多视图立体三维重建中的孔洞修复算法

为了解决多视图立体三维重建算法不能很好地处理弱纹理或无纹理及高光区域的重建问题,提出一种基于可见外壳与多视图三维点云有机融合的多视图立体三维重建孔洞修复算法.该算法以可见外壳及多视图三维点云为输入,首先提取出可见外壳内满足点云稀疏度约束的叶节点,然后利用可见外壳法向量射线约束去除包裹在三维点云外层的叶节点,最后通过加入三维点云曲面曲率约束来消除点云中凹陷区域的影响.实验结果表明,文中算法有效地解决了物体缺乏纹理区域表面的孔洞修复问题,使得最终生成的三维网格模型完整和平滑,具有参数可调、易于实现的特点,对于不同的模型都具有非常好的鲁棒性.     

语义导向多尺度多视图深度估计算法

目前利用深度学习进行多视图深度估计的方法可以根据卷积类型可以大致分为两类.其中,基于2D卷积网络的模型预测计算速度快,但预测精度较低;基于3D卷积网络的模型预测精度高,却存在高硬件消耗.同时,多视图中相机外部参数的变化使得模型无法在物体边缘、遮挡或纹理较弱区域生成高精度预测结果.针对上述问题,提出了基于3D卷积的语义导...     

面向嵌入式平台多视图立体视觉深度感知

针对目前基于神经网络的多视图立体视觉(multi-view stereo, MVS)深度估计算法存在参数量大、内存消耗严重,难以满足当下低算力嵌入式平台的需求.提出基于MVS2D极线注意力机制与MobileNetV3-Small的MVS深度感知网络(Mobile-MVS2D).该网络采用编码器-解码器的结构,使用MobileNetV3-Small网络进行编码特征提取,对源图像与参考图像之间不同特征层的尺度信息耦合采用极线注意力机制,解码阶段引入SE-Net与跳跃连接扩展解码特征细节,提升预测精度.实验结果表明,提出的模型在ScanNet数据集中在深度图的评价指标中展现较高的精度.在与视觉SLAM结合下可以展现出较准确的三维重建效果,具有较好的鲁棒性.在Jeston Xavier NX上推理精度为Float16尺寸为640×480的图片组,仅需0.17 s, GPU消耗仅需1 GB,能够满足低算力嵌入式平台的需求.     

基于多目立体匹配的深度获取方法

提出一种深度获取方法,利用基于颜色分割的多目立体匹配算法,从多个视点图像中提取深度信息。利用mean-shift算法,根据颜色信息分割参考图像,提取图像中的颜色一致性区域,通过局部窗口匹配算法进行多目立体匹配得到多幅初始视差图,根据融合准则将多幅视差图合成为一幅视差图以提高视差图的精度并对视差图进行优化后处理,按照视差与深度的关系,将视差图转化为深度图。该算法能有效处理匹配过程中的遮挡区域,提高匹配精度和视差图的准确度。     

基于全局优化的保细节分层多视图立体匹配

传统的基于全局优化的多视网立体匹配方法由于算法复杂度和内存容量的限制,难以处理高分辨率的图像,为此提出一种保细节的层次化多视图它体匹配方法.首先对每幅分割的图像进行结构微细度评估,为分割区域设定合适的匹配分辨率层;然后在图像的金字塔结构上:进行由粗到细的视差恢复:在一定的分辨率层上,对属于该分辨率层的图像区域进行全局优...     

PMVS算法的CPU多线程和GPU两级粒度并行策略

PMVS(Patch-based Multi-View Stereo)三维重建算法被广泛应用于无人机航拍影像的三维场景重建中。针对PMVS三维重建算法计算量大、时间复杂度高的问题,提出了PMVS算法的CPU多线程和GPU两级粒度并行策略(Multithread and GPU Parallel Schema,MGPS),方法具体包括:基于GPU的PMVS算法特征提取和片面扩散的并行设计;多影像的GPU和CPU任务分配机制,以使得部分任务分配给CPU采用多线程并行,部分任务分配给GPU并行时,程序总运行时间最短。实验采用搭载24核CPU和NVIDIA Tesla K20 GPU的高性能服务器作为测试平台,针对分辨率为4081×2993的16幅无人机影像进行三维重建。实验结果表明,相比串行的PMVS算法,基于MGPS的PMVS算法取得4倍左右的加速比,其中特征提取最高加速13倍,计算误差在10%以内,该方法实现了更高效的PMVS三维重建。基于MGPS的PMVS算法还可用于文物保护、医学图像处理、虚拟现实等领域。     

3D Object Recognition in Cluttered Environments by Segment-Based Stereo Vision

We propose a new method for 3D object recognition which uses segment-based stereo vision. An object is identified in a cluttered environment and its position and orientation (6 dof) are determined accurately enabling a robot to pick up the object and manipulate it. The object can be of any shape (planar figures, polyhedra, free-form objects) and partially occluded by other objects. Segment-based stereo vision is employed for 3D sensing. Both CAD-based and sensor-based object modeling subsystems are available. Matching is performed by calculating candidates for the object position and orientation using local features, verifying each candidate, and improving the accuracy of the position and orientation by an iteration method. Several experimental results are presented to demonstrate the usefulness of the proposed method.     

