Reconstruction of 3D scenes using stereoimages without rectification

The problem of reconstructing 3D scenes using stereopairs of images is considered. In contrast to traditional technologies whereby the rectification and disparity map construction are implemented, the corresponding points are determined directly on the epipolar lines that belong to the given epipolar plane. A complete three-dimensional model of an object is constructed by scanning a 3D scene using pencils of planes. The elimination of fragment-matching errors occurs by using the decimation filtering of the found points. The results of experiments that show the efficiency of the proposed technology are provided.     

Visual SLAM Based on Rigid-Body 3D Landmarks

In current visual SLAM methods, point-like landmarks (As in Filliat and Meyer (Cogn Syst Res 4(4):243–282, 2003), we use this expression to denote a landmark generated by a point or an object considered as punctual.) are used for representation on maps. As the observation of each point-like landmark gives only angular information about a bearing camera, a covariance matrix between point-like landmarks must be estimated in order to converge with a global scale estimation. However, as the computational complexity of covariance matrices scales in a quadratic way with the number of landmarks, the maximum number of landmarks that is possible to use is normally limited to a few hundred. In this paper, a visual SLAM system based on the use of what are called rigid-body 3D landmarks is proposed. A rigid-body 3D landmark represents the 6D pose of a rigid body in space (position and orientation), and its observation gives full-pose information about a bearing camera. Each rigid-body 3D landmark is created from a set of N point-like landmarks by collapsing 3N state components into seven state components plus a set of parameters that describe the shape of the landmark. Rigid-body 3D landmarks are represented and estimated using so-called point-quaternions, which are introduced here. By using rigid-body 3D landmarks, the computational time of an EKF-SLAM system can be reduced up to 5.5%, as the number of landmarks increases. The proposed visual SLAM system is validated in simulated and real video sequences (outdoor). The proposed methodology can be extended to any SLAM system based on the use of point-like landmarks, including those generated by laser measurement.     

Visual 3-D SLAM from UAVs

The aim of the paper is to present, test and discuss the implementation of Visual SLAM techniques to images taken from Unmanned Aerial Vehicles (UAVs) outdoors, in partially structured environments. Every issue of the whole process is discussed in order to obtain more accurate localization and mapping from UAVs flights. Firstly, the issues related to the visual features of objects in the scene, their distance to the UAV, and the related image acquisition system and their calibration are evaluated for improving the whole process. Other important, considered issues are related to the image processing techniques, such as interest point detection, the matching procedure and the scaling factor. The whole system has been tested using the COLIBRI mini UAV in partially structured environments. The results that have been obtained for localization, tested against the GPS information of the flights, show that Visual SLAM delivers reliable localization and mapping that makes it suitable for some outdoors applications when flying UAVs.     

基于3D激光雷达的SLAM算法研究现状与发展趋势

即时定位与地图构建(simultaneous localization and mapping, SLAM)算法是移动机器人实现自主移动的关键环节。激光雷达(LiDAR)具有测距精度高、不易受外部干扰和地图构建直观方便等优点,广泛应用于大型复杂室内外场景地图的构建。随着3D激光器的应用与普及,国内外学者围绕基于3D激光雷达的SLAM算法的研究已取得丰硕的成果。梳理了3D激光SLAM算法在前端数据关联、后端优化等环节的国内外研究现状,分析总结了目前各种3D激光SLAM算法以及改进方案的原理和优缺点,阐述了深度学习和多传感器融合理论与技术在3D激光SLAM算法中的应用情况,指出多源信息融合、与深度学习结合、应用场景的鲁棒性、 SLAM算法通用框架及移动传感器和无线信号体制的技术渗透是3D激光SLAM算法的研究热点和发展趋势。研究成果对3D激光SLAM算法和未知环境中移动机器人即时定位和地图构建的研究具有重要的参考价值和指导意义。     

