CN109344750B - Complex structure three-dimensional object identification method based on structure descriptor - Google Patents
Complex structure three-dimensional object identification method based on structure descriptor Download PDFInfo
- Publication number
- CN109344750B CN109344750B CN201811101377.0A CN201811101377A CN109344750B CN 109344750 B CN109344750 B CN 109344750B CN 201811101377 A CN201811101377 A CN 201811101377A CN 109344750 B CN109344750 B CN 109344750B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- dimensional
- units
- center
- unit
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 15
- 238000012216 screening Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 abstract description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a complex structure three-dimensional object identification method based on a structure descriptor, which comprises the steps of extracting point cloud units corresponding to all parts in a target real three-dimensional point cloud, constructing a standard three-dimensional structure descriptor by adopting the size of an enclosure box of the point cloud unit corresponding to each part, the distance between the center of a core point cloud unit and the centers of other point cloud units and the included angle formed by connecting lines of the centers of each pair of point cloud units and the center of the core point cloud unit, segmenting a three-dimensional scene to be identified to obtain all point cloud units in the three-dimensional scene, constructing a candidate three-dimensional structure descriptor, calculating the Manhattan distance between the candidate three-dimensional structure descriptor and the standard structure descriptor, and if the distance is smaller than a set second threshold value, determining that the point cloud unit corresponding to the candidate three-dimensional structure descriptor is an object to be identified. The method has low calculation consumption, and can realize rapid and accurate identification of the three-dimensional object with the complex structure in the three-dimensional point cloud scene.
Description
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a complex-structure three-dimensional object identification method based on a structure descriptor.
Background
Object recognition in complex scenes is a very important research area for computer vision. Two-dimensional object recognition has been extensively studied over the past few decades and has become a relatively mature field. The three-dimensional point cloud may provide more geometric information than the two-dimensional image, and thus estimating the pose of the object in the three-dimensional point cloud is more accurate than estimating the pose in the two-dimensional image.
The three-dimensional object descriptor based on the point cloud is a popular research, and the aim of the three-dimensional object descriptor is to achieve the aim of identifying the point cloud by describing the local part or the whole of the point cloud. The existing three-dimensional point cloud descriptor can effectively identify a single target, but cannot be applied to identification of three-dimensional objects with complex structures in a three-dimensional point cloud scene. In the process of segmenting and identifying the three-dimensional scene, a three-dimensional object with a complex structure is often segmented into a plurality of point clouds, and the point cloud descriptor which can only be applied to a single point cloud at this time is invalid.
How to identify the three-dimensional object with the complex structure is still a problem which needs to be solved urgently in the industry.
Disclosure of Invention
The invention provides a complex structure three-dimensional object identification method based on a structure descriptor, which solves the problem that the existing three-dimensional descriptor can not identify a complex structure three-dimensional object and can effectively identify the complex structure three-dimensional object in a three-dimensional point cloud scene.
A structure descriptor-based complex structure three-dimensional object recognition method comprises the following steps:
extracting point cloud units corresponding to all parts in the target real three-dimensional point cloud, and calculating to obtain the size and the center of an enclosure box of the point cloud unit corresponding to each part;
selecting one point cloud unit from the point cloud units corresponding to the components as a core point cloud unit, and calculating the distance from the center of the core point cloud unit to the centers of other point cloud units;
pairing other point cloud units except the core point cloud unit in pairs, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit;
constructing a standard three-dimensional structure descriptor by adopting the size of an enclosure box of a point cloud unit corresponding to each component, the distance from the center of a core point cloud unit to the centers of other point cloud units and an included angle formed by connecting the center of each pair of point cloud units and the center of the core point cloud unit;
dividing a three-dimensional scene to be identified to obtain all point cloud units in the three-dimensional scene, and screening out all point cloud units with the number of points greater than a first threshold;
traversing the screened point cloud units, taking any point cloud unit as a core point cloud unit, optionally selecting n point cloud units in a three-dimensional neighborhood of the core point cloud unit, and constructing a candidate three-dimensional structure descriptor, wherein n is equal to the number of the point cloud units corresponding to each part in the target real three-dimensional point cloud minus one;
and calculating the Manhattan distance between the candidate three-dimensional structure descriptor and the standard structure descriptor, and if the distance is smaller than a set second threshold, the point cloud unit corresponding to the candidate three-dimensional structure descriptor is the object to be identified.
