Sensors

Research

20 pages, 103920 KB

Open AccessArticle

Validating Foundation Models for Automated Cattle Detection

by Petra Pejić, Andrej Bošnjak, Robert Cupec, Emmanuel Karlo Nyarko, Josip Job and Boris Lukić

Sensors 2026, 26(16), 5074; https://doi.org/10.3390/s26165074 - 10 Aug 2026

Viewed by 261

Automated monitoring of cattle behavior through computer vision requires robust detection as a foundational step for tracking, re-identification, and behavior analysis. However, training accurate detection models typically demands extensive manually annotated datasets, creating a significant bottleneck for scaling cattle monitoring systems. In this [...] Read more.

Automated monitoring of cattle behavior through computer vision requires robust detection as a foundational step for tracking, re-identification, and behavior analysis. However, training accurate detection models typically demands extensive manually annotated datasets, creating a significant bottleneck for scaling cattle monitoring systems. In this work, we investigate whether automated annotation using foundation models can reliably replace manual labeling for cattle detection tasks. We introduce EMA (Extensive Mitrovac Annotations), a dataset of barn images with manually annotated cows with oriented bounding boxes including head orientation, posture labels (standing/lying), and visibility status (whole/partially visible). We systematically compare manualy annotated oriented bounding boxes with those generated by the Segment Anything Model 3 (SAM 3), demonstrating high agreement between automated and ground truth annotations. Furthermore, we train YOLO11-OBB detectors on both manual and SAM-generated annotations, showing that models trained on automated annotations achieve comparable performance to those trained on manual labels when evaluated on our ground truth test set. Our analysis reveals that only a small fraction of SAM-annotated data is sufficient to achieve high detection accuracy, proving the feasibility of automated annotation at scale. These findings suggest that foundation models show promise for generating training data in cattle detection pipelines under controlled conditions, potentially reducing annotation costs and supporting scalable deployment of monitoring systems. The EMA dataset and trained models are publicly available to support further research in precision livestock farming. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

27 pages, 12718 KB

Open AccessArticle

SGL-Mamba: Structure-Aware Global–Local Mamba for Crack Segmentation

by Mozi Gao, Zhonghua Weng, Jiasheng Wu, Xiaoman Qi, Guanghui Liao, Qinying Zou and Mengyu Wang

Sensors 2026, 26(16), 5040; https://doi.org/10.3390/s26165040 - 8 Aug 2026

Viewed by 371

Abstract

Accurate crack segmentation based on optical sensor imagery is of paramount importance for infrastructure health monitoring and disaster early warning. However, in complex natural scenes, cracks typically exhibit characteristics such as being thin and elongated, multi-branched, and having low contrast. Existing segmentation models [...] Read more.

Accurate crack segmentation based on optical sensor imagery is of paramount importance for infrastructure health monitoring and disaster early warning. However, in complex natural scenes, cracks typically exhibit characteristics such as being thin and elongated, multi-branched, and having low contrast. Existing segmentation models struggle to balance low computational overhead with the simultaneous modeling of global topological continuity and the precise extraction of local details. To overcome the aforementioned limitations, we introduce a Structure-Aware Global–Local Mamba (SGL-Mamba). Firstly, the SGL-Mamba Block is designed in the encoding stage, which significantly enhances the joint modeling capacity for continuous topological structures and edge textures through the synergy of a parallel directional scanning mechanism and a local perception branch. Secondly, we design a High–Low Frequency Separation Enhancement (HLFSE) module to reconstruct the skip connections. This module leverages frequency decoupling to adaptively suppress high-frequency background noise and alleviate the semantic gap. Finally, in the decoding stage, the Deformable Large Kernel Attention (D-LKA) is integrated, utilizing a dynamic spatial receptive field to precisely adapt to irregular crack orientations. Extensive experiments on three public datasets (Crack500, DeepCrack, and CrackMap) demonstrate that SGL-Mamba outperforms other state-of-the-art (SOTA) methods, achieving an F1 score (F1) of 0.7982 ± 0.0038 and an mIoU of 0.8011 ± 0.0036 on the Crack500 dataset. While ensuring lightweight architecture and high computational efficiency, the proposed method provides an effective and practical solution for automatic crack detection. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

25 pages, 3454 KB

Open AccessArticle

Mitigating Spectral Imbalance and Detail Attenuation in RGB-Thermal Object Detection via Frequency-Guided Multimodal Fusion

by Quan Du, Ming Zhao, Lu Song, Minnan Hu, Zhengqiang Wang and Wangyu Wu

Sensors 2026, 26(13), 4145; https://doi.org/10.3390/s26134145 - 1 Jul 2026

Viewed by 489

Abstract

RGB-T object detection combines visible texture information with thermal saliency cues to improve detection under degraded illumination. Existing RGB-T fusion methods usually perform feature interaction in the spatial domain or treat spectral responses jointly, which may allow coarse background components to dominate the [...] Read more.