基于语义概率预测的动态场景单目视觉SLAM

目的 基于视觉的同步定位与建图（visual-based simultaneous localization and mapping,vSLAM）是计算机视觉以及机器人领域中的关键技术,其通过对输入的图像进行处理分析来感知周围的3维环境以及进行自身的定位。现有的SLAM系统大多依赖静态世界假设,在真实环境中的动态物体会严重影响视觉SLAM系统的稳定运行。同时,场景中静止与运动部分往往和其语义有密切关系,因而可以借助场景中的语义信息来提升视觉SLAM系统在动态环境下的稳定性。为此,提出一种新的基于语义概率预测的面向动态场景的单目视觉SLAM算法。方法 结合语义分割的结果以及鲁棒性估计算法,通过对分割进行数据关联、状态检测,从概率的角度来表示观测的静止/运动状态,剔除动态物体上的观测对相机位姿估计的干扰,同时借助运动概率及时剔除失效的地图点,使系统在复杂动态的场景中依然能够稳定运行。结果 在本文构建的复杂动态场景数据集上,提出的方法在跟踪精度和完整度上都显著优于现有的单目视觉SLAM方法,而且在TUM-RGBD数据集中的多个高动态序列上也取得了更好的结果。此外,本文定性比较了动态场景下的建图质量以及AR（augmented reality）效果。结果表明,本文方法明显优于对比方法。结论 本文通过结合语义分割信息以及鲁棒性估计算法,对分割区域进行数据关联以及运动状态检测,以概率的形式表示2D观测的运动状态,同时及时剔除失效地图点,使相机位姿估计的精度以及建图质量有了明显提升,有效提高了单目视觉SLAM在高度动态环境中运行的鲁棒性。     

A practical 2D/3D SLAM using directional patterns of an indoor structure

This paper presents a practical two-dimensional (2D)/three-dimensional (3D) simultaneous localization and mapping (SLAM) algorithm using directional features for ordinary indoor environments; this algorithm is adaptable to various conditions, computationally inexpensive, and accurate enough to use for practical applications. The proposed algorithm uses odometry acquired from other sensors or other algorithms as the initial estimate and the directional features of indoor structures as landmarks. The directional features can only correct the rotation error of the odometry. However, we show that the greater part of the translation error of the odometry can also be corrected when the directional features are detected at almost positions accurately. In that case, there is no need to use other kinds of features to correct translation error. The directions of indoor structures have two advantages as landmarks. First, the extraction of them is not affected by obstacles. Second, the number of them is small regardless of the size of the building. Because of these advantages, the proposed SLAM algorithm shows robustness for parameters and lightweight properties. From extensive experiments with 2D/3D datasets taken from different buildings, we show the practicality of the proposed algorithm. We also demonstrate that the 2D algorithm runs in real time on a low-end smartphone.     

基于半直接法SLAM的大场景稠密三维重建系统

当前三维重建系统大多基于特征点法和直接法的同时定位与地图重建(SLAM)系统,特征点法SLAM难以在特征点缺失的地方具有较好的重建结果,直接法SLAM在相机运动过快时难以进行位姿估计,从而造成重建效果不理想.针对上述问题,文中提出基于半直接法SLAM的大场景稠密三维重建系统.通过深度相机(RGB-D相机)扫描,在特征点丰富的区域使用特征点法进行相机位姿估计,在特征点缺失区域使用直接法进行位姿估计,减小光度误差,优化相机位姿.然后使用优化后较准确的相机位姿进行地图构建,采用面元模型,应用构建变形图的方法进行点云的位姿估计和融合,最终获得较理想的三维重建模型.实验表明,文中系统可适用于各个场合的三维重建,得到较理想的三维重建模型.     

