基于车载激光点云的道路交叉口检测与识别

道路交叉口是道路交通网的重要组成部分,其位置和类型是高精地图、自动驾驶等应用服务的基础数据.目前研究多关注车载激光点云的道路边界提取,较少关注道路交叉口类型识别.为此,本文提出一种基于动态图神经网络的道路交叉口分类方法.首先分析地面超体素的几何和空间分布差异进行提取道路边界点;然后计算道路边界点曲率,利用滑动窗口中曲率...     

钛合金钻削加工及其新发展

从钛合金自身的切削加工性能出发,系统介绍了钛合金的切削加工基本原则。阐述了传统钛合金在钻削加工中常出现的几点问题:钻头材料、钻头几何参数分析和冷却介质的选择。根据钛合金的导热率低、弹性模量低、硬度较低和化学活性强等特点合理地调整钻头材料的选择,对原有钻头几何参数进行优化,正确选取冷却介质。并介绍了当前钛合金切削加工技术的新发展。     

膨胀石墨负载纳米铁的制备及其对水中Cr(Ⅵ)及染料的去除

为有效去除印花废水中的染料和花筒剥铬时形成的Cr(Ⅵ),合成了纳米零价铁(GS-NZVI)和膨胀石墨负载纳米零价铁(GS-EG-NZVI)2种新型吸附剂.借助场发射扫描电子显微镜、能谱仪、X射线衍射仪等对二者进行表征.探讨了GS-EG-NZVI对水中的Cr(Ⅵ)和染料去除效果以及动力学和稳定性.结果表明:纳米级零价铁成...     

Multitask bilateral learning for real‐time image enhancement

Nowadays, deep neural networks (DNNs) for image processing are becoming more complex; thus, reducing computational cost is increasingly important. This study highlights the construction of a DNN for real‐time image processing, training various image processing operators efficiently through multitask learning. For real‐time image processing, the proposed algorithm takes a joint upsampling approach through bilateral guided upsampling. For multitask learning, the overall network is based on an encoder‐decoder architecture, which consists of encoding, processing, and decoding components, in which the encoding and decoding components are shared by all the image processing operators. In the processing component, a semantic guidance map, which contains processing information for each image processing operator, is estimated using simple linear shifts of the shared deep features. Through these components, the proposed algorithm requires an increase of only 5% in the number of parameters to add another image processing operator and achieves faster and higher performance than that of deep‐learning‐based joint upsampling methods in local image processing as well as global image processing.     

Terahertz narrow bandstop, broad bandpass filter using double-layer S-shaped metamaterials

Abstract In this study, double-layer S-shaped metamaterials （MMs） are analyzed by terahertz time-domain spectroscopy. These materials exhibit narrow bandstop and broad bandpass transmission properties at both horizontal and vertical electric-field polarizations. A 117% increase in the unloaded quality factor is experimen- tally observed for these materials. The center frequency is approximately 0.45 THz, with a 3-dB bandwidth of 0.52 THz from 0.20 to 0.72 THz at normal incidence. The measured average insertion loss is 0.5 dB with a ripple of 1 dB. These results show that double-layer S-shaped MMs are effective in designing tunable terahertz devices.     

基于BCH和HOG的Mean Shift跟踪算法

为解决基于核的Mean Shift跟踪算法在颜色相近时,容易出现跟踪不准确或者跟踪丢失的问题,同时提高跟踪的可靠性与准确度,分析了近年来的目标跟踪情况,提出一种基于分块颜色直方图(BCH)和方向梯度直方图(HOG)的Mean Shift跟踪算法.其中BCH包含目标的空间信息,能够很好地应对形变;HOG能够很好地解决光照变化和阴影的影响.使用标准数据集进行测试,结果表明:所提算法可以达到高于80帧/s的跟踪速度,分别优于跟踪-学习-检测(TLD)和核相关滤波(KCF)算法1.7％和2.2％.     

一遍完成的平衡布点外存模型简化算法

现存的自适应采样的外存模型简化算法均需要多次读取原模型，算法效率较低。该文给出一种仅仅需要读取原模型一遍的自适应顶点聚类算法－平衡布点算法（Balanced Tilling,BT），用于外存模型简化。其关键思想在于通过表面编码记录模型表面，通过对原模型的二次量化（quadric quantization）得到原模型上的细节分布。该算法可以定位出所有类型的细节区域，而其它一些算法只能定位细节边。细节区域将被进一步细化，而平滑区域将被进一步简化。该算法大大减少了输入输出时间，尤其适合处理超大规模模型。内存需求很小，只与输出模型规模有关。     

Controllability and local accessibility - a normal form approach

Given a system with an uncontrollable linearization at the origin, we study the controllability of the system at equilibria around the origin. If the uncontrollable mode is nonzero, we prove that the system always has other equilibria around the origin. We also prove that these equilibria are linearly controllable provided a coefficient in the normal form is nonzero. Thus, the system is qualitatively changed from being linearly uncontrollable to linearly controllable when the equilibrium point is moved from the origin to a different one. This is called a bifurcation of controllability. As an application of the bifurcation, systems with a positive uncontrollable mode can be stabilized at a nearby equilibrium point. In the last part of this paper, simple sufficient conditions are proved for local accessibility of systems with an uncontrollable mode. Necessary conditions of controllability and local accessibility are also proved for systems with a convergent normal form.     

基于神经网络的柴油发动机故障预测研究

为了保证设备安全可靠地运行,需要对设备进行故障预测,及时消除故障;探讨了故障预测技术的应用及神经网络预测原理和建模的方法,给出了神经网络模型在预测过程中的学习算法和改进的神经网络的训练算法,并用于柴油发动机运行中的故障预测,经过多种柴油机的试验证明试验结果满意,对故障的实时检测有重要意义,具有良好的应用前景.     

一种基于Markov链模型的动态聚类方法

对单变量时间序列的聚类，是一类有着广泛应用背景的特殊的聚类问题。由于该问题的特殊性，现有的聚类方法无法直接使用，故提出了一种新的基于Markov链模型的动态聚类方法。该方法首先对每一个时间序列建立一个描述其动态特征的Markov链模型，从而把对时间序列的聚类问题转化为对Markov链模型的聚类问题。然后通过定义各个Markov链之间的“距离”，采用动态聚类算法完成对这些Markov链模型的聚类，使用该方法，分别对一经真实数据和仿真数据进行了聚类试验，都获得了比较好的聚类结果。