Unlike the detection of marked on-street parking spaces,detecting unmarked spaces poses significant challenges due to the absence of clear physical demarcation and uneven gaps caused by irregular parking.In urban citi...Unlike the detection of marked on-street parking spaces,detecting unmarked spaces poses significant challenges due to the absence of clear physical demarcation and uneven gaps caused by irregular parking.In urban cities with heavy traffic flow,these challenges can result in traffic disruptions,rear-end collisions,sideswipes,and congestion as drivers struggle to make decisions.We propose a real-time detection system for on-street parking spaces using YOLO models and recommend the most suitable space based on KD-tree search.Lightweight versions of YOLOv5,YOLOv7-tiny,and YOLOv8 with different architectures are trained.Among the models,YOLOv5s with SPPF at the backbone achieved an F1-score of 0.89,which was selected for validation using k-fold cross-validation on our dataset.The Low variance and standard deviation recorded across folds indicate the model’s generalizability,reliability,and stability.Inference with KD-tree using predictions from the YOLO models recorded FPS of 37.9 for YOLOv5,67.2 for YOLOv7-tiny,and 67.0 for YOLOv8.The models successfully detect both marked and unmarked empty parking spaces on test data with varying inference speeds and FPS.These models can be efficiently deployed for real-time applications due to their high FPS,inference speed,and lightweight nature.In comparison with other state-of-the-art models,our models outperform them,further demonstrating their effectiveness.展开更多
【目的】长输油气管道环焊缝射线底片是完整性评价与风险管控的核心依据之一,底片上的焊缝序列号、依据标准、焊缝位置标记等标志信息需实现数字化归档。传统人工判读方式工作量大、效率低、成本高,且易因视觉疲劳导致漏判、误判,亟需...【目的】长输油气管道环焊缝射线底片是完整性评价与风险管控的核心依据之一,底片上的焊缝序列号、依据标准、焊缝位置标记等标志信息需实现数字化归档。传统人工判读方式工作量大、效率低、成本高,且易因视觉疲劳导致漏判、误判,亟需研发一种兼顾识别精度与模型轻量化的智能识别方法。【方法】以YOLO(You Only Look Once)v11n为基准模型,构建面向焊缝底片标志检测的YOLO-MPWR(You Only Look Once for the Marks of Pipeline Weld Radiographs)模型,并实施以下3项关键改进措施:①设计卷积门控线性单元转换器CFCGLU(ConvFormer with Convolutional Gated Linear Unit),并嵌入C3k2模块,再利用门控机制动态分配特征权重,强化对关键字符区域的响应,抑制背景及遮挡噪声;②设计了轻量化检测头LDH(Lightweight Detection Head),采用深度可分离卷积替代标准卷积,在保持精度的同时,显著减少模型参数、降低复杂度;③引入采样算子CARAFE(Content-Aware ReAssembly of FEatures),增强了YOLO-MPWR模型对重要特征的响应、特征图语义信息的利用。【结果】以中国华南地区某长输管道焊缝射线底片为数据集进行训练验证,与YOLOv11n基准模型相比,YOLO-MPWR模型的平均精度均值(mean Average Precision,mAP)mAP@0.50提升2.5%,参数量、计算量分别降低17.2%、18.2%;与RT-DETR(Real-Time Detection Transformer)、YOLOv3tiny、YOLOv5n等9种主流模型相比,YOLO-MPWR模型在精度、参数量、计算复杂度3个方面均实现了最优,在重叠、遮挡、翻转等复杂工况下漏检率更低,且对目标边缘及不规则形状区域关注更均匀。【结论】YOLO-MPWR模型在管道焊缝射线底片标志识别中实现了“高精度+超轻量”协同突破,可满足现场实时检测需求,为管道完整性数字化管理提供了可复制的技术路径,可应用于油气站场、炼化装置、船舶焊缝等工业影像目标检测场景,具有极好的工程推广价值。展开更多
[目的/意义]针对自然环境干扰下检测模型对辣椒叶片病虫害的特征提取不充分、容易忽视目标物体的边缘信息,以及小块病斑与虫害病灶易漏检等问题,本研究提出一种轻量化辣椒叶片病害检测算法,即YOLOMDFR(You Only Look Once Version 12-MD...[目的/意义]针对自然环境干扰下检测模型对辣椒叶片病虫害的特征提取不充分、容易忽视目标物体的边缘信息,以及小块病斑与虫害病灶易漏检等问题,本研究提出一种轻量化辣椒叶片病害检测算法,即YOLOMDFR(You Only Look Once Version 12-MDFR)。[方法]基于YOLOv12s模型做出改进。首先用两个堆叠的3×3的深度可分离卷积代替一个5×5的深度可分离卷积以改进MobileNetV4,并将其代替YOLOv12s的原始骨干网络实现骨干网络轻量化。其次为提高小目标物体的特征提取能力,提出了多维频域互补自注意力机制模块(Dimensional Frequency Reciprocal Attention Mixing Transformer,D-F-Ramit)。最后利用D-F-Ramit与RAGConv(Residual Aggrega⁃tion Gate-Controlled Convolution)重新设计颈部网络,增强模型的特征融合能力和信息传递能力。基于以上改进提出YOLO-MDFR目标检测算法。[结果和讨论]实验结果表明,本研究提出的YOLO-MDFR模型在实验数据集上的平均识别精确度达到95.6%,与YOLOv12s模型相比,平均识别精度提高了2.0%,同时参数量下降了61.5%,计算量下降了68.5%,帧率达到43.4帧/s。[结论]本研究通过系统性的架构优化,在保持模型轻量化的同时显著提升了检测性能,实现了计算效率与检测精度的最佳平衡。展开更多
Remote sensing technology has been widely used for marine monitoring.However,due to the limitations of sensor technologies and data sources,effective monitoring of marine ships at night remains challenging.To address ...Remote sensing technology has been widely used for marine monitoring.However,due to the limitations of sensor technologies and data sources,effective monitoring of marine ships at night remains challenging.To address these challenges,our study developed SDGST,a high-resolution glimmer marine ship dataset from SDGSAT-1 satellite and proposed a ship detection and identification method based on the YOLOv5s model,the Glimmer YOLO model.Considering the characteristics of glimmer images,our model has made several effective improvements to the original YOLOv5s model.In particular,the improved model incorporates a new layer