RGB-T object detection combines visible texture information with thermal saliency cues to improve detection under degraded illumination. Existing RGB-T fusion methods usually perform feature interaction in the spatial domain or treat spectral responses jointly, which may allow coarse background components to dominate the fusion process while weakening boundary and small-target details. In addition, the repeated upsampling and aggregation operations in the detection neck can further smooth high-frequency responses preserved during early fusion. This paper proposes F²Net, a frequency-guided RGB-T object detection framework built on a dual-stream YOLOv11s architecture. The method decomposes RGB and thermal features into low- and high-frequency components for separate cross-modal fusion, mitigates detail attenuation during neck decoding, and regularizes spatial correspondence between RGB and thermal representations during training. On M3FD, F²Net achieves 89.6% mAP@0.5 and 62.1% mAP@0.5:0.95, improving the Dual-YOLOv11s baseline by 7.7 and 6.6 percentage points, respectively, while increasing the parameter count from 13.8M to 15.4M and GFLOPs from 33.9G to 35.6G. Additional experiments on LLVIP and KAIST evaluate the method under low-light and road-scene conditions. The KAIST results show that high-IoU localization remains challenging in dense and occluded pedestrian scenes. This indicates that frequency-guided fusion mainly strengthens target response generation and moderate-IoU detection, but it does not fully solve precise boundary regression under severe occlusion and weak contour conditions. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

21 pages, 3551 KB

Open AccessArticle

Progressive Pixel-Neighborhood Deformable Cross-Attention for Multispectral Object Detection

by Tian Qiu, Jifeng Shen and Xin Zuo

Sensors 2026, 26(12), 3825; https://doi.org/10.3390/s26123825 - 16 Jun 2026

Viewed by 430

Abstract

Effective cross-modal feature alignment and interaction are central challenges in multispectral object detection. Although global cross-attention provides strong long-range modeling ability, its quadratic complexity with respect to feature size limits deployment on resource-constrained platforms. We therefore propose Progressive Pixel-Neighborhood Deformable Cross-Attention for multispectral [...] Read more.

Effective cross-modal feature alignment and interaction are central challenges in multispectral object detection. Although global cross-attention provides strong long-range modeling ability, its quadratic complexity with respect to feature size limits deployment on resource-constrained platforms. We therefore propose Progressive Pixel-Neighborhood Deformable Cross-Attention for multispectral feature fusion, termed PNAFusion. The proposed framework is motivated by two observations: weak misalignment between visible and thermal images is usually concentrated around local neighborhoods, and semantic correspondence across modalities often follows non-linear spatial mappings that fixed receptive fields cannot model well. To address these issues, PNAFusion incorporates local spatial priors into its architectural design to concentrate feature interaction and alignment on the most relevant neighborhoods. Specifically, a Pixel-Neighborhood Cross-Attention (PNCA) module is introduced to avoid redundant global feature matching and suppress background noise. Meanwhile, an Adaptive Deformable Alignment (ADA) module captures non-linear spatial correspondences through learned pixel-wise offsets. These components are further integrated through an iterative feedback mechanism to progressively refine cross-modal feature alignment. Experiments on FLIR, M3FD, and DroneVehicle show that PNAFusion achieves 84.2, 90.5, and 85.5 mAP@0.5, respectively, under the YOLOv5 detector, and further reaches 86.8 mAP@0.5 on FLIR and 90.8 mAP@0.5 on M3FD when transferred to Co-DETR. Efficiency analysis indicates that PNAFusion reduces allocated GPU memory by 33.0% compared with ICAFusion and reduces theoretical FLOPs from 194.8 G to 156.4 G, although the deformable sampling and iterative refinement introduce additional latency. These results demonstrate that PNAFusion provides a practical accuracy–memory trade-off for weakly aligned multispectral object detection. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

19 pages, 2754 KB

Open AccessArticle

Frequency-Aware Refinement Network with Multi-Scale Fusion for Remote Sensing Change Detection

by Xu Zhang, Yue Du, Zeyu Zhang and Kaihua Zhang

Sensors 2026, 26(11), 3538; https://doi.org/10.3390/s26113538 - 3 Jun 2026

Viewed by 420

Abstract

Remote sensing change detection (RSCD) identifies land cover variations by comparing bi-temporal images. However, conventional methods relying solely on RGB domain information often fail to distinguish changed objects from visually similar backgrounds, especially in complex scenarios. To overcome this limitation, we propose a [...] Read more.