基于2D激光雷达的SLAM算法研究综述

移动机器人导航功能的实现需要同时定位与建图(simultaneous localization and mapping,SLAM)和路径规划这两方面的技术,其中由SLAM技术生成的栅格地图是移动机器人运用路径规划算法的前提。2D激光SLAM由于其建图精度较高、性能稳定且价格便宜,在室内移动机器人中应用十分广泛。2D激光SLAM是指移动机器人在自身所处环境及位置先验信息未知的情况下,以2D激光雷达为主要传感器,感知周围环境信息,从而实现自身位姿的估计和地图的构建。将2D激光SLAM分为两部分,第一部分从激光测距原理入手,对三角法和飞行时间法进行了详细介绍和优缺点比较。第二部分从前端扫描匹配、后端优化、回环检测和地图构建这四个方面分别详细阐述了2D激光SLAM系统框架。同时对主流2D激光SLAM算法进行了深入分析和优缺点比较,并对激光SLAM未来的发展进行了展望。     

Information-Efficient 3-D Visual SLAM for Unstructured Domains

This paper presents a novel vision-based sensory package and an information-efficient simultaneous localization and mapping (SLAM) algorithm. Together, we offer a solution for building 3-D dense map in an unknown and unstructured environment with minimal computational costs. The sensory package we adopt consists of a conventional camera and a range imager, which provide range and bearing and elevation inputs as commonly used by 3-D feature-based SLAM. In addition, we propose an algorithm to give the robots the `intelligence” to select, out of the steadily collected data, the maximally informative observations to be used in the estimation process. We show that, although the actual evaluation of information gain for each frame introduces an additional computational cost, the overall efficiency is significantly increased by keeping the matrix compact. The noticeable advantage of this strategy is that the continuously gathered data are not heuristically segmented prior to being input to the filter. Quite the opposite, the scheme lends itself to be statistically optimal and is capable of handling large datasets collected at realistic sampling rates.      

二维概念格图形的三维重构研究

介绍了二维概念格图形向三维空间转化和延伸的必要性和现状。通过对传统概念格图形分层定位布局方法的研究与分析,提出并实现了一种新的以具有大量的平行四边形和有向线段为基本特征的概念格在三维空间的自动布局算法,描述了一种基于该算法的二维概念格图形的三维重构机制,有效地解决了节点横向过度扩张的问题并减少了线段交叉,较好地实现了复杂概念格图形的三维可视化,为知识发现和知识处理提供了良好的基础。     

二维概念格图形的三维重构研究

介绍了二维概念格图形向三维空问转化和延伸的必要性和现状.通过对传统概念格图形分层定位布局方法的研究与分析,提出并实现了一种新的以具有大量的平行四边形和有向线段为基本特征的概念格在三维空间的自动布局算法,描述了一种基于该算法的二维概念格图形的三维重构机制,有效地解决了节点横向过度扩张的问题并减少了线段交叉,较好地实现了复杂概念格图形的三维可视化,为知识发现和知识处理提供了良好的基础.     

稀疏视觉SLAM对平面激光雷达传感的稠密化模拟

为了解决稀疏特征点VSLAM （visual simultaneous localization and mapping）由于其构图稀疏性,在视觉导航应用方面的短板,提出一种高斯滤波插值的方法对其特征点进行稠密化处理,实现对平面激光雷达反馈的模拟.本文利用高斯分布以及迭代滤波实现数据的稠密化,通过建立全局高斯滤波以及局部高斯分布估计,实现对稀疏VSLAM空间点平面投影的数据插补,进而实现对平面激光雷达数据的模拟.仅使用CPU情况下,算法每帧耗时为0.0003s~0.006s,插值结果相对误差为7.956%.实验证明,该插值方法成功实现了稀疏投影点的稠密化,插补结果与真实激光雷达反馈相似度高,为视觉导航提供了一种有效的前端传感处理方法.     

低成本移动机器人2D SLAM算法地图评估研究

随着传感器技术的发展,移动机器人得到了广泛的应用.用于绘制环境地图的2D SLAM技术是移动机器人领域重要的技术之一,选出适合低成本机器人的算法对机器人的发展和普及至关重要.提出在仿真环境和实际场景下运行广泛使用的四种开源算法:Gmapping、Hector SLAM、Karto SLAM和Google Cartogr...     

