Peng Zhang

To address the challenges in synthetic aperture radar (SAR) ship detection, including complex target back-grounds, multi-scale ships, and diverse orientations, while also meeting the lightweight requirements for satellite-based appli-cations, we propose a lightweight multi-scale feature fusion network—MSCF-Net. First, we design a lightweight DRC2F module that enhances the network's multi-scale feature extr-action capability through a two-stage residual mechanism and multi-scale depth-wise separable dilated convolutions. Second, to overcome the limitations of traditional feature fusion methods in SAR ship detection, we propose a multi-scale channel feature fusion pyramid network, MSCF-FPN, which effectively sup-presses background noise interference while highlighting foreground target features through multi-modal pooling and dynamic feature calibration mechanisms. Finally, to further improve detection accuracy on SAR images characterized by low target-background discriminability and diverse target orient-ations, we propose a multi-branch decoupled detection head integrated with receptive-field attention to improve the detection head's capacity to perceive spatial as well as orientation information. Experimental results demonstrate that MSCF-Net achieves detection accuracies of 79.1% (+4.9%) and 92.1% (+2%) on the SRSDD-v1.0 and HRSID, respectively, with mean average precisions (mAP) of 71.7% (+5.8%) and 93.3% (+0.2%). Further-more, the number of model parameters is decreased from 10.89M to 4.66M, a reduction of approximately 57.2%, striking an effective balance between detection performance and model efficiency. In addition, MSCF-Net exhibits robust generalization capabilities on large-scene SAR images, rendering it well-suited for application in complicated real-world scenarios.

Semantic segmentation is a crucial task in computer vision with broad applications in autonomous driving, intelligent surveillance, drone vision, and other fields. The current high-precision segmentation models generally suffer from large parameter sizes, high computational complexity, and substantial memory consumption, which limits their efficient deployment in embedded systems and resource-constrained environments. In addition, traditional methods exhibit significant limitations in handling multi-scale targets and object boundaries, particularly during deep feature extraction, where the loss of shallow spatial information often results in blurred boundaries and reduced segmentation accuracy. To address these challenges, we propose EfficientSegNet, a lightweight and efficient semantic segmentation network. This network features an innovative architecture that integrates the Cascade-Attention Dense Field (CADF) module and the Dynamic Weighting Feature Fusion (DWF) module, effectively reducing computational resource requirements while balancing global semantic information and local detail recovery. Experimental results demonstrate that EfficientSegNet achieves an excellent balance between segmentation accuracy and computational efficiency on multiple public datasets, providing robust support for real-time segmentation tasks and applications on resource-constrained devices.

In recent years, the rapid development of lightweight convolutional neural networks (CNNs) and lightweight vision transformers (ViTs) has led to significant progress in the field of mobile computing. However, deploying facial recognition models on low-power devices (with power consumption below 10 watts) remains challenging. To address this issue, we designed a lightweight facial recognition network specifically optimized for low-power devices—MSKFaceNet (Multi-Scale Kernels Face Network). First, we propose a novel lightweight convolutional neural network module called MSKFNet. MSKFNet adopts a bottleneck design and introduces variable kernel convolutions from VarKNet, combined with channel shuffle and structural re-parameterization techniques, establishing an efficient CNN module for embedded systems. Built upon the MSKFNet module, MSKFaceNet further integrates a lightweight SE module to enhance its feature representation capabilities. Finally, we designed a real-time facial recognition attendance system based on MSKFaceNet and developed a prototype device. Experimental results showthat MSKFaceNet, with only 0.54M parameters and 0.25 GFLOPS, achieves a recognition accuracy of 99.39% on the LFW dataset while delivering an inference speed of 10.8ms on the Jetson Nano platform. The proposed attendance system effectively and accurately performs facial recognition and attendance recording, significantly improving the efficiency and fairness of attendance management.

We revisit the design of convolutional kernels in lightweight convolutional neural networks, and inspired by the recent advances in RepLKNet, we design a Variable Kernel Convolutional Network module VarKNet, which solves the problem of the imbalance between depthwise convolution and pointwise convolution in the case of depthwise separable convolution when the network width is large, and enriches the model’s receptive field. The VarKNet module adopts a multi-branch structure during training and is re-parameterized and fused into a single-path structure during inference to maintain the strong expressive ability of the model and improve the inference speed. In order to further enhance the information exchange between channels, VarKNet adds channel shuffling in the fused branches. Built on VarKNet, we designed a large-scale face recognition network VarKFaceNet. VarKFaceNet achieved A great achievement of 99.5% accuracy on the LFW dataset with 0.7M parameters and 0.24 GFLOPS. At the same time, the measured speed on the NVIDIA Jetson Nano platform is 159 times, 4.2 times, and 2.4 times that of ResNet-50, EfficientNet, and MobileFaceNet, respectively. VarKFaceNet excels in balancing speed and accuracy and is quite suitable for embedded devices with limited resources.

