampling technique) algorithm is applied to eliminate the class imbalance of the dataset. The XGBoost (extreme gradient boosting) algorithm model is examined, and after optimization using the Bayesian optimization algorithm (BOA), its macro⁃average accuracy rate for target recognition reaches 95.1%. Finally, an intelligent bend warning system is designed based on this microwave identification technology. To sum up, this scheme has a certain practical value.
Ionospheric fbEs short⁃term prediction based on Transformer
Zhang Meng1; Huang Chaojun1; Feng Jian2; Feng Jing2The ionospheric Es layer shielding frequency (fbEs) is a key parameter affecting the propagation of radio waves. The accurate prediction of fbEs has important application value in the fields of communication and navigation. In this paper, a prediction method based on Transformer deep learning model is proposed to improve the accuracy of the traditional prediction model for fbEs. A model with historical fbEs observation data and auxiliary parameters such as solar activity index as input and fbEs sequence in the next 72 hours as output is constructed, and the fbEs in Beijing, Haikou and Lhasa are predicted and analyzed. The results show that the prediction accuracy of the model based on Transformer is better than those of the traditional methods such as the model based on ARIMA and the model based on LSTM, and its mean absolute error (MAE) is 12.5% lower on average than that of the model based on LSTM, and its root mean square error (RMSE) is 13.98% lower on average than that of the model based on LSTM. In addition, its prediction error shows geographical differences (The prediction errors of the three stations in Beijing, Haikou, and Lhasa increase in sequence.) and seasonal changes (high in summer and low in winter), which is highly consistent with the physical mechanism of the ionosphere. The experiments verified the feasibility of the model based on Transformer in the prediction of ionospheric fbEs parameters.
Improved lemur optimisation algorithm for solving UAV 3D path planning in mountainous areas
Cheng Qing; Hu Haoxuan; Hu WenhaiThe traditional heuristic methods are slow, have low accuracy and are difficult to provide collision⁃free paths quickly in complex 3D environments, so an improved lemur optimization (ILO) algorithm is proposed. This algorithm integrates the advantages of the crayfish optimization algorithm (COA). The parameter C in the COA is adaptively decreased nonlinearly, and the Lévy flight step size is incorporated into the COA. The jump rate of the original LO algorithm is updated smoothly by the Sigmoid function. A Gaussian function model is utilized to simulate the mountain environment. The objective function establishes a mathematical model of unmanned aerial vehicle (UAV) flight based on the constraint conditions. The fitness function is used to determine the minimum cost of flight that can avoid obstacles within a specified airspace, and cubic spline interpolation is employed to smooth the flight path. Performance tests of the ILO algorithm were conducted on the test set CEC2017. The comparative results of multi⁃algorithm UAV path planning experiments show that the ILO algorithm generates high⁃quality and smooth paths in less iterations, overcomes the premature convergence and insufficient local search ability of the traditional genetic algorithms, adapts to complex terrain, and has reliable performance, so it provides an efficient solution for solving the UAV 3D path planning.
Design of rotating mechanical component monitoring system based on wireless power supply
Li Jiaxing12; Liang Guangsheng2; Liu Shoubao1; Xiong Zhonghao3; Huang Wei14Because of the power supply and communication wiring difficulties caused by the rotating mechanical components in the monitoring system, and the disadvantages of the existing conductive slip ring technology, such as easy wear and short life, this paper designed a rotating mechanical component monitoring system based on wireless power supply. Firstly, theoretical analysis, simulation verification and hardware design were conducted on the LCC⁃S type magnetically coupled wireless power transmission system, realizing a wireless power supply subsystem capable of non⁃contact wireless power transmission. Secondly, a monitoring subsystem was designed, which could collect the temperature and vibration data of multiple points in the rotating mechanical components in real time and transmit them to the upper computer via wireless communication. Finally, the effectiveness of the system was verified by building a rotating experimental platform to simulate the rotational motion of mechanical components. To sum up, the designed system provides a feasible technical solution for the wireless monitoring of rotating mechanical components.