Further, the constructing of the candidate three-dimensional structure descriptor includes:
calculating the size and the center of a bounding box of the core point cloud unit and n point cloud units selected in the three-dimensional neighborhood of the core point cloud unit;
calculating the distance from the center of the core point cloud unit to the centers of other point cloud units;
pairing other point cloud units except the core point cloud unit in pairs, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit;
and constructing a candidate three-dimensional structure descriptor by adopting the size of the bounding box of each point cloud unit, the distance from the center of the core point cloud unit to the centers of other point cloud units and an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit.
The invention provides a three-dimensional structure descriptor based complex structure three-dimensional object recognition method, and aims to recognize complex structure three-dimensional objects in a three-dimensional reconstruction scene by describing the relation of each part of the complex structure three-dimensional object. The bounding box information of the object parts, the distance between the parts and the included angle combination are used as identification information, the calculation consumption is low, and the rapid and accurate identification of the three-dimensional object with the complex structure can be realized in the three-dimensional point cloud scene.
Drawings
FIG. 1 is a flow chart of a complex structure three-dimensional object recognition method based on a structure descriptor according to the present invention;
FIG. 2 is an example of a standard three-dimensional structure descriptor construction according to the present invention;
fig. 3 is a three-dimensional object recognition example of a complex structure according to the present invention.
Detailed Description
The technical solutions of the present invention are further described in detail below with reference to the drawings and examples, which should not be construed as limiting the present invention.
As shown in fig. 1, the method for identifying a three-dimensional object with a complex structure based on a structure descriptor in this embodiment includes:
and step S1, extracting point cloud units corresponding to all parts in the target real three-dimensional point cloud, and calculating to obtain the size and the center of the bounding box of the point cloud unit corresponding to each part.
The technical scheme is applied to the identification of the three-dimensional object with the complex structure, such as the identification of a machined part. In factory production, a plurality of parts may be mixed together, and three-dimensional object recognition is required to select a desired kind.
In the embodiment, a three-dimensional object to be identified is called a target, and a standard three-dimensional structure descriptor is constructed by using a target real three-dimensional point cloud.
In the example shown in fig. 2, the complex structured three-dimensional object (target) is composed of four parts, parts 1, 2, 3, 4 respectively. Fig. 2 shows the point cloud units corresponding to the components 1, 2, 3, 4, and the corresponding point cloud units are also described below with reference to the components 1, 2, 3, 4.
In the step, bounding boxes of the point cloud units corresponding to the components are calculated, and length, width and height information and center point coordinates of the components are obtained, wherein the center points of the four components are B, A, C, O respectively.
And step S2, selecting one point cloud unit from the point cloud units corresponding to the components as a core point cloud unit, and calculating the distance from the center of the core point cloud unit to the centers of other point cloud units.
In the step, one point cloud unit is selected from the point cloud units corresponding to all the parts as a core point cloud unit, and the part 4 is assumed as the core point cloud, and the distances from the parts 1, 2 and 3 to the core point cloud unit are calculated respectively, namely the distances between OB, OA and OC are calculated.
And step S3, pairing other point cloud units except the core point cloud unit, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit.
In the step, except for a component 4, the rest point cloud units are paired pairwise, and an included angle formed by each pair of the point cloud unit center and the component 4 point cloud unit center is calculated, wherein the included angle is angle AOC, angle BOC and angle AOB.
And step S4, constructing a standard three-dimensional structure descriptor by adopting the size of the bounding box of the point cloud unit corresponding to each component, the distance from the center of the core point cloud unit to the centers of other point cloud units and the included angle formed by the connecting lines of each pair of point cloud unit centers and the center of the core point cloud unit.
And linearly combining the size information of the bounding box, the distance between the centers of the point cloud units and the angle formed by the connecting lines between the centers of the point cloud units to obtain the standard three-dimensional structure descriptor.
Specifically, the size data of the bounding box is multiplied by a weight a1And sorting the bounding boxes from large to small according to the length to obtain 3n (each bounding box comprises 3-dimensional data of length, width and height, and n is the number of point cloud units except the core point cloud unit) dimensional data.
Multiplying the distance from the center of each point cloud unit to the center of the core point cloud unit by a1And ordering from large to small to obtain n-dimensional data.