Remote sensing change detection (RSCD) identifies land cover variations by comparing bi-temporal images. However, conventional methods relying solely on RGB domain information often fail to distinguish changed objects from visually similar backgrounds, especially in complex scenarios. To overcome this limitation, we propose a frequency-aware refinement network (FARNet) that follows a coarse-to-fine strategy. In the first stage, we design a frequency-aware module (FAM) that learns frequency domain information to identify the blurred boundaries of changed objects that resemble the background, enabling coarse localization of potential change regions. In the second stage, recognizing that high-resolution RGB domain details provide richer spatial information than frequency-domain features, we design a refinement fusion module (RFM) that leverages these RGB details to correct and refine segmentation boundaries, ensuring precise detection. Finally, edge loss is applied to preserve high-frequency details, enhancing the precision of change detection. Extensive experiments on benchmark datasets demonstrate that FARNet significantly outperforms existing methods, achieving superior accuracy and robustness in complex change detection scenarios. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

24 pages, 5414 KB

Open AccessArticle

SW-Net: A Direction-Aware Deep Learning Model for Shipwreck Segmentation in Side-Scan Sonar Imagery

by Jiani Dai and Jie He

Sensors 2026, 26(11), 3483; https://doi.org/10.3390/s26113483 - 1 Jun 2026

Viewed by 764

Abstract

Side-scan sonar is a critical instrument for underwater cultural heritage preservation, as it allows large-scale detection of shipwrecks in turbid waters where optical methods fail. However, the automated segmentation of these targets remains a significant challenge, as severe speckle noise and complex seabed [...] Read more.

Side-scan sonar is a critical instrument for underwater cultural heritage preservation, as it allows large-scale detection of shipwrecks in turbid waters where optical methods fail. However, the automated segmentation of these targets remains a significant challenge, as severe speckle noise and complex seabed reverberations often obscure the distinctive geometric features of submerged structures. To address this challenge, this paper proposes SW-Net, which utilizes a multi-scale input strategy and a novel Directional Filter Bank to inject physical priors into the feature extraction process. Furthermore, by coupling this with a directional attention mechanism, the network dynamically modulates structural features to accurately segment targets despite intensity inversions and speckle noise. As demonstrated by the experimental results on the AI4Shipwrecks dataset, the SW-Net outperforms seven representative segmentation architectures, achieving the highest intersection over union of 39.43% and an F1-score of 56.56%. In addition, the model exhibits superior robustness against complex seabed interference while maintaining the lowest computational complexity of 4.01 million parameters among the evaluated methods. Taken together, the SW-Net is proposed to offer a practical solution for shipwreck detection on resource-constrained autonomous underwater vehicles. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

20 pages, 2605 KB

Open AccessArticle

Spatial-Frequency Decoupling Alignment Encoding for Remote Sensing Change Detection

by Xu Zhang, Yue Du, Weiran Zhou and Kaihua Zhang

Sensors 2026, 26(6), 1979; https://doi.org/10.3390/s26061979 - 21 Mar 2026

Cited by 1 | Viewed by 841

Abstract

Existing remote sensing change detection methods often struggle to accurately capture the contours of complex change targets and subtle textural differences. This makes it difficult to effectively distinguish between the boundaries of change targets and the background. To address this challenge, we propose [...] Read more.