There are not many high-precision, portable digital compass solutions available right now that can enhance combined navigation systems’ overall functionality. Additionally, there is a dearth of writing about these products. This is why a tunnel magnetoresistance (TMR) sensor-based high-precision portable digital compass system is designed. First, the least-squares method is used to compensate for compass inaccuracy once the ellipsoid fitting method has corrected manufacturing and installation errors in the digital compass system. Second, the digital compass’s direction angle data is utilized to offset the combined navigation system’s mistake. The final objective is to create a high-performing portable TMR digital compass system that will enhance the accuracy and stability of the combined navigation system (abbreviated as CNS). According to the experimental results, the digital compass’s azimuth accuracy was 4.1824° before error compensation and 0.4580° after it was applied. The combined navigation system’s path is now more accurate overall and is closer to the reference route than it was before the digital compass was added. Furthermore, compared to the combined navigation route without the digital compass, the combined navigation route with the digital compass included is more stable while traveling through the tunnel. It is evident that the digital compass system’s design can raise the integrated navigation system’s accuracy and stability. The integrated navigation system’s overall performance may be somewhat enhanced by this approach.

For the state estimation problem of a multi-source localization nonlinear system with unknown and bounded noise, a distributed sequential ellipsoidal intersection fusion estimation algorithm based on the dual set-membership filtering method is proposed to ensure the reliability of the localization system. First, noise with unknown and bounded characteristics is modeled by using bounded ellipsoidal regions. At the same time, local estimators are designed at the sensor link nodes to filter out the noise interference in the localization system. The local estimator is designed using the dual set-membership filtering algorithm. It uses the dual principle to find the minimizing ellipsoid that can contain the nonlinear function by solving the optimization problem with semi-infinite constraints, and a first-order conditional gradient algorithm is used to solve the optimization problem with a low computational complexity. Meanwhile, the communication confusion among multiple sensors causes the problem of unknown correlation. The obtained estimates of local filters are fused at the fusion center by designing a distributed sequential ellipsoid intersection fusion estimation algorithm to obtain more accurate fusion localization results with lower computational cost. Finally, the stability and reliability of the proposed distributed fusion algorithm are verified by designing a simulation example of a multi-source nonlinear system.

Micro-electro-mechanical system (MEMS) inertial devices are small volume, lightweight, low cost, and have mass-production characteristics. The development trend of inertial modules is to reduce cost and improve accuracy, and batch calibration of MEMS devices is one of the most feasible solutions to reduce cost. In this paper, we propose a distributed calibration method based on system-level, discrete calibration. The distributed calibration method requires only one or a few rotations of the combined and arranged devices to excite the individual error parameters of the inertial instruments. In this study, the relationship between the error parameters and the navigation error was rewritten using equivalence transformation, and the 24 error parameters of the device were identified by the distributed least-squares estimation using the velocity error as the observed quantity. In simulation experiments, this method could calibrate more MEMS devices simultaneously than the traditional calibration method with the exact accuracy requirement.

This paper investigates the problem of distributed ellipsoidal intersection (DEI) fusion estimation for linear time-varying multi-sensor complex systems with unknown input disturbances and measurement data transmission delays. For the problem with external unknown input disturbance signals, a non-informative prior distribution is used to model the problem. A set of independent random variables obeying Bernoulli distribution is also used to describe the situation of measurement data transmission delay caused by network channel congestion, and appropriate buffer areas are added at the link nodes to retrieve the delayed transmission data values. For multi-sensor systems with complex situations, a minimum mean square error (MMSE) local estimator is designed in a Bayesian framework based on the maximum a posteriori (MAP) estimation criterion. In order to deal with the unknown correlations among the local estimators and to select the fusion estimator with lower computational complexity, the fusion estimator is designed using ellipsoidal intersection (EI) fusion technique, and the consistency of the estimator is demonstrated. In this paper, the difference between DEI fusion and distributed covariance intersection (DCI) fusion and centralized fusion estimation is analyzed by a numerical example, and the superiority of the DEI fusion method is demonstrated.

Although face recognition has achieved great success due to deep learning, many factors may affect the quality of faces in the wild, such as pose changes, age variations, and light changes, which can seriously affect the performance of face recognition. In this work, an effective approach called Efficient Lightweight Attention Networks (ELANet) is proposed to address the challenge brought by the impacts of poses and ages on face recognition performance. First, similar local patches are particularly important when the geometry and appearance of a face change drastically. To alleviate this challenge, spatial attention is used to capture important locally similar patches and channel attention is employed to focus on features with different levels of importance. Furthermore, Efficient Fusion Attention (EFA) module is designed to achieve better performance, which can alleviate the computational effort required by fusing spatial and channel attention. Second, multi-scale features learning is necessary because pose or large expression changes can cause similar recognition regions to appear at different scales. For this purpose, pyramid multi-scale module is presented, which constructs a series of features at different scales via pooling operations. Third, to unite low-level local detail information with high-level semantic information, the features of different layers are fused by Adaptively Spatial Feature Fusion (ASFF) instead of simply utilizing addition or concatenation. Compared to recent lightweight networks, the ELANet improved performance by 1.83% and 2.17% on the CPLFW and VGG2_FP datasets, respectively, and by 0.92% on the CALFW dataset. The ELANet addresses the challenge regarding the impacts of poses and ages on face recognition performance with few parameters and computational effort and is suitable for embedded and mobile devices.