融合多视图注意力机制的城市交通流量补全

交通流量信息是智能交通系统和城市计算的重要基础.交通流量数据作为新型时序数据,由于数据的采集方式和外部复杂因素的影响,使得数据缺失现象是常见且无法避免的.如何有效地挖掘交通流量数据的时空特性和数据间的关联成为了提高缺失数据补全精度的关键.传统的统计学方法不能满足日益增长的数据需求,深度学习的应用推动了缺失数据的补全方法...     

动态环境下基于旋转-平移解耦的立体视觉里程计算法

在实际应用中,若图像中的动态特征数量多且运动方向一致,这些特征会对视觉里程计的估计结果产生严重的影响.本文针对这类问题提出一种根据图像特征点位置解耦估计摄像机旋转-平移的立体视觉里程计算法.算法通过立体视觉系统将特征点划分成"远点"和"近点".在随机抽样一致性算法(RANSAC)框架下,采用"远点"估计视觉系统的姿态;进而在姿态已知的条件下,通过"近点"估计摄像机平移,实现旋转-平移解耦计算.这样处理可以通过姿态约束减少近距离运动物体对视觉里程计的影响.实验表明,在实际道路环境中,本文基于旋转-平移解耦估计的算法较之传统的同时估计旋转-平移的算法,能有效剔除动态特征.所提出算法对动态特征的抗干扰能力更好,鲁棒性更强,精度更高.     

UAV Path Following in Windy Urban Environments

This paper considers UAV path following in cluttered environments under windy conditions. Unstructured wind patterns in cluttered environments can make path following difficult resulting in high errors and possibly collisions with buildings. Combining a pursuit guidance law philosophy with a line-of-sight guidance law, we develop a novel guidance law that has low computational complexity and can track straight line paths, circular paths, a combination of both and waypaths accurately in the presence of wind blowing as high as fifty percent of the UAV’s air speed. Performance of the guidance law is demonstrated through numerical simulations.     

Robot Localization in Indoor Environments Using Radio Frequency Identification Technology and Stereo Vision

In this paper, a novel method of obstacle recognition and localization for mobile robots using radio frequency identification (RFID) technology and stereo vision is proposed. It is inexpensive, flexible and easy to use in practical environments. As information about the obstacles or environment can be written in ID tags, the proposed method can detect the obstacles easily and quickly compared with other methods. RF is not so stable, so the Bayes rule was introduced to calculate the probability where the ID tag exists in order to improve the accuracy of localizing obstacles with the ID tags. Then the stereo camera starts to process the image of the region of interest where the obstacle exists. The proposed method does not need to process all the image and easily gets some information about obstacles such as size, and color, and thus decreases the processing computation. Research on RFID technology integrating stereo vision to localize an indoor mobile robot has also been performed. This paper introduces the architecture of the proposed method and some experimental results.     

A Vision for Supporting Autonomous Navigation in Urban Environments

The autonomous navigation systems (ANS), such as autonomous ground vehicles (AGVs), unmanned aerial vehicles (UAVs), and unmanned submersible vehicles (USVs), and modern vehicles with actuators, sensors, and computer control perform three basic functions: context gathering using sensors, processing, and action. Most researchers have put all three functions into the ANS or the robot itself to overcome occlusions and handle the environment's dynamics. However, this causes the ANS and robotic systems to be bulky and expensive. It also impedes the introduction of vehicles with ANS in urban environments, where they must coexist with existing cars and highways. The approach presented distributes the context-gathering and processing functions using sensor networks and wireless communications technologies to reduce costs and make ANS widespread. The system uses sensors mounted on moving vehicles and stationary objects such as lampposts, traffic lights, toll plazas, and buildings to gather information at different levels     

一种安全工作于城区环境的消防侦察机器人

介绍了一种应用于城区环境的消防侦察机器人．它采用具有关节链轮的差动轮式结构,适于爬楼梯和斜坡．采用实时泄漏补偿方式维持车体内具有恒定正压,确保机器人的防爆安全性能．在失去无线控制信号时,机器人根据侦察过程中记录的传感器数据,按原路径返回,并自动回撤到安全区域．由于使用单一的编码器获得的位置信息具有累积误差,本文采用新颖的编码器与超声波传感器数据集成的方法,并使用迭代最近点算法进行数据匹配,调整机器人的位置和方向角,获得了较高的自动回撤精度．     

多视角融合的时空动态图卷积网络城市交通流量预测

城市交通流量预测是构建绿色低碳、安全高效的智能交通系统的重要组成部分.时空图神经网络由于具有强大的时空数据表征能力,被广泛应用于城市交通流量预测.当前时空图神经网络在城市交通流量预测中仍存在以下两方面局限性:1)直接构建静态路网拓扑图对城市空间相关性进行表示,忽略了节点的动态交通模式,难以表达节点流量之间的时序相似性,无法捕获路网节点之间在时序上的动态关联.2)只考虑路网节点的局部空间相关性,忽略节点的全局空间相关性,无法建模交通路网中局部区域和全局空间之间的依赖关系.为打破上述局限性,本文提出了一种多视角融合的时空动态图卷积模型用于预测交通流量.首先,从静态空间拓扑和动态流量模式视角出发,构建路网空间结构图和动态流量关联图,并使用动态图卷积学习节点在两种视角下的特征,全面捕获城市路网中多元的空间相关性.其次,从局部视角和全局视角出发,计算路网的全局表示,将全局特征与局部特征融合,增强路网节点特征的表现力,发掘城市交通流量的整体结构特征.接下来,设计了局部卷积多头自注意力机制来获取交通数据的动态时间相关性,实现在多种时间窗口下的准确流量预测.最后,在四种真实交通数据上的实验结果证明了本文模型的有效性和准确性.     

Occluder Shadows for Fast Walkthroughs of Urban Environments

This paper describes a new algorithm that employs image-based rendering for fast occlusion culling in complex urban environments. It exploits graphics hardware to render and automatically combine a relatively large set of occluders. The algorithm is fast to calculate and therefore also useful for scenes of moderate complexity and walkthroughs with over 20 frames per second. Occlusion is calculated dynamically and does not rely on any visibility precalculation or occluder preselection. Speed-ups of one order of magnitude can be obtained.     

具有纹理感知能力的超像素分割方法

目的 超像素分割是计算机视觉领域常用的一项预处理技术,目标是将相邻像素聚集成为具有一定语义的子区域,能够大幅度降低后续处理的计算复杂度,但是对包含强梯度纹理的图像分割效果不佳,为此提出一种具有纹理感知能力的超像素分割方法。方法 提出一种能够区分强梯度噪声和纹理像素的颜色距离,其中利用带方向的1/4圆形窗口均值滤波后的颜色信息,提升包含强梯度噪声和纹理图像的超像素分割性能。利用区间梯度幅值与Sobel梯度幅值相乘得到混合梯度幅值,具有纹理抑制、结构保持以及边缘线条细的优点,能够提升超像素的贴合边缘性能,增强超像素形状规则程度。最后,利用混合梯度的幅值计算具有结构回避能力的综合聚类距离,进一步防止超像素跨越物体的边界,增强超像素的贴边性能。结果 在BSDS500（Berkeley segmentation dataset 500）图像数据集和强纹理马赛克图像等不同类型图像上的测试结果显示,与目前主流的超像素分割方法相比,本文算法在UE （undersegmentation error）、ASA （achievable segmentation accuracy）和CM （compactness measure）等性能指标上分别提高了1.5%、0.2%和4.3%。从视觉效果上看,能够在排除纹理干扰的情况下生成结构边缘贴合程度更好的形状规则超像素。结论 本文算法在包含强梯度纹理图像上的超像素分割性能优于对比方法,在目标识别、目标追踪和显著性检测等易受强梯度干扰的技术领域具有较大应用潜力。     

大数据环境下的城市计算及其应用

介绍城市计算的技术内涵、数学模型及研究方法。包括数据流、图数据模型和排队论以及传感技术、城市数据管理、城市环境异构数据可视化等技术。并指出当前国内外的研究热点课题以及城市计算的应用和面临的挑战。     

Expeditious Modelling of Virtual Urban Environments with Geospatial L-systems

L-systems have been used in Computer Graphics, namely for modelling plants, as well as in a few experiments to model urban environments. However, the lack of geospatial awareness is a limitation and in spite of some developments like open L-systems, that introduced the ability to communicate with the environment, there was a need for more flexibility. This paper presents Geospatial L-systems, a new extension of L-systems that incorporates geospatial awareness, and shows an application in the area of expeditious modelling of urban environments. A modelling system, named XL3D, generates virtual urban environments automatically from a XML based document that contains a modelling specification and accesses data sources in an interoperable way. The integration of geospatial L-systems in this modelling system has increased the potential for automation and the potential to generate virtual urban environments with a higher level of detail and visual fidelity, with a lower level of complexity of the modelling processes. These facts are shown in a case study where a virtual urban environment, taken from an area in the Porto downtown, is generated by this solution.