Existing remote sensing change detection methods often struggle to accurately capture the contours of complex change targets and subtle textural differences. This makes it difficult to effectively distinguish between the boundaries of change targets and the background. To address this challenge, we propose a novel method called spatial-frequency decoupling alignment encoding (SDA-Encoding), which is designed to fully leverage information from both the spatial and frequency domains. Specifically, we first use a Transformer encoder to extract bi-temporal features. Next, we apply wavelet transform to decouple these features into low-frequency and high-frequency components. In the multi-scale high-frequency interaction (MHI) module, we combine local spatial enhancement using spatial pyramid pooling with cross-scale dependency supplementation via the dual-domain alignment fusion (DAF) module. Meanwhile, in the position-aware low-frequency enhancement (PLE) module, spatial position sensitivity is restored using coordinate attention, and region-level contextual dependencies are captured through the selective fusion attention (SFA) module. Finally, the two frequency-domain branches are complementarily fused within the spatial domain to achieve unified detection of both fine-grained and structural changes. Experimental results on three benchmark datasets demonstrate the significant performance improvements of SDA-Encoding. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

22 pages, 2817 KB

Open AccessArticle

A Dual-Branch Spatial Interaction and Multi-Scale Separable Aggregation Driven Hybrid Network for Infrared Image Super-Resolution

by Jiajia Liu, Wenxiang Dong, Xuan Zhao, Jianhua Liu and Xiaoguang Tu

Sensors 2026, 26(4), 1332; https://doi.org/10.3390/s26041332 - 19 Feb 2026

Viewed by 645

Abstract

Single image super-resolution (SISR) is a classical computer vision task that aims to reconstruct a high-resolution image from a low-resolution input, thereby improving detail sharpness and visual quality. In recent years, convolutional neural network (CNN)-based methods and transformer-based methods using self-attention mechanisms have [...] Read more.

Single image super-resolution (SISR) is a classical computer vision task that aims to reconstruct a high-resolution image from a low-resolution input, thereby improving detail sharpness and visual quality. In recent years, convolutional neural network (CNN)-based methods and transformer-based methods using self-attention mechanisms have achieved significant progress in visible-image super-resolution. However, the direct application of these two types of methods to infrared images still poses considerable challenges. On the one hand, infrared images generally suffer from low signal-to-noise ratio, blurred edges, and missing details, and relying only on local convolutions makes it difficult to adequately model long-range dependencies across regions. On the other hand, although pure transformer models have a strong global modeling ability, they usually have large numbers of parameters and are sensitive to the amount of training data, making it difficult to balance efficiency and detail restoration in infrared imaging scenarios. To address these issues, we propose a hybrid neural network architecture for infrared image super-resolution reconstruction, termed RDSR (Residual Dual-branch Separable Super-Resolution Network), which organically integrates multi-scale depthwise separable convolutions with shifted-window self-attention. Specifically, we design a dual-branch spatial interaction module (BDSI, Dual-Branch Spatial Interaction) and a multi-scale separable spatial aggregation module (MSSA, Multi-Scale Separable Spatial Aggregation). The BDSI module models correlations along rows and columns through grouped convolutions in the horizontal and vertical directions, effectively strengthening the spatial information interaction between the convolution branch and the self-attention branch. The MSSA module replaces the conventional MLP with three parallel depthwise separable convolution branches, improving the feature representation and nonlinear modeling through multi-scale spatial aggregation and a star-shaped gating operation. The experimental results on multiple public infrared image datasets show that for

\times 2

and

\times 4

upscaling, the proposed RDSR achieves higher PSNR and SSIM values than CNN-based methods such as EDSR, RCAN, and RDN, as well as transformer-based methods such as SwinIR, DAT, and HAT, demonstrating the effectiveness of the proposed modules and the overall framework. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

27 pages, 4033 KB

Open AccessArticle

DCDW-YOLOv11: An Intelligent Defect-Detection Method for Key Transmission-Line Equipment

by Dezhi Wang, Riqing Song, Minghui Liu, Xingqian Wang, Chengyu Zhang, Ziang Wang and Dongxue Zhao

Sensors 2026, 26(3), 1029; https://doi.org/10.3390/s26031029 - 4 Feb 2026

Cited by 1 | Viewed by 1075

Abstract

The detection of defects in key transmission-line equipment under complex environments often suffers from insufficient accuracy and reliability due to background interference and multi-scale feature variations. To address this issue, this paper proposes an improved defect detection model based on YOLOv11, named DCDW-YOLOv11. [...] Read more.