Sequencing the connecting lines of each pair of point cloud unit centers and the core point cloud unit centers from large to small to obtainDimension data.
Combining the three data to obtain oneA standard three-dimensional structural descriptor of a dimension. In this embodiment, the target is composed of four parts, n being equal to 3.
And step S5, segmenting the three-dimensional scene to be identified to obtain all point cloud units in the three-dimensional scene, and screening out all point cloud units with the number of points greater than a first threshold.
In this embodiment, a three-dimensional scene to be recognized is segmented, and three-dimensional point cloud segmentation is already a relatively mature technology, and is not described herein again.
And after segmentation, obtaining all point cloud units in the three-dimensional scene, and screening out all point cloud units with the number of points greater than a first threshold. In this embodiment, the first threshold is used to remove the influence of noise and interference, and the point cloud units with the number less than or equal to the first threshold are removed.
For example, if the first threshold is 500, the number of points in the screened point cloud units is greater than 500, so that noise point clouds with the number of points of 500 or less are removed.
And step S6, traversing the screened point cloud units, selecting n point cloud units in a three-dimensional neighborhood of any point cloud unit as a core point cloud unit, and constructing a candidate three-dimensional structure descriptor, wherein n is equal to the number of the point cloud units corresponding to each part in the target real three-dimensional point cloud minus one.
In this embodiment, a candidate three-dimensional structure descriptor is constructed, that is, any point cloud unit is a core point cloud unit, n point cloud units are optionally selected in a three-dimensional neighborhood of the point cloud unit, then the size of an enclosure box of each point cloud unit, the distance from the center of the core point cloud unit to the centers of other point cloud units, and an included angle formed by connecting the center of each pair of point cloud units and the center of the core point cloud unit are calculated according to the methods of steps S1 to S3, and a three-dimensional structure descriptor is constructed according to the method of step S4.
Namely:
calculating the size and the center of a bounding box of the core point cloud unit and n point cloud units selected in the three-dimensional neighborhood of the core point cloud unit;
calculating the distance from the center of the core point cloud unit to the centers of other point cloud units;
pairing other point cloud units except the core point cloud unit in pairs, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit;
and constructing a candidate three-dimensional structure descriptor by adopting the size of the bounding box of each point cloud unit, the distance from the center of the core point cloud unit to the centers of other point cloud units and an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit.
The three-dimensional structure descriptor constructed in the step is a candidate three-dimensional structure descriptor, for example, as shown in fig. 3, and if 6 screened point cloud units are provided, the point cloud unit 1 is used as a core point cloud unit, and 3 point cloud units are selected from 2 to 6 to construct; then, constructing by taking the point cloud unit 2 as a core point cloud unit and optionally selecting 3 point cloud units from 1, 3-6; and the description is omitted until all the point cloud units are traversed.
Step S7, calculating the Manhattan distance between the candidate three-dimensional structure descriptor and the standard structure descriptor, and if the distance is smaller than a set second threshold, the point cloud unit corresponding to the candidate three-dimensional structure descriptor is the object to be identified.
In this embodiment, the manhattan distance between the three-dimensional structure descriptor and the standard structure descriptor is calculated, and if the manhattan distance is smaller than a second threshold, for example, 30, the object composed of the current core point cloud and the n point cloud units is the object to be identified.
In this embodiment, it is assumed that a point cloud unit corresponding to the component 4 is used as a core point cloud unit, the components 1, 2, and 3 are selected from the components 2 to 6 to construct a candidate three-dimensional structure descriptor, and the manhattan distance between the candidate three-dimensional structure descriptor and the standard structure descriptor is found to be less than 30 through calculation, so that the three-dimensional point cloud composed of the component 4 and the components 1, 2, and 3 is an object to be identified.
It should be noted that, in this embodiment, the weight a1The first threshold value and the second threshold value are designed according to the point cloud units corresponding to all parts in the target real three-dimensional point cloud and the experimental structure, and a good identification effect can be achieved. Meanwhile, it should be noted that the technical solution is not limited to target recognition of 4 components, and has a better recognition capability for a three-dimensional object with a complex structure.
The above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and those skilled in the art can make various corresponding changes and modifications according to the present invention without departing from the spirit and the essence of the present invention, but these corresponding changes and modifications should fall within the protection scope of the appended claims.
