Challenges such as perspective distortion, irregular reading regions, and complex backgrounds in natural scenes hinder the accuracy and efficiency of automatic meter reading systems. Current mainstream approaches predominantly utilize object-detection-based methods without optimizing for text characteristics, while enhancements in detection robustness under complex backgrounds typically focus on data preprocessing rather than model architecture. To address these limitations, a novel end-to-end framework, i.e., EDPNet (Efficient DB and PARSeq Network), is proposed to integrate efficient boundary detection and text recognition. EDPNet comprises two key components, EDNet for detection and EPNet for recognition, where EDNet employs EfficientNetV2-s as its backbone with the Multi-Scale KeyDrop Attention (MSKA) and Efficient Multi-scale Attention (EMA) mechanisms to address perspective distortion and complex background challenges, respectively. During the recognition stage, EPNet integrates a DropKey Attention module into the PARSeq encoder, enhancing the recognition of irregular readings while effectively mitigating overfitting. Experimental evaluations show that EDNet achieves an F1-score of 0.997988, outperforming DBNet++ (ResNet50) by 0.61%. In challenging scenarios, EDPNet surpasses state-of-the-art methods by 0.7~1.9% while reducing parameters by 20.03%. EPNet achieves 90.0% recognition accuracy, exceeding the current best performance by 0.2%. The proposed framework delivers superior accuracy and robustness in challenging conditions while remaining lightweight.
To address the demand for lightweight, high-precision, real-time, and low-computation detection of targets with limited samples—such as laboratory instruments in portable AR devices—this paper proposes a small dataset object detection algorithm based on a hierarchically deployed attention mechanism. The algorithm adopts Rep-YOLOv8 as its backbone. First, an ECA channel attention mechanism is incorporated into the backbone network to extract image features and adaptively adjust channel weights, improving performance with only a minor increase in parameters. Second, a CBAM-spatial module is integrated to enhance region-specific features for small dataset objects, highlighting target characteristics and suppressing irrelevant background noise. Then, in the neck network, the SE attention module is replaced with an eSE attention module to prevent channel information loss caused by dimensional changes. Experiments conducted on both open-source and self-constructed small datasets show that the proposed hierarchical Rep-YOLOv8 model effectively meets the requirements of lightweight design, real-time processing, high accuracy, and low computational cost. On the self-built small dataset, the model achieves a mAP@0.5 of 0.971 across 17 categories, outperforming the baseline Rep-YOLOv8 (0.871) by 11.5%, demonstrating effective recognition and segmentation capability for small dataset objects.
Traditionally, retrieving both temperature and CO2 concentration in atmospheric remote sensing has relied on two independent lidar systems, leading to increased system complexity and limited coordination. To address this challenge, we propose a coordinated retrieval approach using a three-wavelength differential absorption lidar (DIAL) system. A temperature-sensitive wavelength is selected to distinguish strong absorption from weak absorption, forming the tri-wavelength configuration. By exploiting the different sensitivities of absorption cross-sections to thermal and molecular variations, simultaneous retrieval of both parameters is achieved. A standard atmospheric profile under clean-air conditions is constructed. The CO2 absorption spectrum near 1573 nm is generated using Voigt line shapes and data from the HITRAN database. Extinction and backscatter coefficients are retrieved through the Klett method. A layer-by-layer solution of the coupled differential equations is then performed to extract temperature and concentration simultaneously. Results are benchmarked against the atmospheric model, demonstrating the feasibility of the approach. This method provides a promising pathway for high-precision, multi-parameter DIAL sensing.
To address the issues of insufficient localization accuracy, low matching efficiency, and high cost of fingerprint library construction in complex indoor environments, a multi-source fusion localization method based on mutated signal segmentation fitting (MSSF) and an improved adaptive unscented Kalman filter is proposed. Firstly, a segmentation partitioning strategy integrating Delaunay Triangulation, flood-like lighting algorithm, and localized fitting algorithm for small-area segmentation is designed to construct a double-layer fingerprint library (DLFL) based on a virtual access point signal propagation attenuation model. This achieves high-precision signal fitting and rapid zone matching under sparse sampling conditions. Secondly, an improved adaptive unscented Kalman Filter algorithm incorporating Sage–Husa filtering is proposed. It dynamically estimates the observation noise covariance through an innovation augmentation mechanism, enhancing its robustness in non-linear noise environments characterized by Wi-Fi signal fluctuations and PDR drift. Combined with the DLFL, this further realizes a localization algorithm that fuses Wi-Fi signals and PDR information. Experimental results show that the proposed MSSF algorithm achieves an average error of merely 1.508 dB across 12 test points; the DLFL based on MSSF reduces the online matching time from 0.024 s to 0.004 s; and the multi-source fusion localization algorithm achieves a per-step average error of 0.326 m, representing a 56.7% improvement compared to traditional fusion localization methods. These results validate the comprehensive advantages of the proposed method in terms of accuracy, efficiency, and adaptability.
To address the limited constraints of ground-based lidar with few channels in retrieving aerosol microphysical parameters in urban atmospheres, this study developed a method to retrieve aerosol volume size distribution and effective radius from a 355/532/1064 nm triple-wavelength elastic-scattering, single-polarization lidar system. The method uses 3β + 2α optical quantities as input constraints, applies Mie scattering theory as the forward model, parameterizes the volume size distribution with B-spline functions, and achieves stable solutions through Tikhonov regularization and cross-validation. To reduce uncertainties in prior parameters, including the complex refractive index, particle size range, and lidar ratio, an optimization strategy based on parameter search, retrieval reconstruction, and error minimization was introduced. Numerical simulations showed that the method reproduced the main features of a bimodal lognormal aerosol volume size distribution with good feasibility and stability. Two case studies further showed fine-mode dominance and decreasing extinction coefficient, depolarization ratio, and effective radius with height under good air quality conditions, but enhanced coarse-mode contribution and effective radius in the upper cloud-influenced layer under lightly polluted conditions, as inferred from the combined variations in RSCS, extinction coefficient, depolarization ratio, and effective radius.