The detection of defects in key transmission-line equipment under complex environments often suffers from insufficient accuracy and reliability due to background interference and multi-scale feature variations. To address this issue, this paper proposes an improved defect detection model based on YOLOv11, named DCDW-YOLOv11. The model introduces deformable convolution C2f_DCNv3 in the backbone network to enhance adaptability to geometric deformations of targets, and incorporates the convolutional block attention module (CBAM) to highlight defect features while suppressing background interference. In the detection head, a dynamic head structure (DyHead) is adopted to achieve cross-layer multi-scale feature fusion and collaborative perception, along with the WIoU loss function to optimize bounding box regression and sample weight allocation. Experimental results demonstrate that on the transmission-line equipment defect dataset, DCDW-YOLOv11 achieves an accuracy, recall, and mAP of 94.4%, 92.8%, and 96.3%, respectively, representing improvements of 2.8%, 7.0%, and 4.4% over the original YOLOv11, and outperforming other mainstream detection models. The proposed method can provide high-precision and highly reliable defect detection support for intelligent inspection of transmission lines in complex scenarios. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

15 pages, 3967 KB

Open AccessArticle

Low-Light Image Segmentation on Edge Computing System

by Sung-Chan Choi and Sung-Yeon Kim

Sensors 2026, 26(1), 327; https://doi.org/10.3390/s26010327 - 4 Jan 2026

Viewed by 1300

Abstract

Segmenting low-light images, such as images showing cracks on tunnel walls, is challenging due to limited visibility. Hence, we need to combine image brightness enhancement and a segmentation algorithm. We introduce essential preliminaries, specifically highlighting deep learning-based low-light image enhancement methods and the [...] Read more.

Segmenting low-light images, such as images showing cracks on tunnel walls, is challenging due to limited visibility. Hence, we need to combine image brightness enhancement and a segmentation algorithm. We introduce essential preliminaries, specifically highlighting deep learning-based low-light image enhancement methods and the pixel-level image segmentation algorithm. After that, we provide a three-step low-light image segmentation algorithm. The proposed algorithm begins with brightness and contrast enhancement of low-light images, followed by accurate segmentation using a U-Net model. By various experimental results, we show the performance metrics of the proposed low-light image segmentation algorithm and compare the proposed algorithm’s performance against several baseline models. Furthermore, we demonstrate the implementation of the proposed low-light image segmentation pipeline on an edge computing platform. The implementation results show that the proposed algorithm is sufficiently fast for real-time processing. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

20 pages, 3806 KB

Open AccessArticle

Fusing Multi-Temporal Context for Image Super-Resolution Reconstruction in Cultural Heritage Monitoring

by Caiyan Chen, Fulong Chen, Sheng Gao, Hongqiang Li, Xinru Zhang and Yanni Cheng

Sensors 2026, 26(1), 228; https://doi.org/10.3390/s26010228 - 30 Dec 2025

Cited by 2 | Viewed by 832

Abstract

Effective conservation of World Heritage Sites relies on high-precision and continuous dynamic monitoring of their status. However, cloud cover, limitations in sensor resolution, and the vast distribution of heritage areas make it challenging to consistently acquire high-resolution imagery for key years, thereby hindering [...] Read more.

Effective conservation of World Heritage Sites relies on high-precision and continuous dynamic monitoring of their status. However, cloud cover, limitations in sensor resolution, and the vast distribution of heritage areas make it challenging to consistently acquire high-resolution imagery for key years, thereby hindering accurate characterization of their temporal evolution. To overcome this bottleneck, this paper proposes a temporal change-aware super-resolution reconstruction model. This model innovatively utilizes the temporal evolution information of heritage landscapes as a key clue for reconstructing high-quality imagery of the target year. We design a multi-branch architecture that takes the low-resolution image of the target year as the core input, while also incorporating the high- and low-resolution images from its preceding (t − 1) and subsequent (t + 1) years. Through parallel encoding branches, the model separately learns to: (1) extract spatial features from the multi-temporal low-resolution images, and (2) explicitly model the change patterns recorded in the high-resolution imagery from year t − 1 to t + 1, via a dedicated temporal change encoder. Finally, by deeply fusing these features, the model generates a simulated high-resolution image for the target year (t). Experimental results on a real-world dataset of the Weiyang Palace (WYP) core area (2017–2019), with 2018 as the target year, demonstrate that the proposed method achieves superior performance, significantly outperforming traditional single-image super-resolution models and a contrastive model without explicit temporal change modeling. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

30 pages, 7695 KB

Open AccessEditor’s ChoiceArticle

RTUAV-YOLO: A Family of Efficient and Lightweight Models for Real-Time Object Detection in UAV Aerial Imagery

by Ruizhi Zhang, Jinghua Hou, Le Li, Ke Zhang, Li Zhao and Shuo Gao

Sensors 2025, 25(21), 6573; https://doi.org/10.3390/s25216573 - 25 Oct 2025

Cited by 7 | Viewed by 4312

Abstract

Real-time object detection in Unmanned Aerial Vehicle (UAV) imagery is critical yet challenging, requiring high accuracy amidst complex scenes with multi-scale and small objects, under stringent onboard computational constraints. While existing methods struggle to balance accuracy and efficiency, we propose RTUAV-YOLO, a family [...] Read more.

