基于改进YOLOv3的桥梁底部裂缝目标检测方法

来源 :中国科技论文 | 被引量 : 0次 | 上传用户:ysd007
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
为实现桥梁裂缝的快速、准确定位,考虑光照变化、污渍阴影等干扰因素的影响,提出一种结合桥梁检测机和改进单阶段目标检测(you only look once version 3,YOLOv3)算法的桥梁裂缝检测方法.首先,在预处理阶段,采用改进自适应Mask匀光算法对数据集进行处理,矫正阴影和光照不均等问题,提高算法环境适应能力;其次,在目标检测阶段,针对桥梁裂缝的特点,对数据集使用k-means++算法聚类先验框以适应裂缝的不同尺寸,采用广义交并比对YOLOv3损失函数进行改进以提高定位精度;最后,采用迁移学习对YOLOv3进行训练.实验结果表明,在迭代140个epoch后,检测速度可达到31帧/s,平均精度(average precision,AP)达到94.88%,相比于采用原始数据集的原始YOLOv3网络AP值提高了13.16%,能够满足实时性和高精度的检测要求.
其他文献
Botulinum neurotoxins serotype A(BoNT/A)is the deadliest toxins known to humans and the“Category A”agent for bioterrorism.Over the past 20 years,significant efforts have been put forth to develop effective inhibitors of BoNT/A.Unfortunately,few identified
SPASER nanoparticle(NP),with small size,ultranarrow spectral-line and good biocompatibility,is a potential biomedical nanoprobe.However,owning to the striking light absorption capacity of Au-resonator,huge energy could accumulate under strong pump-ing-las
Core-shell nanostructures constituted by a plasmonic core and an ultrathin shell have drawn enormous attentions in enhanced spectroscopies,catal-ysis,energy,and other fields because of their splendid properties such as multifunctionality,stability,and adj
Amyloid-β protein(Aβ)is a fatal cause of Alzheimer\'s disease,which can trigger a series of cytotoxicity by the abnormal aggregation of Aβ in human brain.The strategies for inhibition and disaggregation of Aβ fibrillation are mostly based on the interac
Chiral pools have been commonly utilized as inexpensive and versatile building blocks in organic synthesis.In this account,we provide an overview of our recent synthetic efforts toward polycyclic natural products using chiral natural small molecules(chira
Nano-catalysis plays a vital role in the industrially chemical transformations.The increasingly severe energy crisis and environmental issues spur sci-entists to exploit novel nano-catalysts in diverse fields.It is essential to understand the fundamental
Chemically active motion is ubiquitous in many nature and artificial systems where each unit can move directionally by converting chemical energy into kinetic energy.Since the systems are driven by activity far from equilibrium,many dynamical behaviors fo
针对高光谱图像语义分割中空间信息利用不充分的问题,提出了一种基于多任务学习的语义分割算法,分别为语义分割及基于遥感指数的图像重建任务.网络主要由3D卷积和2D卷积组成,通过主成分分析(principal component analysis,PCA)来减少冗余信息,通过不同任务的重要程度优化两者的损失函数权重.2个特征在分割任务中融合送入由3D通道注意力及空间池化金字塔(spatial pyramid pooling,SPP)组成的光谱-空间特征,提取模块获取谱间相关性及空间上下文信息,从而实现更好的分割
为增强遮挡场景下所提取行人特征的判别力,有效挖掘样本类别信息,提出了一种基于特征关联和多损失融合的行人再识别方法.首先利用姿态估计器生成的人体关键点作为辅助信息,引导模型关注行人图像未被遮挡区域,提取姿态引导的全局特征;其次引入全局对比池化模块,将平均池化和最大池化的特性进行融合,提取对背景噪声和遮挡抗干扰性更强的全局特征;然后引入One-vs-rest关系模块,挖掘局部分块特征的内在关系,提取能够反映图像整体信息的局部特征;最后将交叉熵损失、难样本采样三元组损失和中心损失这3种损失函数进行加权融合,监督
Organic room-temperature phosphorescence(ORTP)materials have attracted widespread attention in the fields of OLED,optical imaging and sensors.Herein,we successfully prepared a long afterglow IPA-MA crystalline films using iso-phthalic acid(IPA)and melamin