基于三维彩色点云的物体识别算法

随着新一代深度传感器的出现,使用三维(3-D)数据成为物体识别研究的热点,而且提出了很多点云特征描述子.针对传统的采用点云形状特征描述子在目标描述方面的不足,提出了一种基于三维彩色点云的物体识别算法.首先提取点云数据的视点特征直方图(VFH)和颜色直方图(CH),然后对提取的形状特征和颜色特征分别通过支持向量机(SVM)进行预分类,最后将上述2个识别结果进行决策级融合.提出的算法在Washington RGB-D数据集进行训练和测试.结果表明,该方法与传统的采用点云形状特征描述子相比,其物体的正确识别率有了显著的提高.     

模拟技术与模拟装置

Y2001-62856-1394 0119487三维物体采用基于网格小波分析的快速多分辨率建模=Fast multiresolution modeling of 3-D objects usingmesh-based wavelet analysis[会,英]/Zha,H& Mitsu-tomi,T//2000 IEEE International Conference on Sys-tems,Man and Cybernetics.—1394～1399(NiK)本文采用基于三角网格的小波分析方法提出了三维物体的快速多分辨率建模方法。作者采用此模型用于多分辨率控制以达到复杂环境的有效可视化。参9     

3-R定位系统在最佳目标——雷达站几何下的误差传播特性

求解3-R定位系统的最佳目标——雷达站几何是CW雷达定位体制研究中的关键问题之一。本文给出了利用最优化方法求解这种最佳几何的目标函数。这种最佳几何是：三个雷达天线位于正三角形的顶点，三个天线波束在空间正交。本文给出了3-R定位体制的普适性误差椭球的二次曲面方程。研究结果表明，在上述最佳目标——雷达站几何下，椭球形误差二次曲面变为圆球形误差二次曲面。     

基于拓扑图及三维特征的三维物体识别方法

本文提出一种基于拓扑图及三维特征的三维物体识别方法。该方法在图像预处理阶段首先检测物体可视表面的闭合边界,依据闭合边界对物体进行分割,并用一个景物特征图对景物进行表达(SAGR)。在建模阶段,采用多视图表面模型方法进行建模,依据表面分割,对多视图模型用一个模型特征图进行表达(MAGR)。在物体识别阶段,将景体与模型的匹配分成三部分进行,即扫描部件、图的匹配部件及匹配检验部件。匹配首先是基于SAGR与MAGR图的拓扑性质的匹配,然后是基于三维特征的匹配。根据匹配结果,或对未知物体自动建模,或给出识别结果。     

三维网格模型的预测压缩算法研究

通过遍历顶点的位置信息和3-D网格中的角度信息,提出了一种顶点邻域预测和角度预测相结合的3-D网格模型顶点位置预测压缩算法。采用Hausdorff距离测量预测误差,试验的结果表明,提出的算法在不同的测试模型上都优于MPEG-4标准中的三维网格模型编码(3-DMC)算法,较好地减少了3-D网格顶点预测的几何失真。     

一种基于YOLOv3的水下声呐图像目标检测方法

将目标检测框架应用于水下声呐图像处理是近期的高热度话题,现有水下声呐目标检测方法多基于声呐图像的纹理特征识别不同物体,难以解决声呐图像中由于形状畸变造成的几何特征不稳定问题。为此,该文提出一种基于YOLOv3的水下物体检测模型YOLOv3F,该模型将从声呐图像中提取的纹理特征和从深度图中提取的空间几何特征相融合,利用深度图中相对稳定的空间几何特征弥补纹理特征表述能力的不足,再将融合后的特征用于目标检测。实验结果表明,所提改进模型的检测性能相较于3个基线模型在识别精度方面具有明显提升;在对单个类别的物体进行检测的情况下,与YOLOv3相比,改进模型也表现出了更出色的检测效果。     

结合超体素与PFCM的点云分割方法

为了实现点云数据的区域划分,提出了一种结合超体素与粒子群优化模糊C均值(PFCM)的聚类分割算法(SPFCM)。用随机采样一致性算法去除点云平面,根据3-D点云的空间位置、曲率以及快速直方图特征,利用八叉树体素化点云得到超体素。采用PFCM算法对超体初步划分,并对粘连的点云再划分,克服了PFCM算法对于堆叠物体无法分割及较大物体过分割的缺点, 并在OSD-v0.2数据集上对SPFCM算法进行了性能测试。结果表明,相较于PFCM算法,SPFCM不仅保留了其参量少、操作简单等优点,而且指标得到了较大提升,准确率达到86%,查全率达到83%。该研究对3-D点云复杂场景的准确分割提供了帮助与参考。     

