WO2022077561A1 - 一种衡量缺失点云覆盖度的点云补全测评方法 - Google Patents

一种衡量缺失点云覆盖度的点云补全测评方法 Download PDF

Info

Publication number
WO2022077561A1
WO2022077561A1 PCT/CN2020/124247 CN2020124247W WO2022077561A1 WO 2022077561 A1 WO2022077561 A1 WO 2022077561A1 CN 2020124247 W CN2020124247 W CN 2020124247W WO 2022077561 A1 WO2022077561 A1 WO 2022077561A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
missing
coverage
point
complete
Prior art date
Application number
PCT/CN2020/124247
Other languages
English (en)
French (fr)
Inventor
李革
晏玮
张若楠
Original Assignee
北京大学深圳研究生院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京大学深圳研究生院 filed Critical 北京大学深圳研究生院
Publication of WO2022077561A1 publication Critical patent/WO2022077561A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/60Analysis of geometric attributes
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10028Range image; Depth image; 3D point clouds
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30168Image quality inspection

Definitions

  • the invention relates to the technical field of image recognition, in particular to the field of point cloud and point cloud completion, and in particular to a point cloud completion evaluation method for measuring the coverage of missing point clouds.
  • the point cloud data obtained by using the existing point cloud acquisition device or depth perception may exist due to limited scanning area, incomplete angle, physical environmental light influence, limitations of the laser scanner itself, or the collection structure of objects is too complex and changeable, etc. , which inevitably lead to the presence of missing areas in the scan results.
  • some subsequent point cloud processing tasks require a complete point cloud.
  • Point cloud completion is the task of inferring the most likely complete shape of the missing point cloud. Point cloud completion has become a hot research field. Many point cloud completion methods have been proposed, and how to propose an objective indicator or formula to better measure the effect of point cloud completion methods has also become a key issue.
  • the existing methods for measuring point cloud completion all use the method of measuring the difference between two complete point clouds, and do not redesign the evaluation indicators for the characteristics of point cloud completion tasks.
  • the existing evaluation index measures the point-to-point Euclidean distance between two point clouds as the evaluation index for point cloud completion. This measurement method does not have the invariance of missing degree, that is, the same proportion of missing, but different missing parts, the index's The calculation results will also change accordingly, which is not in line with the quantitative performance of missing completeness. At the same time, since this indicator will be affected by outliers, the missing rate will be larger, and the indicator will not appear monotonic. It is also not good for measuring the completeness of the missing point cloud.
  • the method proposed in the present invention has great advantages in the following two aspects: first, it has better robustness of missing parts of point clouds. Second, it has a stronger linear correlation degree of missing rate. The two concepts of robustness of missing parts of point cloud and linear correlation degree of missing rate are explained below.
  • Linear correlation degree of missing rate Given a function f that measures the missing rate of point cloud, the degree of linear correlation between it and missing rate can be measured by linear correlation coefficient, for example, the famous Pearson correlation coefficient (PLCC) and Spearman's correlation coefficient (SROCC).
  • PLCC Pearson correlation coefficient
  • SROCC Spearman's correlation coefficient
  • CD Chamfer Distance
  • CD calculates the degree of difference between two point clouds by calculating the difference of the point-to-point Euclidean distance between the two point clouds, but it cannot measure the degree of missing point cloud relative to the complete point cloud, and its missing parts Robustness and missing rate are less linearly correlated.
  • the present invention adopts the following technical solutions to achieve the purpose:
  • the present invention aims to measure the missing degree of the point cloud, and the present invention does not measure the missing degree of the point cloud by the difference of the Euclidean distance between the points.
  • the invention measures the missing degree of the point cloud by measuring the coverage degree of one point cloud relative to another point cloud.
  • Step 1 Calculate the coverage of the missing point cloud relative to the complete point cloud: record the missing point cloud as A, record the corresponding complete point cloud as B, and use a mapping method for each point in the A point cloud Find one or more points in the B point cloud, and add these points to a set. Since the points in this set are all composed of the points in the B point cloud, it is calculated that this set accounts for the points in the B point cloud. scale, and use it as the coverage of point cloud A relative to point cloud B.
  • Step 2 Calculate the coverage of the complete point cloud relative to the missing point cloud: Step A is also performed on the B point cloud and the A point cloud, that is, the coverage degree of the B point cloud relative to the A point cloud is obtained.
  • Step 3 Combine the values calculated in step 1 and step 2: combine the calculation result of step 1 with the calculation result of step 2 to achieve a symmetry for the two point clouds A and B, and consider A and B at the same time.
  • the point cloud missing degree indicator value of the bidirectional relationship is
  • the point cloud completion evaluation method for measuring the coverage of missing point clouds of the present invention has the following beneficial effects:
  • this method proposes a point cloud completion evaluation method to measure the coverage of missing point clouds. By calculating the coverage between two point sets as the evaluation standard, it will not be affected by individual outliers. This is because the number of outliers is very small compared to the entire point cloud, so the proportion of outliers contributes very little to the final coverage.
  • the CD distance is measured by the difference of the Euclidean distance. If the outlier point is very far away from the main point cloud, it will have a greater impact on the final result.
  • Fig. 1 is the flow chart of the present invention
  • Figure 2b is missing 60% of the point cloud missing position 1;
  • Figure 2c is missing 60% of the point cloud missing position 2;
