3D Modeling of Subterranean Environments by Robotic Survey翻译文

This paper presents a method for autonomously collecting survey-quality data from underground spaces for the purposes of 3D modeling. This overview includes a systems-level description of the robotic tools, data acquisition process and data processing techniques that enable accurate 3D modeling of subterranean interiors. A performance evaluation and accuracy analysis between the capabilities of this robotic system and human surveying is also provided.
本文提出了一种从地下空间自动采集测量质量数据的方法，用于三维建模。这个概述包括系统层面的描述，机器人工具，数据采集过程和数据处理技术，使精确的三维建模地下室内。并对机器人系统的性能和精度进行了分析。

subterranean[,sʌbtə'reɪnɪən]  	adj. 地下的；秘密的；隐蔽的 	n. 地下工作者		

1.Introduction
Subterranean applications offer exceptional opportunities for robots. Hazardous, remote and space constrained, underground spaces such as mines, tunnels, caves and sewers are difficult environments for people to reach and work; yet, information acquired from the subterranean has immense civil and commercial value [1]. Compact, sensory-tailored robotic systems provide practical solutions to subterranean information-gathering efforts by reaching spaces and collecting data on a scale that was once not feasible.

地下应用程序为机器人提供了特殊的机会。危险、偏远和空间受限的地下空间，如矿井、隧道、洞穴和下水
道，是人们难以到达和工作的环境;然而，从地下获取的信息具有巨大的民用和商业价值。紧凑，传感器定
制的机器人系统提供了实用的解决方案，通过到达空间和收集曾经不可行的规模的数据来收集地下信息。
   
Hazardous['hæzədəs] adj. 有危险的；冒险的；碰运气的

One of the significant challenges facing robotic information-gathering in underground applications is the availability of accurate position and pose information in the absence of GPS. This paper explores the use of robotic surveying as a method for acquiring high-fidelity position data such that a robot’s sensor footprint can be accurately registered to a global coordinate frame. In this context, LIDAR scans, photographs, thermal images, gas concentrations, moisture levels and countless other sensing modalities can be spatially co-registered to provide an unrivaled virtual view of subterranean voids. The underlying processes for accruing and fusing subterranean information into a consistent and accurate model, however, are labor-intensive tasks; hence this work seeks to construct a fully automated system for data acquisition, model construction and reporting.

在地下应用中，机器人信息收集面临的一个重大挑战是，在没有GPS的情况下，能否获得准确的位置和姿态
信息。本文探讨了机器人测量法作为获取高保真度位置数据的一种方法，使机器人的传感器足迹能够准确地
记录到全局坐标系中。在这种情况下，激光雷达扫描、照片、热图像、气体浓度、湿度水平和无数其他传感
方式可以在空间上共同注册，以提供无与伦比的地下空洞虚拟视图。然而，积累和融合地下信息到一致和准
确的模型中的基本过程是劳动密集型任务;因此，这项工作试图建立一个完全自动化的系统，用于数据采
集、模型构建和报告。

This paper presents a systems-level overview of the subterranean modeling process starting from data collection and concluding with model presentation. The structure of this paper is as follows: Section 2 describes a robotic survey team that cooperatively gathers the necessary sensor data from a subterranean space. Section 3 details the techniques employed to reconstruct the void’s interior from logged robot data. Section 4 showcases
a collection of results obtained by this procedure and Section 5 concludes with a forecast of research and enterprise for this technology.

本文从系统层面概述了地下建模过程，从数据采集开始，最后以模型表示结束。本文的结构如下:第2节描述
了一个机器人测量队合作从地下空间收集必要的传感器数据。第3节详细介绍了通过记录机器人数据重建空
间内部的技术。第4节展示了本程序所获得的结果的集合，第5节最后对该技术的研究和企业进行了预测。

conclude with 以…为结论
interior [ɪn’tɪɜːrɪə] n. 内部；本质 adj. 内部的；国内的；本质的
logged 把…输入计算机

2
Data Collection and Robotic Survey Robotic equipment is revolutionizing geological data
collection in work environments such mines and quarries by reducing the time,managing the repetitive nature and minimizing the error associated with measuring natural surfaces over vast ranges. Laser profiling systems, for example, integrate video with LIDAR images, which record of hundreds of thousands of individual range measurements, to create high-resolution surface models for detailed burden analysis in quarries and surface mines [2].