Optimizing 3D Irregular Object Packing from 3D Scans Using Metaheuristics

For efficient construction-assemblies transportation, volume constrained 3D printing, dry stacking, and facility waste management, a common problem must be solved. It is the practical problem of packing irregular 3D rigid objects into a container with fixed dimensions so that the volume of the final packed objects is minimized. To solve this problem, a methodology is presented that begins with capturing the initial as-is 3D shape data for each object, followed by a metaheuristic-based packing optimization algorithm. This methodology is demonstrated to be effective in two situations with known optimum solutions and in a third situation involving packing of real-life as-is objects. A high-level selection algorithm that is designed to guide the search of possible object subsets, when not all objects can fit into a single predefined container, is discussed as well. Performance is examined for variations, and a preliminary sensitivity analysis is performed. The methodology and its key algorithms are demonstrated to produce effective packing solutions in a mostly automatic manner. Object packing for this class of applications in civil engineering can thus be potentially improved in terms of outcome efficiency and level of planning effort required.     

Extending distortion viewing from 2D to 3D

Addresses the visual exploration of 3D information layouts. Several visual exploration techniques have been proposed for 2D information layouts. Many of these try to take advantage of humans' natural visual pattern-recognition abilities to understand global relationships while simultaneously integrating this knowledge with local details. This desire for detail-in-context views (also called fisheye, multiscale and distortion views) has fueled considerable research in the development of distortion viewing tools. Generally, these tools provide space for magnification of local detail by compressing the rest of the image. In considering a possible detail-in-context view for 3D layouts, we first examine 2D distortion techniques, bearing in mind the particular 3D problem of occlusion. Comparing 2D and 3D information layout adjustment tools leads directly to a 3D visual access tool that clears a line of sight to any region of interest. While our technique can extend to any type of 3D information display, we focus on graphs     

Transformations between Pictures from 2D to 3D

This paper describes programs for 3-dimensional engraving. The programs use raster or vector images to create a 3D model and, subsequently, convert this model into a sequence of control commands for 3D engraving machines. Three programs have been developed. A program for engraving general 3D surfaces from grey-scale images, a program for preparing these grey-scale images from patterns and vector images, and a program for fast 2D engraving. A simple and fast preparation of the 3D model, a user-friendly environment, and small hardware requirements were the principal goals.     

3D foot shape generation from 2D information

Two methods to generate an individual 3D foot shape from 2D information are proposed. A standard foot shape was first generated and then scaled based on known 2D information. In the first method, the foot outline and the foot height were used, and in the second, the foot outline and the foot profile were used. The models were developed using 40 participants and then validated using a different set of 40 participants. Results show that each individual foot shape can be predicted within a mean absolute error of 1.36 mm for the left foot and 1.37 mm for the right foot using the first method, and within a mean absolute error of 1.02 mm for the left foot and 1.02 mm for the right foot using the second method. The second method shows somewhat improved accuracy even though it requires two images. Both the methods are relatively cheaper than using a scanner to determine the 3D foot shape for custom footwear design.     

Bayesian 3D shape from silhouettes

This paper introduces a smooth posterior density function for inferring shapes from silhouettes. Both the likelihood and the prior are modelled using kernel density functions and optimisation is performed using gradient ascent algorithms. Adding a prior allows for the recovery of concave areas of the shape that are usually lost when estimating the visual hull. This framework is also extended to use colour information when it is available in addition to the silhouettes. In these cases, the modelling not only allows for the shape to be recovered but also its colour information. Our new algorithms are assessed by reconstructing 2D shapes from 1D silhouettes and 3D faces from 2D silhouettes. Experimental results show that using the prior can assist in reconstructing concave areas and also illustrate the benefits of using colour information even when only small numbers of silhouettes are available.