The design provisions of current seismic codes are generally not very accurate for assessing effects of near-fault ground motions on reinforced concrete （r.c.） spatial frames, because only far-fault ground motions a...The design provisions of current seismic codes are generally not very accurate for assessing effects of near-fault ground motions on reinforced concrete （r.c.） spatial frames, because only far-fault ground motions are considered in the seismic codes. Strong near-fault earth- quakes are characterized by long-duration （horizontal） pulses and high values of the ratio ~PGA of the peak value of the vertical acceleration, PGAv, to the analogous value of the horizontal acceleration, PGAH, which can become critical for girders and columns. In this work, six- and twelve-storey r.c. spatial frames are designed according to the provisions of the Italian seismic code, considering the horizontal seismic loads acting （besides the gravity loads） alone or in combination with the vertical ones. The non- linear seismic analysis of the test structures is performed using a step-by-step procedure based on a two-parameter implicit integration scheme and an initial stress-like itera- tive procedure. A lumped plasticity model based on the Haar-K^n~m principle is adopted to model the inelastic behaviour of the frame members. For the numerical investigation, five near-fault ground motions with high values of the acceleration ratio C^p6A are considered. Moreover, following recent seismological studies, which allow the extraction of the largest （horizontal） pulse from a near-fault ground motion, five pulse-type （horizontal） ground motions are selected by comparing the original ground motion with the residual motion after the pulse has been extracted. The results of the nonlinear dynamic analysis carried out on the test structures highlighted thathorizontal and vertical components of near-fault ground motions may require additional consideration in the seis- mic codes.展开更多
To proceed from sensation to movement, integration and transformation of information from different senses and reference frames are required. Several brain areas are involved in this transformation process, but previo...To proceed from sensation to movement, integration and transformation of information from different senses and reference frames are required. Several brain areas are involved in this transformation process, but previous neuroanatomical and neurophysiological studies have implicated the caudal area 7b as one particular component of this transformation system. In this study, we present the first quantitative report on the spatial coding properties of caudal area 7b. The results showed that neurons in this area had intermediate component characteristics in the transformation system; the area contained bimodal neurons, and neurons in this area encode spatial information using a hybrid reference frame. These results provide evidence that caudal area 7b may belong to the reference frame transformation system, thus contributing to our general understanding of the transformation system.展开更多
街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的...街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的问题,本文提出锚框校准和空间位置信息补偿视频实例分割(Anchor frame calibration and Spatial position information compensation for Video Instance Segmentation,AS-VIS)网络.首先,在预测头3个分支中添加锚框校准模块实现同锚框纵横比匹配的多类型感受野采样,解决目标边缘提取不充分问题.其次,设计多感受野下采样模块将各种感受野采样后的特征融合,解决下采样信息缺失问题.最后,应用多感受野下采样模块将特征金字塔低层目标区域激活特征映射嵌入到高层中实现空间位置信息补偿,解决高层特征空间细节位置信息匮乏问题.在Youtube-VIS标准库中提取街道场景视频数据集,其中包括训练集329个视频和验证集53个视频.实验结果与YolactEdge检测和分割精度指标定量对比表明,锚框校准平均精度分别提升8.63%和5.09%,空间位置信息补偿特征金字塔平均精度分别提升7.76%和4.75%,AS-VIS总体平均精度分别提升9.26%和6.46%.本文方法实现了街道场景视频序列实例级同步检测、跟踪与分割,为无人驾驶车辆环境感知提供有效的理论依据.展开更多
文摘The design provisions of current seismic codes are generally not very accurate for assessing effects of near-fault ground motions on reinforced concrete （r.c.） spatial frames, because only far-fault ground motions are considered in the seismic codes. Strong near-fault earth- quakes are characterized by long-duration （horizontal） pulses and high values of the ratio ~PGA of the peak value of the vertical acceleration, PGAv, to the analogous value of the horizontal acceleration, PGAH, which can become critical for girders and columns. In this work, six- and twelve-storey r.c. spatial frames are designed according to the provisions of the Italian seismic code, considering the horizontal seismic loads acting （besides the gravity loads） alone or in combination with the vertical ones. The non- linear seismic analysis of the test structures is performed using a step-by-step procedure based on a two-parameter implicit integration scheme and an initial stress-like itera- tive procedure. A lumped plasticity model based on the Haar-K^n~m principle is adopted to model the inelastic behaviour of the frame members. For the numerical investigation, five near-fault ground motions with high values of the acceleration ratio C^p6A are considered. Moreover, following recent seismological studies, which allow the extraction of the largest （horizontal） pulse from a near-fault ground motion, five pulse-type （horizontal） ground motions are selected by comparing the original ground motion with the residual motion after the pulse has been extracted. The results of the nonlinear dynamic analysis carried out on the test structures highlighted thathorizontal and vertical components of near-fault ground motions may require additional consideration in the seis- mic codes.
基金funded by the National Science Foundation of China (NSFC 30770700, 30670669, 30870825,30530270, 31070963, and 31070965)the 973 program(2007CB947703 and 2011CB707800)+3 种基金the Key Program of the Chinese Academy of Sciences, China (KSCX2-EW-J-23, KSCX2-YW-R-261, and KSCX2-EW-R-11)the West Light Foundation of the Chinese Academy of Sciences (0902351081)the National Natural Science Foundation of China (30921064)the project sponsored by Yunnan Development and Reform Commission (2009-1988)
文摘To proceed from sensation to movement, integration and transformation of information from different senses and reference frames are required. Several brain areas are involved in this transformation process, but previous neuroanatomical and neurophysiological studies have implicated the caudal area 7b as one particular component of this transformation system. In this study, we present the first quantitative report on the spatial coding properties of caudal area 7b. The results showed that neurons in this area had intermediate component characteristics in the transformation system; the area contained bimodal neurons, and neurons in this area encode spatial information using a hybrid reference frame. These results provide evidence that caudal area 7b may belong to the reference frame transformation system, thus contributing to our general understanding of the transformation system.
文摘街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的问题,本文提出锚框校准和空间位置信息补偿视频实例分割(Anchor frame calibration and Spatial position information compensation for Video Instance Segmentation,AS-VIS)网络.首先,在预测头3个分支中添加锚框校准模块实现同锚框纵横比匹配的多类型感受野采样,解决目标边缘提取不充分问题.其次,设计多感受野下采样模块将各种感受野采样后的特征融合,解决下采样信息缺失问题.最后,应用多感受野下采样模块将特征金字塔低层目标区域激活特征映射嵌入到高层中实现空间位置信息补偿,解决高层特征空间细节位置信息匮乏问题.在Youtube-VIS标准库中提取街道场景视频数据集,其中包括训练集329个视频和验证集53个视频.实验结果与YolactEdge检测和分割精度指标定量对比表明,锚框校准平均精度分别提升8.63%和5.09%,空间位置信息补偿特征金字塔平均精度分别提升7.76%和4.75%,AS-VIS总体平均精度分别提升9.26%和6.46%.本文方法实现了街道场景视频序列实例级同步检测、跟踪与分割,为无人驾驶车辆环境感知提供有效的理论依据.