数据收集和机器人调查机器人设备正在通过缩短时间、管理重复性质和最大限度地减少测量自然表面的误	
差，使矿山和采石场等工作环境中的地质数据收集发生革命性的变化。例如，激光剖面系统将视频和激光雷
达图像结合在一起，后者记录了数十万个单独的距离测量数据，从而创建高分辨率的地表模型，以便在采石
场和地表矿井[2]中进行详细的负荷分析。

quarry ['kwɒrɪ]

n. 采石场；猎物；来源
vi. 费力地找
vt. 挖出；努力挖掘
n. (Quarry)人名；(英)夸里

Robotic total stations and GPS are enabling rapid acquisition of highly accurate positioning data to spatially geo-reference natural and man-made geologic features [3]. Together, these and other related surface-scanning technologies are saturating the world with geospatial data on and over the Earth’s surface.

机器人全站仪和全球定位系统能够快速获取高度精确的定位数据，以便对自然和人为地质特征进行空间地理
参考。总之，这些和其他相关的地表扫描技术使地球表面和地面上的地理空间数据饱和。

acquisition

英 [,ækwɪ'zɪʃ(ə)n] 美 [,ækwɪ'zɪʃən]
n. 获得物，获得；收购


geospatial

[,dʒiːəʊ'speɪʃ(ə)l]
n. 地理空间

Subterranean environments, characterized by confined spaces,harsh conditions, limited accessibility and the absence of GPS, prohibit direct utilization of modern surface mapping methods [1]. To address these problems, the field of subterranean robotics has produced a number of platforms, sensors and algorithms to reach, sense and model a variety of underground spaces [4,5,Error! Reference source not found.].

地下环境以密闭空间、恶劣条件、有限的可达性和缺乏GPS为特征，禁止直接使用现代地表测绘方法[1]。为
了解决这些问题，地下机器人领域已经开发出了许多平台、传感器和算法，以达到、感知和建模各种地下空间。

The most recent technique to emerge from subterranean robot development is a method for robotic surveying, which provides a direct mechanism to maintain a robot’s global position while accruing underground data.

地下机器人发展中出现的最新技术是一种机器人测量方法，它提供了一种直接机制，在积累地下数据的同时
保持机器人的全球位置。

Similar to a method employed to accurately position robotic mining machinery [7], robotic surveying utilizes a total station (e.g.an actuated and highly accurate range-measuring laser) to track the position of a mobile platform equipped with a special reflective prism (Fig. 1).

与用于精确定位机器人采矿机械[7]的方法类似，机器人测量利用全站仪(例如，一个驱动的、高度精确的距	
离测量激光器)来跟踪装有特殊反射棱镜的移动平台的位置(图1)。

utilize
英 [ˈjuːtɪˌlaɪz] 美 ['juːtəlaɪz]
vt. 利用

As the robot moves and scans a subterranean space, the survey instrument monitors the robot’s position by following the robot’s onboard prism.

当机器人移动并扫描地下空间时，测量仪器通过机器人的机载棱镜监视机器人的位置。

onboard prism 机载棱镜

prism
英 ['prɪz(ə)m] 美 ['prɪzəm]
n. 棱镜；[晶体][数] 棱柱

Pose information (i.e. the orientation of the robot) may also be obtained by either outfitting the robot with multiple prisms or mounting one prism to an actuated platform located on the robot.