Claims (2)
1. A structure descriptor-based complex structure three-dimensional object recognition method is characterized by comprising the following steps:
extracting point cloud units corresponding to all parts in the target real three-dimensional point cloud, and calculating to obtain the size and the center of an enclosure box of the point cloud unit corresponding to each part;
selecting one point cloud unit from the point cloud units corresponding to the components as a core point cloud unit, and calculating the distance from the center of the core point cloud unit to the centers of other point cloud units;
pairing other point cloud units except the core point cloud unit in pairs, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit;
constructing a standard three-dimensional structure descriptor by adopting the size of an enclosure box of a point cloud unit corresponding to each component, the distance from the center of a core point cloud unit to the centers of other point cloud units and an included angle formed by connecting the center of each pair of point cloud units and the center of the core point cloud unit;
dividing a three-dimensional scene to be identified to obtain all point cloud units in the three-dimensional scene, and screening out all point cloud units with the number of points greater than a first threshold;
traversing the screened point cloud units, taking any point cloud unit as a core point cloud unit, optionally selecting n point cloud units in a three-dimensional neighborhood of the core point cloud unit, and constructing a candidate three-dimensional structure descriptor, wherein n is equal to the number of the point cloud units corresponding to each part in the target real three-dimensional point cloud minus one;
and calculating the Manhattan distance between the candidate three-dimensional structure descriptor and the standard structure descriptor, and if the distance is smaller than a set second threshold, the point cloud unit corresponding to the candidate three-dimensional structure descriptor is the object to be identified.
2. The method for identifying a three-dimensional object with a complex structure based on a structure descriptor as claimed in claim 1, wherein the constructing the candidate three-dimensional structure descriptor comprises:
calculating the size and the center of a bounding box of the core point cloud unit and n point cloud units selected in the three-dimensional neighborhood of the core point cloud unit;
calculating the distance from the center of the core point cloud unit to the centers of other point cloud units;
pairing other point cloud units except the core point cloud unit in pairs, and calculating an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit;
and constructing a candidate three-dimensional structure descriptor by adopting the size of the bounding box of each point cloud unit, the distance from the center of the core point cloud unit to the centers of other point cloud units and an included angle formed by connecting the center of each pair of point cloud units with the center of the core point cloud unit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811101377.0A CN109344750B (en) | 2018-09-20 | 2018-09-20 | Complex structure three-dimensional object identification method based on structure descriptor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811101377.0A CN109344750B (en) | 2018-09-20 | 2018-09-20 | Complex structure three-dimensional object identification method based on structure descriptor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109344750A CN109344750A (en) | 2019-02-15 |
CN109344750B true CN109344750B (en) | 2021-10-22 |
Family
ID=65305859
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811101377.0A Active CN109344750B (en) | 2018-09-20 | 2018-09-20 | Complex structure three-dimensional object identification method based on structure descriptor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109344750B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110223297A (en) * | 2019-04-16 | 2019-09-10 | 广东康云科技有限公司 | Segmentation and recognition methods, system and storage medium based on scanning point cloud data |
CN110136181B (en) * | 2019-05-17 | 2021-08-20 | 百度在线网络技术(北京)有限公司 | Method and apparatus for generating information |
CN111174761A (en) * | 2019-12-31 | 2020-05-19 | 中国电建集团河北省电力勘测设计研究院有限公司 | Circular pole tower inclination deformation rapid calculation method based on laser point cloud |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2375212A (en) * | 1999-04-29 | 2002-11-06 | Mitsubishi Electric Inf Tech | Representing and searching for an object using shape |
CN102236794A (en) * | 2010-05-07 | 2011-11-09 | Mv科技软件有限责任公司 | Recognition and pose determination of 3D objects in 3D scenes |
CN103150544A (en) * | 2011-08-30 | 2013-06-12 | 精工爱普生株式会社 | Method and apparatus for object pose estimation |