for detecting small targets and integrates the CA(Coordinate Attention)mechanism.To enhance the original feature fusion strategy,we introduced BiFPN(Bi-directional Feature Pyramid Network).We also adopted the EIOU Loss function and replaced the initially defined anchors with clustering results,thus improving detection performance.The mean Average Precision(mAP%)reaches 96.7%,which is a 5.1%improvement over the YOLOv5s model.Notably,it significantly improves the detection of small ships.This model demonstrates superior performance in ship detection under glimmer conditions compared to the original YOLOv5s model and other popular target detection models,and may serve as a valuable reference for achieving high-precision nighttime marine monitoring.展开更多
To map the rock joints in the underground rock mass,a method was proposed to semiautomatically detect the rock joints from borehole imaging logs using a deep learning algorithm.First,450 images containing rock joints ...To map the rock joints in the underground rock mass,a method was proposed to semiautomatically detect the rock joints from borehole imaging logs using a deep learning algorithm.First,450 images containing rock joints were selected from borehole ZKZ01 in the Rumei hydropower station.These images were labeled to establish ground truth which was subdivided into training,validation,and testing data.Second,the YOLO v2 model with optimal parameter settings was constructed.Third,the training and validation data were used for model training,while the test data was used to generate the precision-recall curve for prediction evaluation.Fourth,the trained model was applied to a new borehole ZKZ02 to verify the feasibility of the model.There were 12 rock joints detected from the selected images in borehole ZKZ02 and four geometric parameters for each rock joint were determined by sinusoidal curve fitting.The average precision of the trained model reached 0.87.展开更多
Fuel station drive-offs,wherein the drivers simply drive off without paying,are a major issue in the UK(United Kingdom)due to rising fuel costs and financial hardships.The phenomenon has increased greatly over the las...Fuel station drive-offs,wherein the drivers simply drive off without paying,are a major issue in the UK(United Kingdom)due to rising fuel costs and financial hardships.The phenomenon has increased greatly over the last few years,with reports indicating a substantial increase in such events in the major cities.Traditional prevention measures such as Avutec and Driveoffalert rely primarily on expensive infrastructure and blacklisted databases.Such systems typically involve costly camera installation andmaintenance and are consequently out of the budget of small fuel stations.These conventional approaches also fall short regarding real-time recognition,particularly regarding first-time impostors using fictitious plates,which represent an increasingly significant proportion of such forgery.This research presents an AI(Artificial Intelligence)-driven detection system using the MOT(Ministry of Transport)History API(Application Programming Interface)to scan in real-time at gas stations to recognize and prevent such fraud.The system integrates various state-of-the-art technologies to offer a foolproof system.Using the latestYOLO(YouOnly Look Once)model to recognize number plates and EasyOCR(Optical Character Recognition)to recognize characters,the system correctly reads license plates in various environmental conditions like lighting,viewpoint,and weather conditions.This approach minimizes the utilization of expensive camera systems and employs cheaper ANPR(AutomaticNumber Plate Recognition)gear,availing existing installed surveillance cameras on filling stations.The system operates with a basic web-based application to notify operators of stolen vehicles in real-time,enabling them to react immediately.Real-world testing achieves 84%success with CCTV(Closed-Circuit Television)images,depicting its real-world applicability.The results indicate that the AI-driven solution offers a monumental leap compared to current practices,giving fuel stations a cost-effective and efficient means of reducing financial loss from drive-off incidents.展开更多
基金supports this paper.Project Nos.NSTC-112-2221-E-324-003 MY3,NSTC-111-2622-E-324-002 and NSTC-112-2221-E-324-011-MY2.
文摘Unlike the detection of marked on-street parking spaces,detecting unmarked spaces poses significant challenges due to the absence of clear physical demarcation and uneven gaps caused by irregular parking.In urban cities with heavy traffic flow,these challenges can result in traffic disruptions,rear-end collisions,sideswipes,and congestion as drivers struggle to make decisions.We propose a real-time detection system for on-street parking spaces using YOLO models and recommend the most suitable space based on KD-tree search.Lightweight versions of YOLOv5,YOLOv7-tiny,and YOLOv8 with different architectures are trained.Among the models,YOLOv5s with SPPF at the backbone achieved an F1-score of 0.89,which was selected for validation using k-fold cross-validation on our dataset.The Low variance and standard deviation recorded across folds indicate the model’s generalizability,reliability,and stability.Inference with KD-tree using predictions from the YOLO models recorded FPS of 37.9 for YOLOv5,67.2 for YOLOv7-tiny,and 67.0 for YOLOv8.The models successfully detect both marked and unmarked empty parking spaces on test data with varying inference speeds and FPS.These models can be efficiently deployed for real-time applications due to their high FPS,inference speed,and lightweight nature.In comparison with other state-of-the-art models,our models outperform them,further demonstrating their effectiveness.
文摘【目的】长输油气管道环焊缝射线底片是完整性评价与风险管控的核心依据之一,底片上的焊缝序列号、依据标准、焊缝位置标记等标志信息需实现数字化归档。传统人工判读方式工作量大、效率低、成本高,且易因视觉疲劳导致漏判、误判,亟需研发一种兼顾识别精度与模型轻量化的智能识别方法。【方法】以YOLO(You Only Look Once)v11n为基准模型,构建面向焊缝底片标志检测的YOLO-MPWR(You Only Look Once for the Marks of Pipeline Weld Radiographs)模型,并实施以下3项关键改进措施:①设计卷积门控线性单元转换器CFCGLU(ConvFormer with Convolutional Gated Linear Unit),并嵌入C3k2模块,再利用门控机制动态分配特征权重,强化对关键字符区域的响应,抑制背景及遮挡噪声;②设计了轻量化检测头LDH(Lightweight Detection Head),采用深度可分离卷积替代标准卷积,在保持精度的同时,显著减少模型参数、降低复杂度;③引入采样算子CARAFE(Content-Aware ReAssembly of FEatures),增强了YOLO-MPWR模型对重要特征的响应、特征图语义信息的利用。【结果】以中国华南地区某长输管道焊缝射线底片为数据集进行训练验证,与YOLOv11n基准模型相比,YOLO-MPWR模型的平均精度均值(mean Average Precision,mAP)mAP@0.50提升2.5%,参数量、计算量分别降低17.2%、18.2%;与RT-DETR(Real-Time Detection Transformer)、YOLOv3tiny、YOLOv5n等9种主流模型相比,YOLO-MPWR模型在精度、参数量、计算复杂度3个方面均实现了最优,在重叠、遮挡、翻转等复杂工况下漏检率更低,且对目标边缘及不规则形状区域关注更均匀。【结论】YOLO-MPWR模型在管道焊缝射线底片标志识别中实现了“高精度+超轻量”协同突破,可满足现场实时检测需求,为管道完整性数字化管理提供了可复制的技术路径,可应用于油气站场、炼化装置、船舶焊缝等工业影像目标检测场景,具有极好的工程推广价值。
文摘[目的/意义]针对自然环境干扰下检测模型对辣椒叶片病虫害的特征提取不充分、容易忽视目标物体的边缘信息,以及小块病斑与虫害病灶易漏检等问题,本研究提出一种轻量化辣椒叶片病害检测算法,即YOLOMDFR(You Only Look Once Version 12-MDFR)。[方法]基于YOLOv12s模型做出改进。首先用两个堆叠的3×3的深度可分离卷积代替一个5×5的深度可分离卷积以改进MobileNetV4,并将其代替YOLOv12s的原始骨干网络实现骨干网络轻量化。其次为提高小目标物体的特征提取能力,提出了多维频域互补自注意力机制模块(Dimensional Frequency Reciprocal Attention Mixing Transformer,D-F-Ramit)。最后利用D-F-Ramit与RAGConv(Residual Aggrega⁃tion Gate-Controlled Convolution)重新设计颈部网络,增强模型的特征融合能力和信息传递能力。基于以上改进提出YOLO-MDFR目标检测算法。[结果和讨论]实验结果表明,本研究提出的YOLO-MDFR模型在实验数据集上的平均识别精确度达到95.6%,与YOLOv12s模型相比,平均识别精度提高了2.0%,同时参数量下降了61.5%,计算量下降了68.5%,帧率达到43.4帧/s。[结论]本研究通过系统性的架构优化,在保持模型轻量化的同时显著提升了检测性能,实现了计算效率与检测精度的最佳平衡。
基金funded by Operation and Maintenance Project of Big Earth Data Center of the Chinese Academy of Sciences[grant no CAS-WX2022SDC-XK13]Joint HKU-CAS Laboratory for iEarth[grant no 313GJHZ2022074MI].
文摘Remote sensing technology has been widely used for marine monitoring.However,due to the limitations of sensor technologies and data sources,effective monitoring of marine ships at night remains challenging.To address these challenges,our study developed SDGST,a high-resolution glimmer marine ship dataset from SDGSAT-1 satellite and proposed a ship detection and identification method based on the YOLOv5s model,the Glimmer YOLO model.Considering the characteristics of glimmer images,our model has made several effective improvements to the original YOLOv5s model.In particular,the improved model incorporates a new layer for detecting small targets and integrates the CA(Coordinate Attention)mechanism.To enhance the original feature fusion strategy,we introduced BiFPN(Bi-directional Feature Pyramid Network).We also adopted the EIOU Loss function and replaced the initially defined anchors with clustering results,thus improving detection performance.The mean Average Precision(mAP%)reaches 96.7%,which is a 5.1%improvement over the YOLOv5s model.Notably,it significantly improves the detection of small ships.This model demonstrates superior performance in ship detection under glimmer conditions compared to the original YOLOv5s model and other popular target detection models,and may serve as a valuable reference for achieving high-precision nighttime marine monitoring.
基金supported by the National Key R&D Program of China(No.2023YFC3081200)the National Natural Science Foundation of China(No.42077264)。
文摘To map the rock joints in the underground rock mass,a method was proposed to semiautomatically detect the rock joints from borehole imaging logs using a deep learning algorithm.First,450 images containing rock joints were selected from borehole ZKZ01 in the Rumei hydropower station.These images were labeled to establish ground truth which was subdivided into training,validation,and testing data.Second,the YOLO v2 model with optimal parameter settings was constructed.Third,the training and validation data were used for model training,while the test data was used to generate the precision-recall curve for prediction evaluation.Fourth,the trained model was applied to a new borehole ZKZ02 to verify the feasibility of the model.There were 12 rock joints detected from the selected images in borehole ZKZ02 and four geometric parameters for each rock joint were determined by sinusoidal curve fitting.The average precision of the trained model reached 0.87.
文摘Fuel station drive-offs,wherein the drivers simply drive off without paying,are a major issue in the UK(United Kingdom)due to rising fuel costs and financial hardships.The phenomenon has increased greatly over the last few years,with reports indicating a substantial increase in such events in the major cities.Traditional prevention measures such as Avutec and Driveoffalert rely primarily on expensive infrastructure and blacklisted databases.Such systems typically involve costly camera installation andmaintenance and are consequently out of the budget of small fuel stations.These conventional approaches also fall short regarding real-time recognition,particularly regarding first-time impostors using fictitious plates,which represent an increasingly significant proportion of such forgery.This research presents an AI(Artificial Intelligence)-driven detection system using the MOT(Ministry of Transport)History API(Application Programming Interface)to scan in real-time at gas stations to recognize and prevent such fraud.The system integrates various state-of-the-art technologies to offer a foolproof system.Using the latestYOLO(YouOnly Look Once)model to recognize number plates and EasyOCR(Optical Character Recognition)to recognize characters,the system correctly reads license plates in various environmental conditions like lighting,viewpoint,and weather conditions.This approach minimizes the utilization of expensive camera systems and employs cheaper ANPR(AutomaticNumber Plate Recognition)gear,availing existing installed surveillance cameras on filling stations.The system operates with a basic web-based application to notify operators of stolen vehicles in real-time,enabling them to react immediately.Real-world testing achieves 84%success with CCTV(Closed-Circuit Television)images,depicting its real-world applicability.The results indicate that the AI-driven solution offers a monumental leap compared to current practices,giving fuel stations a cost-effective and efficient means of reducing financial loss from drive-off incidents.