 姿态信息(即机器人的方向)也可以通过给机器人配备多个棱镜或将一个棱镜安装到机器人
 上的驱动平台来获得。

The computed differential between relative prism locations on the robot provides an estimate of body orientation that can align each of the robot’s sensors to a particular vantage. The accuracy of this pose estimate can be further enhanced through the incorporation of inertial sensing.
通过计算出机器人上相对棱镜位置之间的差异，可以估算出机器人的身体方向，从而使机器人的每个传感器对准一个特定的有利位置。这种姿态估计的精度可以通过加入惯性传感进一步提高。
differential
英 [,dɪfə’renʃ(ə)l] 美 ['dɪfə’rɛnʃəl]

adj. 微分的；差别的；特异的
n. 微分；差别

align
英 [ə’laɪn] 美 [ə’laɪn]

vt. 使结盟；使成一行；匹配
vi. 排列；排成一行

vantage
英 ['vɑːntɪdʒ] 美 ['væntɪdʒ]

n. 优势；有利情况

inertial sensor

	惯性传感器 

incorporation
英 [ɪn,kɔːpə’reɪʃən] 美 [ɪn,kɔrpə’reʃən]

n. 公司；合并，编入；团体组织

inertial
英 [ɪ’nɜːʃ(ə)l] 美 [ɪ’nɝʃəl]

adj. 惯性的；不活泼的

Data collection is a completely automated procedure for the reasons previously described: efficiency, repeatability and accuracy. The robot seen in Fig. 1, know as Cave Crawler, is designed to use an assortment of sensors that are well-suited for subterranean model construction such as 3D LIDAR, low-light video, a high-resolution still camera, radar, a thermal camera and inertial.

数据收集是一个完全自动化的过程，原因如前所述:效率、可重复性和准确性。如图1所示的机器人被称为洞
穴爬虫，它被设计用来使用各种各样的传感器，这些传感器非常适合于地下模型的构建，例如3D激光雷
达、低光视频、高分辨率静止相机、雷达、热相机和惯性。

To obtain a sensor coverage capable of reconstructing subterranean surfaces to within centimeters of accuracy, the robot moves,stops and thoroughly scans the subterranean interior in continuous repetition. During each scan, the robot moves its survey prism to four locations and queries the survey instrument to measure each prism position.

为了获得一个能以厘米的精度重建地下表面的覆盖范围，机器人要不断的移动，停止和彻
底扫描地下。在每次扫描过程中，机器人将测量棱镜移动到四个位置，并询问测量仪器来
测量每个棱镜的位置。

interior
英 [ɪn’tɪɜːrɪə] 美 [ɪn’tɪrɪɚ]

n. 内部；本质
adj. 内部的；国内的；本质的

After obtaining these measurements, the robot resets the prism position to a
specified start position and moves to the next scan location. This cyclic operation transpires in clockwork fashion until the robot has completed its survey, is halted by error or is stopped by human intervention.

获得这些测量数据后，机器人将棱镜的位置重置到指定的起始位置，并移动到下一个扫描位置。这个循环操
作以发条的方式进行，直到机器人完成了它的测量，或者由于错误而停止，或者由于人为干预而停止。

Navigation and scan position is robot controlled, but human operators can dynamically adjust trajectory characteristics to suit the environment. The navigation scheme employed on Cave Crawler is a behavior-based system where simple trajectory routines are managed by higher-order control logic. Akin to known mine navigation methods of [8,9], low-level steering behaviors are reactive; however, Cave Crawler extends these prior 2D schemes to 3D to take advantage of its 3D LIDAR 1 .

导航和扫描位置是机器人控制的，但人工操作人员可以根据环境动态调整轨迹特征。洞穴爬虫导航方案是一
种基于行为的系统，简单的轨迹轨迹由高阶控制逻辑进行管理。与已知的[8,9]矿井导航方法类似，低水平转
向行为是被动的;然而，洞穴爬虫扩展这些先前的二维方案到三维利用其3D激光雷达1。



trajectory
英 [trə'dʒekt(ə)rɪ; 'trædʒɪkt(ə)rɪ] 美 [trə'dʒɛktəri]
n. [物] 轨道，轨线；[航][军] 弹道

reactive
英 [rɪ’æktɪv] 美 [rɪ’æktɪv]

adj. 反应的；电抗的；反动的

At moderate speeds (human walking pace) this simple reactive 3D navigation scheme has shown to be empirically robust.Higher-order navigation is based upon a topological
command structure [4].