  • Figure 2d is missing 60% of the point cloud missing position 3;
  • Fig. 4 is the variance variation diagram of CD distance and the method of the present invention.
  • Fig. 5 is a change trend diagram of the CD distance and the method of the present invention.
  • Fig. 1 is a flowchart of the present invention, as shown in the figure, an embodiment of a point cloud completion evaluation method for measuring the coverage of missing point clouds of the present invention: the evaluation method includes,
  • Step 1 Calculate the coverage S1 of the missing point cloud relative to the complete point cloud: First, calculate an implementation method Cov(S eval , S GT ) index S1 of the coverage defined in this paper, where Cov(S eval , S GT ) represents coverage metrics,
  • S eval , S GT respectively represent the missing point cloud and its corresponding complete point cloud
  • d(x, y) represents any distance metric function, such as Euclidean distance
  • argmin represents the minimum value that satisfies the conditions on the right. Elements.
  • the text of formula 1 is interpreted as: for each point x in the missing point cloud S eval to be measured, find the point y closest to its distance d in the complete point cloud S GT , and divide the number of points in the set formed by y by The number of points in the complete point cloud S GT is the coverage.
  • the coverage defined in this paper can measure the proportion of the missing point cloud to the original point cloud to a certain extent.
  • Step 2 Calculate the coverage S2 of the complete point cloud relative to the missing point cloud: Calculate Cov(S GT , S eval ), the formula is the same as that of step 1.
  • Step 3 Perform a combination S3 of the values calculated in Step 1 and Step 2.
  • the combination is: F-score-cov S2, where F-score-cov represents the point cloud completion evaluation index,
  • F-score-cov has the following property: when Cov(S eval , S GT ) tends to zero or Cov(S GT , S eval ) tends to zero, F-score-cov also tends to zero.
  • This embodiment discloses a point cloud completion evaluation method for measuring the coverage of missing point clouds, which specifically includes the following steps:
  • Step S1 Calculate two coverages Cov(S eval , S GT ) and Cov(S GT , S eval ) of the missing point cloud and its corresponding complete point cloud according to formula (1), where S eval is the missing point cloud, S eval GT is the complete point cloud;
  • the distance index d in formula (1) can be any distance metric.
  • Step S2 Calculate the evaluation index F-score-cov(S eval , S GT ) according to formula (2). In this way, the point cloud completion index proposed in this paper to measure the coverage of missing point clouds is obtained.
  • Effect drawing The following shows the effect drawing of the present invention compared to the CD distance [1].
  • Figures 2a to d show the index effects of different missing positions under the same missing rate.
  • the value of CD is 0.101
  • the Cov value of the present invention is 0.393
  • the F-score-cov value of the present invention is 0.562
  • the value of CD is 0.052
  • the Cov value of the present invention is 0.394
  • the F-score-cov value of the present invention is 0.563
  • the CD value is 0.074
  • the Cov value of the present invention is 0.395
  • the F-score-cov value of the present invention is 0.564.
  • Figures 3a to d show that the missing rate is getting larger and larger, and the index proposed by the present invention presents a monotonic trend, which can be used to measure the degree of missing.
  • the indicators Cov (coverage, a form as shown in formula (1) and F-score-cov proposed in the figure are both monotonic with the missing rate decrease, while the CD distance does not have this monotonic change phenomenon.
  • Fig. 4 is the variance change diagram of CD distance and the method of the present invention.
  • variance in Fig. 4 is the variance
  • GT is the complete point cloud
  • different missing parts are different missing parts
  • CD distance and the method proposed by the present invention (represented by F-score)
  • F-score the variance of the value changes with the change of the missing part; it can be seen from Figure 4 that the value of the CD distance fluctuates to a large extent with the change of the missing part , while the index calculated by the present invention remains stable with the change of the missing part, and the variance of the index calculated by the present invention with the change of the missing part is significantly smaller than CD.
  • Fig. 5 is the change trend diagram of CD distance and the method of the present invention, as shown in Fig. 5, the missing ratio in Fig. 5 is missing ratio, PLCC is the Pierce correlation coefficient, SROCC is the Spearman correlation coefficient, missing missing, CD distance and the present invention
  • the change trend of the proposed method represented by F-score) with the increase of the missing rate, it can be seen from Figure 5 that the linear correlation degree of the missing rate of the method of the present invention is significantly higher than that of CD.
  • this embodiment discloses a point cloud completion evaluation method for measuring the coverage of missing point clouds.
  • the calculation result of the indicator will also change, which is not in line with the quantitative performance of missing completeness.
  • the existing method indicators will be affected by The impact of outliers, so the missing rate will be larger, and the indicator will not show monotonic performance, which is not a good measure of the degree of missing. Therefore, the point cloud completion evaluation method for measuring the coverage of missing point clouds proposed by the present invention can solve the above problems by calculating the ratio of matching points between two point sets as the evaluation standard, and provides a method for point cloud completion. New evaluation method.
  • a point cloud completion evaluation method for measuring the coverage of missing point clouds can be used in the field of image recognition technology, especially in the field of point cloud and point cloud completion. It is suitable for 3D scanning equipment (laser, radar, etc.), widely used in the construction of urban digital maps, and plays a technical supporting role in many popular researches such as smart cities, unmanned driving, and cultural relics protection.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明涉一种衡量缺失点云覆盖度的点云补全测评方法,该方法包括步骤一:计算缺失点云相对于完整点云的覆盖度;步骤二:计算完整点云相对于缺失点云的覆盖度;步骤三:将步骤一和步骤二计算的值进行组合;本发明的衡量缺失点云覆盖度的点云补全测评方法能够满足衡量缺失点云的缺失不变性,根据缺失率的单调变化而呈现单调变化趋势,能够直观的衡量点云补全方法的补全效果。

