在复杂构造、复杂岩性、复杂地表条件等地区进行油气地震勘探面临诸多难题,其中剧烈起伏的地表已成为制约上述地区地震勘探发展的瓶颈之一。最小二乘逆时偏移(Least-Squares Reverse Time Migration,LSRTM)相对于常规偏移具有更高的成...在复杂构造、复杂岩性、复杂地表条件等地区进行油气地震勘探面临诸多难题,其中剧烈起伏的地表已成为制约上述地区地震勘探发展的瓶颈之一。最小二乘逆时偏移(Least-Squares Reverse Time Migration,LSRTM)相对于常规偏移具有更高的成像分辨率、振幅保真性及均衡性等优势,但基于矩形网格的LSRTM在面对复杂地表时无法很好地适用山前带等剧烈起伏地形。此外,山前带地震资料中包含较多的噪声,T分布相比Huber范数和混合模,在缺失数据的条件下更稳健,且没有多余参数,因而简单实用,而Huber范数和混合模的结果严重依赖参数选取,需要大量的尝试。为此,将全交错网格引入贴体网格,将T分布推广到起伏地表LSRTM,进一步推导了贴体网格线性Born正演方程,在此基础上提出了基于贴体全交错网格的起伏地表LSRTM算法,较好地克服了起伏地形的影响。模型试算验证了算法的有效性和对复杂模型的适应性。展开更多
睡眠呼吸暂停综合征(Sleep Apnea Syndrome,SAS)是一种常见的睡眠呼吸系统疾病.目前有很多关于SAS自动检测的研究仅仅是在一段时间内判断是否发病,无法识别发病具体时段.针对这一局限性,本文提出一种新的SAS识别模型SD-FCE(SAS Detectio...睡眠呼吸暂停综合征(Sleep Apnea Syndrome,SAS)是一种常见的睡眠呼吸系统疾病.目前有很多关于SAS自动检测的研究仅仅是在一段时间内判断是否发病,无法识别发病具体时段.针对这一局限性,本文提出一种新的SAS识别模型SD-FCE(SAS Detection based on Functional Connectivity of Electroencephalography),该模型具备识别不规则发病时段的能力.首先,为提取发病时段的神经系统异常,本文利用脑电功能连接性构建脑电功能连接矩阵.其次,为识别发病的具体时段,本文基于目标检测算法改进,针对SAS设计不同的预选框将不规则目标网格化,以进行模型搭建.分类与位置回归模块依托于网格化处理的结果输出发病的分类结果、中心时间与持续时间.对比以往相关研究,SD-FCE模型的识别结果更利于医学诊断,同时性能优于以往其它类似模型.展开更多
Through the analysis and comparison of shortcomings and advantages of existing technologies on object modeling in 3D applications,we propose a new modeling method for virtual scene based on multi-view image sequence t...Through the analysis and comparison of shortcomings and advantages of existing technologies on object modeling in 3D applications,we propose a new modeling method for virtual scene based on multi-view image sequence to model irregular objects efficiently in 3D application.In 3D scene,this method can get better visual effect by tracking the viewer's real-time perspective position and projecting the photos from different perspectives dynamically.The philosophy of design,the steps of development and some other relevant topics are discussed in details,and the validity of the algorithm is analyzed.The results demonstrate that this method represents more superiority on simulating irregular objects by applying it to the modeling of virtual museum.展开更多
文摘在复杂构造、复杂岩性、复杂地表条件等地区进行油气地震勘探面临诸多难题,其中剧烈起伏的地表已成为制约上述地区地震勘探发展的瓶颈之一。最小二乘逆时偏移(Least-Squares Reverse Time Migration,LSRTM)相对于常规偏移具有更高的成像分辨率、振幅保真性及均衡性等优势,但基于矩形网格的LSRTM在面对复杂地表时无法很好地适用山前带等剧烈起伏地形。此外,山前带地震资料中包含较多的噪声,T分布相比Huber范数和混合模,在缺失数据的条件下更稳健,且没有多余参数,因而简单实用,而Huber范数和混合模的结果严重依赖参数选取,需要大量的尝试。为此,将全交错网格引入贴体网格,将T分布推广到起伏地表LSRTM,进一步推导了贴体网格线性Born正演方程,在此基础上提出了基于贴体全交错网格的起伏地表LSRTM算法,较好地克服了起伏地形的影响。模型试算验证了算法的有效性和对复杂模型的适应性。
文摘睡眠呼吸暂停综合征(Sleep Apnea Syndrome,SAS)是一种常见的睡眠呼吸系统疾病.目前有很多关于SAS自动检测的研究仅仅是在一段时间内判断是否发病,无法识别发病具体时段.针对这一局限性,本文提出一种新的SAS识别模型SD-FCE(SAS Detection based on Functional Connectivity of Electroencephalography),该模型具备识别不规则发病时段的能力.首先,为提取发病时段的神经系统异常,本文利用脑电功能连接性构建脑电功能连接矩阵.其次,为识别发病的具体时段,本文基于目标检测算法改进,针对SAS设计不同的预选框将不规则目标网格化,以进行模型搭建.分类与位置回归模块依托于网格化处理的结果输出发病的分类结果、中心时间与持续时间.对比以往相关研究,SD-FCE模型的识别结果更利于医学诊断,同时性能优于以往其它类似模型.
文摘Through the analysis and comparison of shortcomings and advantages of existing technologies on object modeling in 3D applications,we propose a new modeling method for virtual scene based on multi-view image sequence to model irregular objects efficiently in 3D application.In 3D scene,this method can get better visual effect by tracking the viewer's real-time perspective position and projecting the photos from different perspectives dynamically.The philosophy of design,the steps of development and some other relevant topics are discussed in details,and the validity of the algorithm is analyzed.The results demonstrate that this method represents more superiority on simulating irregular objects by applying it to the modeling of virtual museum.