在中等速度(人的步行速度)下，这个简单的无反应3D导航方案已经被证明是经验上可靠的。高阶导航基于拓
扑命令结构[4]。

The robot is provided with pairings of <distance, action>commands prior to the start of the survey. A typical command list, for example, would have the robot drive straight for 90 meters, turn left, continue for another 30 meters, turn right, and etc. Scanning locations are therefore defined as waypoints (i.e. specific increments) along this path.

机器人具有命令的配对，在调查开始前。例如，一个典型的命令列表会让机器人沿着90米直线行驶，左转，
继续行驶30米，右转，等等。因此，扫描位置被定义为沿着这条路径的路径点(即特定增量)。

In addition, both the topological commands and waypoint increments can be dynamically altered during the robot’s traverse.At present, human interaction is required to place and initialize the survey instrument at the start of each operation and to continue the operation once the robot and survey instrument lose line-of-sight contact.

此外，拓扑命令和路径点增量都可以在机器人的遍历过程中动态改变。目前，在每次操作开始时都需要进行
人工交互，将测量仪器放置并初始化，一旦机器人与测量仪器失去视线接触，则需要继续操作。

increment
英 ['ɪŋkrɪm(ə)nt] 美 ['ɪŋkrəmənt]

n. [数] 增量；增加；增额；盈余

traverse
英 ['trævəs; trə’vɜːs] 美 [trə’vɝs]

n. 穿过；横贯；横木
vt. 穿过；反对；详细研究；在…来回移动
vi. 横越；旋转；来回移动
adj. 横贯的
n. (Traverse)人名；(英)特拉弗斯；(法)特拉韦尔斯

For this reason, this system is restricted to surveying areas where humans have access to manually place the survey instrument. Section 5 discusses future plans to remove this restriction by robotically mobilizing the survey instrument.

由于这个原因，这个系统被限制在人们可以手动放置测量仪器的区域。第5节讨论了通过机器人移动测量仪
器来消除这一限制的未来计划。

mobilizing

动用,动员
使...松动
使流通

3
Automated Model Construction
Just as data collection requires automation for reasons of efficiency and accuracy, data processing is tedious and error-prone without the proper autonomous tools to manage the volumes of data logged from robot sensors.

正如由于效率和准确性的原因，数据收集需要自动化一样，如果没有合适的自动化工具来管理机器人传感
器记录的数据量，数据处理就会变得单调乏味且容易出错。

tedious
英 ['tiːdɪəs] 美 ['tidɪəs]

adj. 沉闷的；冗长乏味的

Model synthesis requires that all data be synchronized, fused and spatially registered into a global coordinate system. As such, data synchronization begins online while the robot is collecting data.

模型合成要求所有数据同步、融合并在空间上登记到一个全局坐标系中。因此，当机器人在收集数据时，数
据同步就会在线开始。

synthesis
英 ['sɪnθɪsɪs] 美 ['sɪnθəsɪs]

n. 综合，[化学] 合成；综合体

synchronization
英 [,sɪŋkrənaɪ’zeɪʃən] 美 [,sɪnkrənaɪ’zeʃən]

n. [物] 同步；同时性

The Cave Crawler system is constructed with a modular software architecture that specifically monitors and regulates data flow [12]. Every sensor and input device is time stamped and logged.

洞穴爬虫系统是用一个专门监控和调节数据流[12]的模块化软件架构构建的。每个传感器和输入设备都有时
间戳和记录。

Handling data in this manner allows for complete offline playback of the data and provides a timeline to organize streaming data into discrete model-building blocks (see Fig. 2).

以这种方式处理数据允许对数据进行完全的脱机回放，并提供了将流数据组织为离散模型构建块的时间线(见图2)。

Once data is blocked into time segments, these blocks are aligned and registered into a model.LIDAR data streams and survey measurements, for example, are blocked into point clouds and robot position/pose data, respectively. Point clouds are filtered for outliers and
globally registered using the surveyed position and pose estimates as the starting point for multi-view surface matching [10] and global ICP [11]algorithms. Fig, 4 shows a 2D map generated from 172 scans of varying views that were autonomously registered from these algorithms.