基于注意力机制和时空信息的群组行为识别

群组行为识别是对个体的共同行为进行识别。群组行为与群组状态密不可分,也与群组内个体时空特征息息相关,时空信息既能描述空间语义信息,更能反映行为的动态变化情况。针对有效精细的时空特征提取问题,本文提出了一种基于注意力机制和深度时空信息的群组行为识别方法。首先将ShuffleAttention引入双流特征提取网络中,有效提取个体外观和运动信息。其次使用改进Non-Local网络提取深度时序信息。最后将个体特征送到图卷积网络中进行空间交互信息建模,得到群组行为识别结果。在CAD和CAED数据集上的准确率达到了93.6%和97.8%,在CAD数据集上与凝聚群组搜索算法(CCS)和成员关系图(ARG)方法相比,准确率提高了1.2%和2.6%,这表明本文方法能有效提取深度时空特征,提升群组行为识别准确率。     

一种基于2D和3DSIFT特征级融合的一般物体识别算法

如何选择合适的特征表示一般物体类间差异和类内共性至关重要,因此,本文在2D SIFT(Scale Invariant Feature Transform,SIFT)的基础上,提出了基于点云模型的3D SIFT特征描述子,进而提出一种基于2D和3D SIFT特征级融合的一般物体识别算法.分别提取物体2维图像和3维点云的2D和3D SIFT特征描述子,利用"词袋"(Bag of Words,BoW)模型得到物体特征向量,根据特征级融合将两个特征向量进行融合实现物体描述,运用有监督分类器支持向量机(Support Vector Machine,SVM)实现分类识别,给出最终识别结果.最后,实验验证了本文提出算法的好处.     

单像素几何矩探测运动物体定位技术（特邀）

针对快速实时定位运动物体的需求,提出了一种使用几何矩探测的单像素快速定位运动物体的方法。该方法的核心是通过探测运动物体的质心实现快速定位运动物体。根据几何矩性质构造3个几何矩照明光,并照射运动物体,利用单像素探测器收集运动物体与调制光相互作用后的反射或透射光的强度值。根据单像素成像理论,探测强度值与物体的零阶和一阶几何矩值相对应。标识物体位置的质心参数可由物体的零阶和一阶几何矩值获取。结合数字微镜调制器(DMD)和时间抖动的方法产生几何矩照明光,利用所提方法在不成像的前提下分别实现了帧频约为500 fps和1 000 fps运动物体定位。所提方法获取质心的误差在1.63个像素以内,均方误差为0.118 3个像素。文中所提方法为使用单像素探测器实现快速追踪运动物体提供了一种新思路。     

Plane surface detection and reconstruction using segment-based tensor voting

A Segment-based Tensor Voting (SBTV) algorithm is presented for planar surface detection and reconstruction of man-made objects. Our work is inspired by piecewise planar stereo reconstruction. During the vital procedure to detect and label the planar surface, the two main contributions are: first, tensor voting is used for obtaining the geometry attribute of the 3D points cloud. The candidate planar patches are generated through scene image segment of low variation of color and intensity. Second, we over-segment the scene image into the segment and the candidate 3D planar patch is generated. The SBTV algorithm is used on 3D points cloud sets to identify the co-plane on the candidate patch. After detecting every planar patch, the geometry architecture of object is obtained. The experiments demonstrate the effectiveness of our proposed approach on either outdoor or indoor datasets.     

Exploiting color for graph-based 3D point cloud denoising

A point cloud is a representation of a 3D scene as a discrete collection of geometry plus other attributes such as color, normal, transparency associated with each point. The traditional acquisition process of a 3D point cloud, e.g. using depth information acquired directly by active sensors or indirectly from multi-viewpoint images, suffers from a significant amount of noise. Hence, the problem of point cloud denoising has recently received a lot of attention. However, most existing techniques attempt to denoise only the geometry of each point, based on the geometry information of the neighboring points; there are very few works at all considering the problem of denoising the color attributes of a point cloud. In this paper, we move beyond the state of the art and we propose a novel technique employing graph-based optimization, taking advantage of the correlation between geometry and color, and using it as a powerful tool for several different tasks, i.e. color denoising, geometry denoising, and combined geometry and color denoising. The proposed method is based on the notion that the correct location of a point also depends on the color attribute and not only the geometry of the neighboring points, and the correct color also depends on the geometry of the neighbors. The proposed method constructs a suitable k-NN graph from geometry and color and applies graph-based convex optimization to obtain the denoised point cloud. Extensive simulation results on both real-world and synthetic point clouds show that the proposed denoising technique outperforms state-of-the-art methods using both subjective and objective quality metrics.     

