针对树木三维重构过程中面临的处理速度慢、重构精度低等问题,提出一种采用激光点云数据的单木骨架三维重构方法。首先,根据点云数据类型确定组合滤波方式,以去除离群点和地面点;其次,采用一种基于内部形态描述子(ISS)和相干点漂移算法(...针对树木三维重构过程中面临的处理速度慢、重构精度低等问题,提出一种采用激光点云数据的单木骨架三维重构方法。首先,根据点云数据类型确定组合滤波方式,以去除离群点和地面点;其次,采用一种基于内部形态描述子(ISS)和相干点漂移算法(CPD)的混合配准算法(Intrinsic Shape-Coherent Point Drift,IS-CPD),以获取单棵树木的完整点云数据;最后,采用Laplace收缩点集和拓扑细化相结合的方法提取骨架,并通过柱体构建枝干模型,实现骨架三维重构。试验结果表明,相比传统CPD算法,研究设计的配准方案精度和执行速度分别提高50%和95.8%,最终重构误差不超过2.48%。研究结果证明可有效地重构单棵树木的三维骨架,效果接近树木原型,为构建林木数字孪生环境和林业资源管理提供参考。展开更多
In order to establish a restoring-force model for modified concrete columns with recycled aggregates concrete(RAC), cyclic loading tests were carried out on five concretes with RAC columns and ordinary concrete frame ...In order to establish a restoring-force model for modified concrete columns with recycled aggregates concrete(RAC), cyclic loading tests were carried out on five concretes with RAC columns and ordinary concrete frame columns under the combined influence with different admixtures and admixtures ratios(silica fume and hybrid fiber). The expressions for characteristic nodes of the skeleton curve were given by the analysis and numerical regression of the test results. In addition, the hysteretic rules of the restoring-force model and the expression for unloading stiffness were presented. Finally, we summed up the complete calculation method of the hysteretic restoring force, whose results were in good agreement with experiment. The results demonstrated that the proposed model could simulate and reflect the corresponding hysteretic behaviors, and the calculation method can provide the theoretical basis for the engineering application.展开更多
文摘针对树木三维重构过程中面临的处理速度慢、重构精度低等问题,提出一种采用激光点云数据的单木骨架三维重构方法。首先,根据点云数据类型确定组合滤波方式,以去除离群点和地面点;其次,采用一种基于内部形态描述子(ISS)和相干点漂移算法(CPD)的混合配准算法(Intrinsic Shape-Coherent Point Drift,IS-CPD),以获取单棵树木的完整点云数据;最后,采用Laplace收缩点集和拓扑细化相结合的方法提取骨架,并通过柱体构建枝干模型,实现骨架三维重构。试验结果表明,相比传统CPD算法,研究设计的配准方案精度和执行速度分别提高50%和95.8%,最终重构误差不超过2.48%。研究结果证明可有效地重构单棵树木的三维骨架,效果接近树木原型,为构建林木数字孪生环境和林业资源管理提供参考。
基金Project(51178388)supported by the National Natural Science Foundation of ChinaProject(2013SZS01-Z02)supported by Key Laboratory Fund of Shaanxi Province,China
文摘In order to establish a restoring-force model for modified concrete columns with recycled aggregates concrete(RAC), cyclic loading tests were carried out on five concretes with RAC columns and ordinary concrete frame columns under the combined influence with different admixtures and admixtures ratios(silica fume and hybrid fiber). The expressions for characteristic nodes of the skeleton curve were given by the analysis and numerical regression of the test results. In addition, the hysteretic rules of the restoring-force model and the expression for unloading stiffness were presented. Finally, we summed up the complete calculation method of the hysteretic restoring force, whose results were in good agreement with experiment. The results demonstrated that the proposed model could simulate and reflect the corresponding hysteretic behaviors, and the calculation method can provide the theoretical basis for the engineering application.
文摘识别非驾驶行为是提高驾驶安全性的重要手段之一。目前基于骨架序列和图像的融合识别方法具有计算量大和特征融合困难的问题。针对上述问题,本文提出一种基于多尺度骨架图和局部视觉上下文融合的驾驶员行为识别模型(skeleton-image based behavior recognition network,SIBBR-Net)。SIBBR-Net通过基于多尺度图的图卷积网络和基于局部视觉及注意力机制的卷积神经网络,充分提取运动和外观特征,较好地平衡了模型表征能力和计算量间的关系。基于手部运动的特征双向引导学习策略、自适应特征融合模块和静态特征空间上的辅助损失,使运动和外观特征间互相引导更新并实现自适应融合。最终在Drive&Act数据集进行算法测试,SIBBR-Net在动态标签和静态标签条件下的平均正确率分别为61.78%和80.42%,每秒浮点运算次数为25.92G,较最优方法降低了76.96%。