一旦数据被阻塞成时间段，这些块就被对齐并注册到模型中。例如，激光雷达数据流和测量数据分别被阻塞
成点云和机器人位置/姿态数据。将点云过滤为离群值，并使用所调查的位置和姿态估计作为多视图表面匹
配[10]和全局ICP[11]算法的起点进行全局注册。图4显示了一个2D地图，它是由172次不同视图的扫描生成
的，这些视图是由这些算法自主注册的。

survey measurement

调查测量

outliers

n. [分化] 异常值；离群值；离开本体的东西；局外人（outlier的复数）

Following scan registration, updated position and pose information (obtained through the registration process) corrects orientation for all robot sensors. This step allows data from each sensor to be globally registered and thereby visually represented as texture on a 3D model. Spatial visualization of non-geometric data greatly enhances the aesthetic appeal
and contextual understanding of 3D models.

扫描注册后，更新位置和姿势信息(通过注册过程获得)，修正所有机器人传感器的方向。这一步允许每个传
感器的数据进行全局注册，从而直观地表示为3D模型上的纹理。非几何数据的空间可视化极大地提高了三
维模型的审美情趣和语境理解能力。

registration
英 [redʒɪ’streɪʃ(ə)n] 美 ['rɛdʒɪ’streʃən]

n. 登记；注册；挂号

texture
英 ['tekstʃə] 美 ['tɛkstʃɚ]

n. 质地；纹理；结构；本质，实质

spatial
英 ['speɪʃ(ə)l] 美 ['speʃl]

adj. 空间的；存在于空间的；受空间条件限制的

aesthetic
英 [i:sˈθetɪk] 美 [esˈθetɪk; ɛsˈθɛtɪk]

adj. 美的；美学的；审美的，具有审美趣味的

contextual
英 [kɒn’tekstjʊəl] 美 [kən’tɛkstʃuəl]

adj. 上下文的；前后关系的

One example of this texture mapping capability is shown in Fig. 3 where high-resolution fisheye photographs are un-warped and applied to the surface model. .

这个纹理映射功能的一个例子如图3所示，其中高分辨率鱼眼照片没有扭曲并应用于表面模型。

This procedure takes a projected 2D image plane and, though knowledge of camera and lens parameters (obtained via calibration), maps every image pixel to ray on a unit hemisphere. Using this transformation and the relative LIDAR and camera poses, each point in the range image is colored with information from the fisheye image.

这个过程需要一个投影的二维图像平面，虽然知道相机和镜头参数(通过校准获得)，但是将每个图像像素映
射到单元半球上的光线。利用这种转换和相对的激光雷达和摄像机的姿态，距离图像中的每个点都用鱼眼图
像的信息着色。

ray
英 [reɪ] 美 [re]

n. 射线；光线；【鱼类】鳐形目(Rajiformes)鱼
vt. 放射；显出
vi. 放射光线；浮现

4
Results
The described system has undergone extensive utilization in a variety of underground domains. The results show in Fig. 2,3,4 were acquired from a research coal mine located outside of Pittsburgh, PA.

所述系统在各种地下领域得到了广泛的应用。图2、图3、图4是在宾夕法尼亚州匹兹堡市外的一个煤矿进行的研究。

Similar results have also been logged in limestone mines, concrete passageways, hard rock tunnels and indoor settings.

石灰石矿、混凝土通道、硬岩隧道和室内设置也有类似的结果。

The earlier claims that robotic surveying provide efficiency,repeatability and accuracy for subterranean data collection are supported by the results shown in Table 1 and Fig. 5.

表1和图5所示的结果支持了早先关于机器人测量为地下数据收集提供效率、重复性和准确性的说法。

Table 1 shows a comparison among human, mixed human and robot and robot-only surveying operations in terms of operational performance. Fig. 5 shows a comparison between a human-surveyed mine map and one generated by taking slices from a robot-surveyed mine model.

表1显示了人类、混合人类和机器人以及只使用机器人的测量操作在操作性能方面的比较。图5显示了人工勘
测矿图与由机器测量矿井模型生成的图之间的比较。

Table 1 sumarizes performance results recorded from actual field data 2 in three modes of operation: H for human-only, HR for mixed human-robot and R for robot-only. All three data sets utilized the same equipment (i.e. robot and surveying instrument), followed the same procedure of moving and scanning as described in Section 2 and occurred in different yet comparable subterranean environments.

表1将现场实测数据2在三种操作模式下的性能结果进行拼接:H为纯人模式，HR为混合人-机器人模式，R为
纯机器人模式。这三个数据集使用相同的设备(即机器人和测量仪器)，按照第2节中描述的相同的移动和扫描
过程，并且发生在不同但类似的地下环境中。

Human-only data required a team of two people to manually operate the survey instrument and puppeteer the robot. Human-robot data required one person to operate the survey instrument while the robot autonomously navigated, scanned and signaled
the human to take survey measurements. Robot-only required a human to monitor the surveying process, but the operation was conducted without direct human interaction.

只有人类的数据需要一个两个人的团队来手动操作测量仪器和机器人木偶。人类-机器人数据需要一个人操	
作测量仪器，而机器人则自动导航、扫描并向人类发出信号进行测量。机器人只需要一个人来监视测量过
程，但操作是在没有人直接参与的情况下进行的。

Table 1 is composed of the following information: The sample size for each of these results is denoted by the Size column and is in units of robot stations (i.e. the number of scan-transition cycles). To accurately portray performance characteristics, stations where errors occurred were removed from statistical computations.

表1由以下信息组成:每个结果的样本量用size列表示，单位为机器人站(即扫描-转换周期数)。为了准确地描	
述性能特征，从统计计算中删除发生错误的站点。

portray
英 [pɔː’treɪ] 美 [pɔr’tre]

vt. 描绘；扮演

In this work, errors are considered to be stations where one or more of the survey readings were corrupted or the cycle time was larger than two standard deviations from the sample mean. Average error is reported in the Err/Size column.

在这项工作中，误差被认为是一个或多个调查读数被损坏或周期时间比样本均值大两个标准
差的站点。在Err/Size列中报告平均错误。

deviation
英 [diːvɪ’eɪʃ(ə)n] 美 ['divɪ’eʃən]

n. 偏差；误差；背离

The error-free averages (fist number) and standard deviation (highlighted second number)
of the scan and transition times (in seconds) are reported in columns S-Time and T-Time, respectively. The estimated period (in seconds) for scan-transition cycles are show in the Period column.

扫描的无错误平均(第一个数字)和标准偏差(突出显示的第二个数字)以及转换时间(以秒为单位)分别在S-Time
和T-Time列中报告。扫描-转换周期的估计周期(以秒为单位)显示在period列中。

The results in Table 1 show clear performance improvements as robot autonomy is incorporated into the surveying process. The error rates are a factor of 2 higher in human surveying verses robot autonomy, with full autonomy showing a slightly lower error rate than mixed human-robot.

表1中的结果表明，机器人的自主性在测量过程中得到了明显的改善。在人类的调查中，错误率比机器人的
自主性高2倍，完全自主性比混合人类机器人的错误率略低。

incorporate
英 [ɪn’kɔːpəreɪt]

vt. 包含，吸收；体现；把……合并
vi. 合并；混合；组成公司
adj. 合并的；一体化的；组成公司的

The mean survey time is consistent over all modes; however, the reported deviation in human operation time is double the repeatability of robot-only.Transition time shows the greatest difference in performance with human surveying appearing two times slower than automation.
各模式平均调查时间一致;然而，报告的人类操作时间偏差是robot-only的两倍。过渡时间的性能差异最大，人工测量的速度比自动化慢两倍。

This result is a consequence of the costly preparation time required to manually move the
robot verses the small overhead involved in allowing the robot to move itself. The error margin in these between-scan transition times is 1/20 th of second for the robot and 18 seconds for the human.

这个结果是由于手动移动机器人需要花费昂贵的准备时间，而不是让机器人自己移动所涉及的小开销造成
的。这些扫描间转换时间的误差对于机器人是1/20秒，而人类的是18秒。

Interestingly, this transition-time statistic is repeated for HR and R modes because the robot is fully automated during this part of the process in both modes.
Altogether, these numbers show the automated system being 21.5% faster than the human-only system and 28% more productive.

有趣的是，这种转换时间统计量在HR和R模式下是重复的，因为在这部分过程中，机器人在这两种模式
是完全自动化的。总的来说，这些数字表明自动化系统比人工系统快21.5%，效率高28%。

In prior evaluations, the reported accuracy of this process has been shown to be within 8cm of surface measurements acquired by alternative profiling technologies such as human survey and stationary laser profiling systems.

在先前的评估中，报告的这个过程的精度已经被证明是在8厘米以内的表面测量获得的替代分析技术，如人	
类调查和固定激光分析系统。

Fig. 5, for example, shows the difference between a decades-old human-surveyed mine map and one generated through robotic survey over a traverse of 183 meters. In areas where concrete walls exist, which define a flat and easily comparable surface, the prior map and robot survey are within 7.24cm of each other.

例如，图5显示了几十年前人类勘测的矿井地图与机器人勘测183米导线生成的矿井地图之间的差异。在混凝
土墙存在的区域，定义了一个平坦且易于比较的表面，先前的地图和机器人的测量彼此之间的距离为7.24厘米。

Natural surfaces, such as the rock wall, are within 15cm of each other; however, the human survey tends to “idealize” walls whereas as robotic survey captures the true interior surface.

自然表面，如岩壁，彼此之间距离在15厘米以内;然而，人类的调查倾向于“理想化”墙壁，而机器人的调查捕
捉到真实的内部表面。

Topologically, intersections align exactly to locations specified on the prior map. The notable differences are caused by alterations in the environment.

从拓扑结构上讲，交叉点精确地与先前地图上指定的位置对齐。显著的差异是由环境的变化引起的。

topologically

adv. 拓扑地

alteration
英 [ɔːltə’reɪʃ(ə)n; 'ɒl-] 美 [,ɔltə’reʃən]

n. 修改，改变；变更

As shown in magnified portion of Fig. 5, equipment and structural modifications dramatically alter the subterranean surface, which suggest yet another interesting application for this technology: the ability to produce temporal subterranean maps that monitor change over time.

如图5放大的部分所示，设备和结构的改变极大地改变了地下表面，这表明了这一技术的另一个有趣的应用:
随时间产生监视变化的地下地图的能力。

modification
英 [,mɒdɪfɪ’keɪʃ(ə)n] 美 [,mɑdɪfɪ’keʃən]

n. 修改，修正；改变

temporal
英 ['temp(ə)r(ə)l] 美 ['tɛmpərəl]

adj. 暂时的；当时的；现世的
n. 世间万物；暂存的事物

Visualization and analysis tools are show in Fig. 6. The left portion of Fig. 6 shows a multi-view registration utility that colors individual point clouds inside of a 3D model. This colored model assists in quickly assessing potential problems that may have occurred during the automated registration process.

可视化分析工具如图6所示。图6的左边部分显示了一个多视图注册工具，该工具可以在3D模型中为单个点
云着色。这种彩色模型有助于快速评估在自动注册过程中可能发生的潜在问题。

The right portion of Fig. 6 shows a tool for extracting area and volumetric measurements from model information. Such tools are invaluable for data interaction and can also be automated to produce sectional views and record measurements over long stretches of model.

图6的右侧显示了一个工具，用于从模型信息中提取面积和体积测量数据。这样的工具对于数据交互是非常
宝贵的，并且可以自动生成剖视图，并在模型的很长一段时间内记录测量结果。

long stretches of

很长一段的

volumetric
英 [,vɒljʊ’metrɪk] 美 [,vɑljə’mɛtrɪk]

adj. [物] 体积的；[物] 容积的；[物] 测定体积的