Description

一种衡量缺失点云覆盖度的点云补全测评方法 技术领域
本发明涉及图像识别技术领域,具体为点云及点云补全领域,特别涉及一种衡量缺失点云覆盖度的点云补全测评方法。
背景技术
使用现有点云采集装置或深度感知获取的点云数据,由于可能存在扫描区域有限、角度不完全、物理环境光线影响、激光扫描仪本身的局限性、或者物体的集合结构过于复杂多变等情况,都不可避免的致使扫描结果中存在缺失区域。而后续的一些点云处理任务需要一个完整的点云,这时如何将缺失点云补全成完整点云就是一个值得研究的问题。点云补全就是完成这样的任务,即根据缺失点云推测出其最有可能的完整形状。点云补全已经成为一个热门的研究领域,很多点云补全的方法都被提出,而怎样提出一个客观指标或者公式来更好的衡量点云补全方法的效果也成为一个关键问题。现有衡量点云补全的方法都是使用衡量两个完整点云之间的方法,没有为点云补全任务的特点来重新设计评价指标。现有的评价指标通过衡量两个点云之间的点对点欧式距离作为点云补全的评价指标,这种衡量方法不具有缺失程度不变性,即缺失相同的比例,但缺失部位不同,指标的计算结果也会随之变化,这不符合缺失完整度的量化表现,同时,由于这种指标会受局外点的影响,因此会出现缺失率越大,而指标不会出现单调表现。这也不利于衡量缺失点云的完整程度。
发明内容
本发明提出的方法与现有衡量点云补全的测评方法相比,在如下两个方面具有较大优势:一是具有更好的点云缺失部位鲁棒性。二是具有更强的缺失率线性相关程度。下面分别解释点云缺失部位鲁棒性和缺失率线性相关程度这两个概念。
1、点云缺失部位鲁棒性:给定一个完整点云X,一种产生缺失点云的遮罩函数mask:X′=mask(X,α,p)。此遮罩函数用于根据给定的缺失率α和缺失位置中心p∈R 3产生不完整点云X′。现有一个衡量点云缺失率的函数f(X,X′)=f(X,mask(X,α,p))。如果f的函数值不会随着p的位置变化而变化,则函数f具有缺失位置鲁棒性。
2、缺失率线性相关程度:给定一个衡量点云缺失率的函数f,其与缺失率之间的线性相关程度可以用线性相关系数来衡量,例如,著名的皮尔森相关系数(PLCC)和斯皮尔曼相关系数(SROCC)。
现有的方法通常使用倒角距离(Chamfer Distance,简称CD)来衡量缺失点云相对于完整的点的完整程度,CD的公式如下:
Figure PCTCN2020124247-appb-000001
CD通过计算两个点云之间点对点的欧氏距离的差来计算两个点云之间的差异程度,但是其无法衡量缺失点云相对于完整点云之间的缺失程度,并且其缺失部位鲁棒性和缺失率线性相关程度较差。
从图2a至图2d可以看出,在缺失率相同的情况下,CD距离的变化幅度较大,即说明其缺失部位鲁棒性差。
从图3a至图3d可以看出,在随着缺失率越来越大的时候,CD距离并不会随着缺失率的增大而具有单调变化趋势,即说明其缺失率线性相关程度较差。
为解决上述技术问题,本发明采用如下技术方案来达到目的:
与CD距离不同,本发明旨在衡量点云缺失程度,本发明不通过点与点之间的欧式距离的差异来衡量点云的缺失程度。本发明通过衡量一个点云相对于另一个点云的覆盖程度大小来衡量点云的缺失程度。
步骤一、计算缺失点云相对于完整点云的覆盖度:将缺失点云记为A,将其对应的完整点云记为B,将A点云中的每个点都用一种映射方式找到B点云中的一个或多个点,将这些点加入到一个集合中,由于这个集合中的点都是由B点云中的点组成,因此计算这个集合占B点云中的点的比例大小,并将其作为A点云相对于B点云的覆盖程度。
步骤二、计算完整点云相对于缺失点云的覆盖度:对B点云也对A点云同样进行步骤A,即得到B点云相对于A点云的覆盖程度。