基于多幅未标定图像的三维度量重构

提出了一种基于多幅未标定图像，恢复场景三维几何的优化矩阵分解算法，该算法先通过迭代逼近得到射影重构，然后通过估计绝对二次曲面，将射影重构更新到度量重构。采用真实图像测试，获得好的实验结果，说明该方法简单易用。     

NURBS曲面重构与点云-曲面误差分析

通过对某汽车零件进行光栅扫描、解相和去包裹处理,获取物体三维点云数据,对点云数据进行降噪、精简和网格化处理。然后根据点云曲率分布云图将点云数据分割成11块区域,首先对A区域点云数据进行曲面拟合,生成4×4阶均匀曲面,然后对其余分块点云数据分别进行曲面拟合,最后通过曲面延伸、拼接、倒角、修剪等处理,获取物体NURBS自由曲面,总体点云-曲面误差为0.2645 mm,并且曲面间符合G1相切连续和法向曲率连续,解决了在曲率较大的地方拟合误差较大的问题,提高了曲面的重构精度。     

基亏移动最小二乘法表面的曲率计算及应用

图像工程中,离散的点云模型表面曲率的精确计算非常重要。笔者在Levin的MLS（Moving Least—Square）表面的基础上,将原始三维点云模型的表面投影到一个MLS表面,然后直接从MLS表面计算点云曲率,并将该方法应用在隧道三维扫描点云模型数据压缩中,取得了良好的效果。     

Microstrip triangular patch antenna fabricated on LiTiZn ferrite substrate and tested in the X band range

An experimental study of microstrip triangle patch antenna designed and fabricated on LiTiZn ferrite substrate is presented. This antenna geometry has been optimized with IE3D and lithographed on 2 mm thick LiTiZn ferrite substrate. LiTiZn polycrystalline substrate has been synthesized using solid state reaction technique and further characterized for microwave application. We have computed various important antenna parameters such as return-loss, bandwidth, directivity, gain, radiation power, etc. for synthesized ferrite as well as on RT-duroid substrate. A comparison of these results has also been presented.     

基于混合属性的零样本图像分类

对于具有相似属性的类别而言,在有限维度的语义属性下,基于属性的零样本图像分类器难以对它们进行正确区分.考虑到语义属性描述类别的有限性,在直接属性预测(Direct Attribute Prediction,DAP)模型的基础上,提出一种基于混合属性的零样本图像分类模型(Hybrid Attribute-Based DAP,HA-DAP).首先,对样本的底层特征进行稀疏编码并利用编码后的非语义属性来辅助现有的语义属性;将非语义属性与语义属性构成混合属性并将其作为DAP模型的属性中间层,利用属性预测模型的思想进行混合属性分类器的训练;最后,根据预测的混合属性以及属性与类别之间的关系进行测试样本类别标签的预测.在OSR、Pub Fig以及Shoes数据集上的实验结果表明,HA-DAP的分类性能优于DAP,不仅能够取得较高的零样本图像分类精度,而且还获得了较高的AUC值.     

点云编码综述

点云编码是支撑点云广泛应用的关键技术之一,是近期技术研究和标准化领域的热点。对点云几何信息和属性信息编码技术演进进行了回顾,并针对稠密点云和稀疏点云的几种典型编码方法的编码效率进行了比较。未来点云编码研究将集中于利用帧间预测去除动态点云的不同帧之间的相关性,以及端到端点云编码、任务驱动的点云编码等方面。     

四阶模型方位NCS高分辨SAR成像

针对长积累时间SAR回波耦合性强,聚焦困难的问题,从提高成像模型近似精度和距离徙动校正精度两方面开展了SAR成像研究,推导了一种考虑方位非线性调频变标(NCS)的四阶近似成像模型。首先,为了提高模型近似度,将斜距展开到四次项;然后,为了将距离向和方位向进行初步解耦,在时域进行走动补偿,但是带来了聚焦深度问题,高分辨成像时,这种现象更明显;为此,二维频域解耦后又在方位向采用二阶非线性变标方法解决这一问题。仿真结果表明,从主观分析和量化指标两方面来讲,该算法优于传统的高分辨斜视SAR成像算法。     

Effects of directional antennas on outband D2D mmWave communications in heterogeneous networks

Device-to-Device (D2D) communications are considered as a keystone of the fifth generation wireless technology (5G). This new approach is very promising in terms of energy and spectrum efficiency. However, the integration of such communications in a typical cellular network increases inevitably the amount of interference. Several researches propose to lower the interference thanks to either sharing the cellular spectrum intelligently, or using non-cellular bands for D2D links. In this paper, we focus on the latter opportunity, and consider that the D2D communications are used with millimeter waves (mmWaves).For what comes to modeling a D2D-enabled (D2D-e) network, many works propose to use stochastic geometry so as to evaluate the impact of interference and noise on the various links. In this work, we aim to analyze the SINR and the average data rate of Outband D2D links for user equipments (UEs) with conventional omnidirectional antennas and with various directional mmWave antennas: patch antennas, horn antennas and uniform linear array antennas. Analytical and empirical evaluations of the Signal-to-Interference-plus Noise Ratio (SINR) are made with stochastic geometry. We propose to discuss the advantages and drawbacks of directional mmWave antennas in Outband D2D for various antenna designs, and their interest in various environments.