Real-time object detection in Unmanned Aerial Vehicle (UAV) imagery is critical yet challenging, requiring high accuracy amidst complex scenes with multi-scale and small objects, under stringent onboard computational constraints. While existing methods struggle to balance accuracy and efficiency, we propose RTUAV-YOLO, a family of lightweight models based on YOLOv11 tailored for UAV real-time object detection. First, to mitigate the feature imbalance and progressive information degradation of small objects in current architectures multi-scale processing, we developed a Multi-Scale Feature Adaptive Modulation module (MSFAM) that enhances small-target feature extraction capabilities through adaptive weight generation mechanisms and dual-pathway heterogeneous feature aggregation. Second, to overcome the limitations in contextual information acquisition exhibited by current architectures in complex scene analysis, we propose a Progressive Dilated Separable Convolution Module (PDSCM) that achieves effective aggregation of multi-scale target contextual information through continuous receptive field expansion. Third, to preserve fine-grained spatial information of small objects during feature map downsampling operations, we engineered a Lightweight DownSampling Module (LDSM) to replace the traditional convolutional module. Finally, to rectify the insensitivity of current Intersection over Union (IoU) metrics toward small objects, we introduce the Minimum Point Distance Wise IoU (MPDWIoU) loss function, which enhances small-target localization precision through the integration of distance-aware penalty terms and adaptive weighting mechanisms. Comprehensive experiments on the VisDrone2019 dataset show that RTUAV-YOLO achieves an average improvement of 3.4% and 2.4% in mAP50 and mAP50-95, respectively, compared to the baseline model, while reducing the number of parameters by 65.3%. Its generalization capability for UAV object detection is further validated on the UAVDT and UAVVaste datasets. The proposed model is deployed on a typical airborne platform, Jetson Orin Nano, providing an effective solution for real-time object detection scenarios in actual UAVs. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

21 pages, 3489 KB

Open AccessArticle

GA-YOLOv11: A Lightweight Subway Foreign Object Detection Model Based on Improved YOLOv11

by Ning Guo, Min Huang and Wensheng Wang

Sensors 2025, 25(19), 6137; https://doi.org/10.3390/s25196137 - 4 Oct 2025

Cited by 5 | Viewed by 2855

Abstract

Modern subway platforms are generally equipped with platform screen door systems to enhance safety, but the gap between the platform screen doors and train doors may cause passengers or objects to become trapped, leading to accidents. Addressing the issues of excessive parameter counts [...] Read more.

Modern subway platforms are generally equipped with platform screen door systems to enhance safety, but the gap between the platform screen doors and train doors may cause passengers or objects to become trapped, leading to accidents. Addressing the issues of excessive parameter counts and computational complexity in existing foreign object intrusion detection algorithms, as well as false positives and false negatives for small objects, this article introduces a lightweight deep learning model based on YOLOv11n, named GA-YOLOv11. First, a lightweight GhostConv convolution module is introduced into the backbone network to reduce computational resource waste in irrelevant areas, thereby lowering model complexity and computational load. Additionally, the GAM attention mechanism is incorporated into the head network to enhance the model’s ability to distinguish features, enabling precise identification of object location and category, and significantly reducing the probability of false positives and false negatives. Experimental results demonstrate that in comparison to the original YOLOv11n model, the improved model achieves 3.3%, 3.2%, 1.2%, and 3.5% improvements in precision, recall, mAP@0.5, and mAP@0.5: 0.95, respectively. In contrast to the original YOLOv11n model, the number of parameters and GFLOPs were reduced by 18% and 7.9%, respectfully, while maintaining the same model size. The improved model is more lightweight while ensuring real-time performance and accuracy, designed for detecting foreign objects in subway platform gaps. Full article

(This article belongs to the Special Issue Image Processing and Analysis for Object Detection: 3rd Edition)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Image Processing and Analysis for Object Detection: 3rd Edition

Share This Special Issue

Editor

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Related Special Issue

Published Papers (13 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI