Road defect detection plays a pivotal role in enhancing traffic safety,optimizing urban management,and fostering sustainable urban development.Nevertheless,the limited availability of detection resources constrains th...Road defect detection plays a pivotal role in enhancing traffic safety,optimizing urban management,and fostering sustainable urban development.Nevertheless,the limited availability of detection resources constrains the deployment and effectiveness of many existing models.To address this challenge,we propose SCD(space-to-depth convolution,ConvTranspose,distance intersection over union(DIoU))-YOLO11s(you only look once version 11 small),an enhanced variant of YOLO11s.The proposed method substantially improves detection accuracy and model adaptability for small-scale defects by integrating the SPD-Conv(space-to-depth convolution)module to capture fine-grained target features,the ConvTranspose module to mitigate resolution degradation of feature maps induced by repeated downsampling,and DIoU loss function to refine multi-scale target localization.Experimental evaluations conducted on the RDD2022 public dataset demonstrate that SCD-YOLO11s achieves a 3.4%improvement in detection accuracy and a 3.5%increase in mAP@0.5%,with only a 1.9 M parameter overhead.These findings highlight the effectiveness and practical significance of the proposed approach in advancing automatic road defect detection systems.展开更多
针对部署于有限算力平台的YOLOv3(you only look once v3)算法对电容器外观缺陷存在检测速度较慢的问题,提出了基于YOLOv3算法改进的轻量化算法MQYOLOv3。首先采用轻量化网络MobileNet v2作为特征提取模块,通过利用深度可分离式卷积替...针对部署于有限算力平台的YOLOv3(you only look once v3)算法对电容器外观缺陷存在检测速度较慢的问题,提出了基于YOLOv3算法改进的轻量化算法MQYOLOv3。首先采用轻量化网络MobileNet v2作为特征提取模块,通过利用深度可分离式卷积替换一般卷积操作,使得模型的参数量大幅度降低进而提高模型的检测速度,同时也带来了检测精度的降低;然后在网络结构中嵌入空间金字塔池化结构实现局部特征与全局特征的融合、引入距离交并比(distance intersection over union,DIoU)损失函数优化交并比(intersection over union,IoU)损失函数以及使用Mish激活函数优化Leaky ReLU激活函数来提高模型的检测精度。本文采用自制的电容器外观缺陷数据集进行实验,轻量化MQYOLOv3算法的平均精度均值(mean average precision,mAP)为87.96%,较优化前降低了1.16%,检测速度从1.5 FPS提升到7.7 FPS。实验表明,本文设计的轻量化MQYOLOv3算法在保证检测精度的同时,提高了检测速度。展开更多
基金Supported by the National Science Foundation of China(No.62571164)the Natural Science Foundation of Heilongjiang Province(PL2024F025)the Fundamental Scientific Research Funds of Heilongjiang Province(No.2022-KYYWF-1050).
文摘Road defect detection plays a pivotal role in enhancing traffic safety,optimizing urban management,and fostering sustainable urban development.Nevertheless,the limited availability of detection resources constrains the deployment and effectiveness of many existing models.To address this challenge,we propose SCD(space-to-depth convolution,ConvTranspose,distance intersection over union(DIoU))-YOLO11s(you only look once version 11 small),an enhanced variant of YOLO11s.The proposed method substantially improves detection accuracy and model adaptability for small-scale defects by integrating the SPD-Conv(space-to-depth convolution)module to capture fine-grained target features,the ConvTranspose module to mitigate resolution degradation of feature maps induced by repeated downsampling,and DIoU loss function to refine multi-scale target localization.Experimental evaluations conducted on the RDD2022 public dataset demonstrate that SCD-YOLO11s achieves a 3.4%improvement in detection accuracy and a 3.5%increase in mAP@0.5%,with only a 1.9 M parameter overhead.These findings highlight the effectiveness and practical significance of the proposed approach in advancing automatic road defect detection systems.
文摘针对部署于有限算力平台的YOLOv3(you only look once v3)算法对电容器外观缺陷存在检测速度较慢的问题,提出了基于YOLOv3算法改进的轻量化算法MQYOLOv3。首先采用轻量化网络MobileNet v2作为特征提取模块,通过利用深度可分离式卷积替换一般卷积操作,使得模型的参数量大幅度降低进而提高模型的检测速度,同时也带来了检测精度的降低;然后在网络结构中嵌入空间金字塔池化结构实现局部特征与全局特征的融合、引入距离交并比(distance intersection over union,DIoU)损失函数优化交并比(intersection over union,IoU)损失函数以及使用Mish激活函数优化Leaky ReLU激活函数来提高模型的检测精度。本文采用自制的电容器外观缺陷数据集进行实验,轻量化MQYOLOv3算法的平均精度均值(mean average precision,mAP)为87.96%,较优化前降低了1.16%,检测速度从1.5 FPS提升到7.7 FPS。实验表明,本文设计的轻量化MQYOLOv3算法在保证检测精度的同时,提高了检测速度。