步骤三、将步骤一和步骤二计算的值进行一种组合:将步骤一的计算结果与步骤二计算结果进行结合,达到一个对于A和B这两个点云对称,并且同时考虑A和B的双向关系的点云缺失程度指标值。
本发明的衡量缺失点云覆盖度的点云补全测评方法与现有技术相比,具有如下的有益效果:
1、具有较强的点云缺失部位鲁棒性,弥补了现有方法点云缺失部位鲁棒性较差的缺点。与现有的CD距离相比,本方法提出一种衡量缺失点云覆盖度的点云补全测评方法,通过计算两个点集之间的覆盖程度来作为评价标准,不会受到个别离群点的影响,这是因为离群点相比于整个点云的点数是非常少的,所以离群点的占比对于最终的覆盖度贡献也非常小。而CD距离用过欧式距离的差值作为衡量标准,如果离群点与主点云离得非常远,那么将会对最终得结果产生较大的影响。
2、具有较强的缺失率线性相关程度,弥补了现有方法缺失率线性相关程度较 低的缺点。与现有CD距离相比,本方法通过计算两个点集之间的覆盖程度来作为评价指标,如果一个缺失点云缺失率越大,那么其占完整点云的占比必然越小,与欧式距离的差值完全无关,因此本发明相对于CD具有更强的缺失率线性相关程度。
附图说明
图1本发明的流程图;
图2a完整点云;
图2b缺失60%点云缺失位置1;
图2c缺失60%点云缺失位置2;
图2d缺失60%点云缺失位置3;
图3a完整点云;
图3b缺失40%点云;
图3c缺失50%点云;
图3d缺失60%点云;
图4为CD距离和本发明方法的方差变化图;
图5为CD距离和本发明方法的变化趋势图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚,下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整的描述,显然,所描述的实施例是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例, 本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本方面保护的范围。
图1本发明的流程图,如图所示本发明一种衡量缺失点云覆盖度的点云补全测评方法的实施例:该测评方法包括,
步骤一、计算缺失点云相对于完整点云的覆盖度S1:首先计算本文定义的覆盖度的一种实现方法Cov(S eval,S GT)指标S1,其中Cov(S eval,S GT)表示覆盖度指标,
Figure PCTCN2020124247-appb-000002
上式中,S eval,S GT分别表示被测量的缺失点云和它对应的完整点云,d(x,y)代表任意距离度量函数,例如欧式距离,argmin代表取满足右侧条件中最小的元素。公式1的文字解释为:对于待测量的缺失点云S eval中的每个点x,找到完整点云S GT中与其距离d最近的点y,将y组合而成的的集合的点数除以完整点云S GT的点数,即为覆盖度。本文定义的覆盖度在一定程度上能够衡量缺失点云占原始点云的比例。
步骤二、计算完整点云相对于缺失点云的覆盖度S2:计算Cov(S GT,S eval),公式同步骤一。
步骤三、将步骤一和步骤二计算的值进行一种组合S3,该组合为:F-score-cov S2,其中F-score-cov表示点云补全评价指标,
Figure PCTCN2020124247-appb-000003
F-score-cov具有如下性质:当Cov(S eval,S GT)趋于零或者Cov(S GT,S eval)趋于零时,F-score-cov也会趋于零。
这就是本发明提出的一种衡量缺失点云覆盖度的点云补全测评方法的一种实现方式。
本实施例公开了一种衡量缺失点云覆盖度的点云补全测评方法,具体包括下列步骤:
步骤S1、按照公式(1)计算缺失点云及其对应完整点云的两种覆盖度Cov(S eval,S GT)和Cov(S GT,S eval),其中S eval是缺失点云,S GT是完整点云;
公式(1)中的距离指标d可以是任意距离度量。
步骤S2、按照公式(2)计算评价指标F-score-cov(S eval,S GT)。这样就得到了本文提出的衡量缺失点云覆盖度的点云补全指标。
效果图:下面展示了本发明相比于CD距离[1]的效果图。
图2a至d展示的是相同缺失率下,缺失位置不同的指标效果。其中,图2b中,CD的值为0.101,本发明的Cov值为0.393,本发明的F-score-cov值为0.562;图2c中,CD的值为0.052,本发明的Cov值为0.394,本发明的F-score-cov值为0.563;图2d中,CD的值为0.074,本发明的Cov值为0.395,本发明的F-score-cov值为0.564。从图2a至图2d可以看出,在相同缺失率的情况下,CD距离的值随着缺失部位变化较大,而本发明提出的方法计算出来的值相对较为稳定。
图3a至d展示的是缺失率越来越大,本发明提出的指标呈现单调趋势,可以用于衡量缺失程度大小。从图3a至d可以看出,随着缺失率越来越大,图中提出的指标Cov(覆盖度,一种形式如公式(1)所示和F-score-cov都随着缺失率单调减少,而CD距离则无此单调变化现象。
图4为CD距离和本发明方法的方差变化图,如图4所述,图4中variance为方差,GT为完整点云,different missing part为不同的缺失部位,CD距离和本发明提出的方法(用F-score表示)对于缺失相同比例的缺失点云,随着缺失部位变化的值的方差变化;从图4可以看出,CD距离的值随着缺失部位变化而发生较 大程度的波动,而本发明计算的指标随着缺失部位变化而保持稳定,本发明计算的指标随着缺失部位的变化的方差显著小于CD。
图5为CD距离和本发明方法的变化趋势图,如图5所示,图5中missing ratio缺失率,PLCC为皮尔斯相关系数,SROCC为斯皮尔曼相关系数,missing缺失,CD距离和本发明提出的方法(用F-score表示)随着缺失率增大的变化趋势,从图5可以看出,本发明的方法的缺失率线性相关程度显著高于CD。
综上所述,本实施例公开了一种衡量缺失点云覆盖度的点云补全测评方法。针对目前衡量方法不具有缺失程度不变性,即缺失相同的比例,但缺失部位不同,指标的计算结果也会随之变化,这不符合缺失完整度的量化表现,同时,现有方法指标会受局外点的影响,因此会出现缺失率越大,而指标不会出现单调表现,这也无法很好的衡量缺失程度。因而本发明所提出的衡量缺失点云覆盖度的点云补全测评方法,通过计算两个点集之间的匹配点数比例来作为评价标准,能够解决上述问题,为点云补全提供一种新的评价方法。
上述实施例为本发明较佳的实施方式,但本发明的实施方式并不受上述实施例的限制,其他的任何未背离本发明的精神实质与原理下的所作的改变、修饰、替代、组合、简化等均应为等效的置换方式,都包含在本发明的保护范围之内。
参考附录:
[1]Panos Achlioptas,Olga Diamanti,et al.2018.Learning Representations and Generative Models for 3D Point Clouds.ICML(2018).
工业应用性
一种衡量缺失点云覆盖度的点云补全测评方法,可使用于图像识别技术领域,特别是点云及点云补全领域。适用于三维扫描设备(激光、雷达等),广泛应用于城市数字化地图的构建,在如智慧城市、无人驾驶、文物保护等众多热门研究中起技术支撑作用。

Claims (4)

  1. 一种衡量缺失点云覆盖度的点云补全测评方法,其特征在于,所述的包括以下步骤:
    步骤一、计算缺失点云相对于完整点云的覆盖度:将缺失点云记为A,将其对应的完整点云记为B,将A点云中的每个点都用一种映射方式找到B点云中的一个或多个点,将这些点加入到一个集合中,由于这个集合中的点都是由B点云中的点组成,因此计算这个集合占B点云中的点的比例大小,并将其作为A点云相对于B点云的覆盖程度;
    步骤二、计算完整点云相对于缺失点云的覆盖度:对B点云也对A点云同样进行步骤A,即得到B点云相对于A点云的覆盖程度;
    步骤三、将步骤一和步骤二计算的值进行组合:将步骤一的计算结果与步骤二计算结果进行结合,达到一个对于A和B这两个点云对称,并且同时考虑A和B的双向关系的点云缺失程度指标值。
  2. 根据权利要求1所述的点云补全测评方法,其他特征在于,该步骤一、计算缺失点云相对于完整点云的覆盖度中,所述的计算覆盖度Cov(S eval,S GT)的计算方法是:
    Figure PCTCN2020124247-appb-100001
    式(1)中,S eval,S GT分别表示被测量的缺失点云和它对应的完整点云,d(x,y)代表任意距离度量函数,例如欧式距离;公式1的文字解释为:对于待测量的缺失点云S eval中的每个点x,找到完整点云S GT中与其距离d最近的点y,将y组合而成的的集合的点数除以完整点云S GT的点数,即为覆盖度。
  3. 根据权利要求1所述的点云补全测评方法,其他特征在于,步骤二、计算完整点云相对于缺失点云的覆盖度,该步骤与步骤一相同。
  4. 根据权利要求1所述的点云补全测评方法,其他特征在于,步骤三、将 步骤一和步骤二计算的值进行组合中;该组合为F-score-cov:
    Figure PCTCN2020124247-appb-100002
PCT/CN2020/124247 2020-10-12 2020-10-28 一种衡量缺失点云覆盖度的点云补全测评方法 WO2022077561A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011083681.4 2020-10-12
CN202011083681.4A CN112215881A (zh) 2020-10-12 2020-10-12 一种衡量缺失点云覆盖度的点云补全测评方法

Publications (1)

Publication Number Publication Date
WO2022077561A1 true WO2022077561A1 (zh) 2022-04-21

Family

ID=74054479

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/124247 WO2022077561A1 (zh) 2020-10-12 2020-10-28 一种衡量缺失点云覆盖度的点云补全测评方法

Country Status (2)

Country Link
CN (1) CN112215881A (zh)
WO (1) WO2022077561A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114842180A (zh) * 2022-07-06 2022-08-02 山东海量信息技术研究院 一种点云补全方法、装置、设备及介质
CN115496881A (zh) * 2022-10-19 2022-12-20 南京航空航天大学深圳研究院 单目图像辅助的大型飞机点云补全方法
CN116519021A (zh) * 2023-06-29 2023-08-01 西北工业大学 一种惯性导航系统故障诊断方法、系统及设备

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170234775A1 (en) * 2016-02-11 2017-08-17 Caterpillar Inc. Wear measurement system using a computer model
CN108257213A (zh) * 2018-01-17 2018-07-06 视缘(上海)智能科技有限公司 一种点云轻量级的多边形曲面重建方法
CN109684005A (zh) * 2018-12-11 2019-04-26 浙江口碑网络技术有限公司 图形界面中组件相似度确定方法及装置

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104063898B (zh) * 2014-06-30 2017-05-03 厦门大学 一种三维点云自动补全方法
CN108492329B (zh) * 2018-03-19 2022-04-12 北京航空航天大学 一种三维重建点云精度和完整度评价方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170234775A1 (en) * 2016-02-11 2017-08-17 Caterpillar Inc. Wear measurement system using a computer model
CN108257213A (zh) * 2018-01-17 2018-07-06 视缘(上海)智能科技有限公司 一种点云轻量级的多边形曲面重建方法
CN109684005A (zh) * 2018-12-11 2019-04-26 浙江口碑网络技术有限公司 图形界面中组件相似度确定方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
FU ZHENXIN FUZHENXIN@PKU.EDU.CN; JI FENG ZHONGXIU.JF@ALIBABA-INC.COM; HU WENPENG WENPENG.HU@PKU.EDU.CN; ZHOU WEI FAYI.ZW@ALIBABA-I: "Query-bag Matching with Mutual Coverage for Information-seeking Conversations in E-commerce", PROCEEDINGS OF THE 28TH ACM JOINT MEETING ON EUROPEAN SOFTWARE ENGINEERING CONFERENCE AND SYMPOSIUM ON THE FOUNDATIONS OF SOFTWARE ENGINEERING, ACM, NEW YORK, NY, USA, 3 November 2019 (2019-11-03) - 13 November 2020 (2020-11-13), New York, NY, USA , pages 2337 - 2340, XP058640005, ISBN: 978-1-4503-7043-1, DOI: 10.1145/3357384.3358099 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114842180A (zh) * 2022-07-06 2022-08-02 山东海量信息技术研究院 一种点云补全方法、装置、设备及介质
WO2024007616A1 (zh) * 2022-07-06 2024-01-11 山东海量信息技术研究院 一种点云补全方法、装置、设备及介质
CN115496881A (zh) * 2022-10-19 2022-12-20 南京航空航天大学深圳研究院 单目图像辅助的大型飞机点云补全方法
CN115496881B (zh) * 2022-10-19 2023-09-22 南京航空航天大学深圳研究院 单目图像辅助的大型飞机点云补全方法
CN116519021A (zh) * 2023-06-29 2023-08-01 西北工业大学 一种惯性导航系统故障诊断方法、系统及设备
CN116519021B (zh) * 2023-06-29 2023-09-15 西北工业大学 一种惯性导航系统故障诊断方法、系统及设备

Also Published As

Publication number Publication date
CN112215881A (zh) 2021-01-12

Similar Documents

Publication Publication Date Title
WO2022077561A1 (zh) 一种衡量缺失点云覆盖度的点云补全测评方法
WO2021218314A1 (zh) 基于位置定位的事件识别方法、装置、设备及存储介质
Goodchild et al. A simple positional accuracy measure for linear features
US9934249B2 (en) Systems and methods for context-aware and personalized access to visualizations of road events
JP2019133646A (ja) 点群データ同士のマッチング関係を確定するための方法及び装置
CN112233240B (zh) 三维矢量地图的三维矢量数据切片方法、装置及电子设备
Florinsky Computation of the third‐order partial derivatives from a digital elevation model
CN106355197A (zh) 基于K‑means聚类算法的导航图像匹配过滤方法
TWI608744B (zh) 估算裝置及其無線網路之通訊品質估算方法及電表安裝方法
CN114387319B (zh) 点云配准方法、装置、设备以及存储介质
WO2022237821A1 (zh) 生成交通标志线地图的方法、设备和存储介质
CN103632338A (zh) 一种基于匹配曲线特征的图像配准评估法
Debella-Gilo Bare-earth extraction and DTM generation from photogrammetric point clouds including the use of an existing lower-resolution DTM
CN112116549A (zh) 用于评估点云地图精度的方法和装置
CN111597986B (zh) 用于生成信息的方法、装置、设备和存储介质
CN113240734A (zh) 一种基于鸟瞰图的车辆跨位判断方法、装置、设备及介质
CN116958434A (zh) 多视图三维重建方法、测量方法及系统
CN111539297B (zh) 一种高分辨率遥感影像道路信息半自动提取方法
TWM596930U (zh) 不動產估價裝置
CN115952248B (zh) 终端设备的位姿处理方法、装置、设备、介质及产品
CN112130166A (zh) 基于反光板网络的agv定位方法及装置
US10546418B2 (en) Visualization of positional geospatial uncertainty
CN115937449A (zh) 高精地图生成方法、装置、电子设备和存储介质
Daudt et al. Learning to understand earth observation images with weak and unreliable ground truth
Lyu et al. Movement-aware map construction

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20957339

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20957339

Country of ref document: EP

Kind code of ref document: A1