Road horizontal alignment contains three elements: straight line,circular arc,and clothoid.In AutoCAD,clothoid can only be fitted by polyline or spline,and the graphics which are separated from the road data are indep...Road horizontal alignment contains three elements: straight line,circular arc,and clothoid.In AutoCAD,clothoid can only be fitted by polyline or spline,and the graphics which are separated from the road data are independent of each other.Therefore,it is necessary to develop a new curve for road horizontal alignment.Firstly,the differential approximation and series integration methods for clothoid were discussed.Secondly,the geometric formulae for road line which contain the three elements were derived as a whole.Then,the advantages and feasibility of customizing a road line class,which was derived from the curve base class,were analyzed based on ObjectARX(AutoCAD Runtime eXtension) techniques.Finally,the data structure and operations for the road line class were stressed.The experimental results show that the road line can integrate the road elements,graphics and data to implement the graphics-oriented design,which can be widely used in road alignment design.展开更多
为解决现有的基于深度学习的结直肠息肉检测算法计算复杂或检测精度较低,不能在速度和精度方面同时满足实时检测的问题,提出了一种基于单阶段目标检测网络YOLO(you only look once)v5s的结直肠腺瘤实时检测方法.在YOLOv5s的主干网中融...为解决现有的基于深度学习的结直肠息肉检测算法计算复杂或检测精度较低,不能在速度和精度方面同时满足实时检测的问题,提出了一种基于单阶段目标检测网络YOLO(you only look once)v5s的结直肠腺瘤实时检测方法.在YOLOv5s的主干网中融入通道注意力机制,并以BCEWithLogitsLoss代替其原有的交叉熵损失函数BCELogits,以此提升网络性能.选取2074张腺瘤图片和包含19700帧的20段腺瘤视频,按照3∶1的比例构建结直肠腺瘤训练集和测试集.测试结果表明,结直肠腺瘤检测的平均精度为93.6%,检测速度为93帧/s,验证了该系统可以在肠镜的退镜过程中实时检测腺瘤性息肉,且具有较好的检测性能.展开更多
基金Project(51108049)supported by the National Natural Science Foundation of ChinaProject(kfj090207)supported by Open Fund of Key Laboratory of Road Structure and Material of Ministry of Transport(Changsha University of Science and Technology),China
文摘Road horizontal alignment contains three elements: straight line,circular arc,and clothoid.In AutoCAD,clothoid can only be fitted by polyline or spline,and the graphics which are separated from the road data are independent of each other.Therefore,it is necessary to develop a new curve for road horizontal alignment.Firstly,the differential approximation and series integration methods for clothoid were discussed.Secondly,the geometric formulae for road line which contain the three elements were derived as a whole.Then,the advantages and feasibility of customizing a road line class,which was derived from the curve base class,were analyzed based on ObjectARX(AutoCAD Runtime eXtension) techniques.Finally,the data structure and operations for the road line class were stressed.The experimental results show that the road line can integrate the road elements,graphics and data to implement the graphics-oriented design,which can be widely used in road alignment design.
文摘为解决现有的基于深度学习的结直肠息肉检测算法计算复杂或检测精度较低,不能在速度和精度方面同时满足实时检测的问题,提出了一种基于单阶段目标检测网络YOLO(you only look once)v5s的结直肠腺瘤实时检测方法.在YOLOv5s的主干网中融入通道注意力机制,并以BCEWithLogitsLoss代替其原有的交叉熵损失函数BCELogits,以此提升网络性能.选取2074张腺瘤图片和包含19700帧的20段腺瘤视频,按照3∶1的比例构建结直肠腺瘤训练集和测试集.测试结果表明,结直肠腺瘤检测的平均精度为93.6%,检测速度为93帧/s,验证了该系统可以在肠镜的退镜过程中实时检测腺瘤性息肉,且具有较好的检测性能.