EP2610783A2 (en) * | 2012-01-02 | 2013-07-03 | Samsung Electronics Co., Ltd | Object recognition method and descriptor for object recognition |
CN106250881A (en) * | 2016-08-25 | 2016-12-21 | 深圳大学 | A kind of target identification method based on three dimensional point cloud and system |
CN107077735A (en) * | 2014-10-28 | 2017-08-18 | 惠普发展公司,有限责任合伙企业 | Three dimensional object is recognized |
CN107273831A (en) * | 2017-06-05 | 2017-10-20 | 苏州大学 | A kind of Three-dimensional target recognition method based on spherical space |
CN107577984A (en) * | 2017-07-17 | 2018-01-12 | 华南理工大学 | A kind of 3D Object identifying layered approach based on significance analysis |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8929602B2 (en) * | 2013-01-31 | 2015-01-06 | Seiko Epson Corporation | Component based correspondence matching for reconstructing cables |
-
2018
- 2018-09-20 CN CN201811101377.0A patent/CN109344750B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2375212A (en) * | 1999-04-29 | 2002-11-06 | Mitsubishi Electric Inf Tech | Representing and searching for an object using shape |
CN102236794A (en) * | 2010-05-07 | 2011-11-09 | Mv科技软件有限责任公司 | Recognition and pose determination of 3D objects in 3D scenes |
CN103150544A (en) * | 2011-08-30 | 2013-06-12 | 精工爱普生株式会社 | Method and apparatus for object pose estimation |
EP2610783A2 (en) * | 2012-01-02 | 2013-07-03 | Samsung Electronics Co., Ltd | Object recognition method and descriptor for object recognition |
CN107077735A (en) * | 2014-10-28 | 2017-08-18 | 惠普发展公司,有限责任合伙企业 | Three dimensional object is recognized |
CN106250881A (en) * | 2016-08-25 | 2016-12-21 | 深圳大学 | A kind of target identification method based on three dimensional point cloud and system |
CN107273831A (en) * | 2017-06-05 | 2017-10-20 | 苏州大学 | A kind of Three-dimensional target recognition method based on spherical space |
CN107577984A (en) * | 2017-07-17 | 2018-01-12 | 华南理工大学 | A kind of 3D Object identifying layered approach based on significance analysis |
Non-Patent Citations (1)
Title |
---|
A method of 3D object recognition and localization in a cloud of points;Jerzy Bielicki et al;《 Journal on Advances in Signal Processing》;20130221;第1-13页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109344750A (en) | 2019-02-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111932688B (en) | Indoor plane element extraction method, system and equipment based on three-dimensional point cloud | |
US20190188856A1 (en) | Systems and methods for block based edgel detection with false edge elimination | |
CN108303037B (en) | Method and device for detecting workpiece surface shape difference based on point cloud analysis | |
WO2021052283A1 (en) | Method for processing three-dimensional point cloud data and computing device | |
US11367195B2 (en) | Image segmentation method, image segmentation apparatus, image segmentation device | |
CN112164145B (en) | Method for rapidly extracting indoor three-dimensional line segment structure based on point cloud data | |
CN109344750B (en) | Complex structure three-dimensional object identification method based on structure descriptor | |
JP2015079505A (en) | Noise identification method and noise identification device of parallax depth image | |
CN107871321B (en) | Image segmentation method and device | |
JP6611681B2 (en) | Method and system for segmenting images | |
CN114359042A (en) | Point cloud splicing method and device, three-dimensional scanner and electronic equipment | |
US20160232420A1 (en) | Method and apparatus for processing signal data | |
CN116721228B (en) | Building elevation extraction method and system based on low-density point cloud | |
WO2024120437A1 (en) | Bottom surface segmentation method and apparatus for three-dimensional point cloud, and electronic device and storage medium | |
CN113111741B (en) | Assembly state identification method based on three-dimensional feature points | |
CN106530336A (en) | Stereo matching algorithm based on color information and graph-cut theory | |
CN113628343B (en) | Three-dimensional grid combination processing method and device and storage medium | |
US11256949B2 (en) | Guided sparse feature matching via coarsely defined dense matches | |
CN113379826A (en) | Method and device for measuring volume of logistics piece | |
CN113592976B (en) | Map data processing method and device, household appliance and readable storage medium | |
KR101178015B1 (en) | Generating method for disparity map | |
Haque et al. | Robust feature-preserving denoising of 3D point clouds | |
Kovacs et al. | Edge detection in discretized range images | |
Guo et al. | An efficient voxel-based segmentation algorithm based on hierarchical clustering to extract lidar power equipment data in transformer substations | |
KR101220003B1 (en) | Generating method for disparity map |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |