CN113012161B - Stacked scattered target point cloud segmentation method based on convex region growth - Google Patents
Stacked scattered target point cloud segmentation method based on convex region growth Download PDFInfo
- Publication number
- CN113012161B CN113012161B CN202110247203.0A CN202110247203A CN113012161B CN 113012161 B CN113012161 B CN 113012161B CN 202110247203 A CN202110247203 A CN 202110247203A CN 113012161 B CN113012161 B CN 113012161B
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- neighborhood
- curvature
- seed
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 50
- 239000013598 vector Substances 0.000 claims abstract description 63
- 230000000694 effects Effects 0.000 claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 6
- 238000009499 grossing Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention provides a stacking scattered target point cloud segmentation method based on convex region growth, which mainly comprises the following steps: smoothing stacked scattered target point clouds acquired by a depth camera by adopting a mobile least square algorithm, solving normal vector and curvature information of all points in the point clouds based on a principal component analysis method for the smoothed point clouds, and finally clustering according to the normal vector, curvature and concave-convex characteristics of the point clouds to realize point cloud segmentation. The point cloud segmentation algorithm only depends on the characteristics of each point in the point cloud, can realize the good segmentation effect of stacking scattered target point clouds without training a target model in advance, has higher operation efficiency, and has good adaptability and robustness for point cloud segmentation in different scenes.
Description
Technical Field
The invention belongs to the technical field of three-dimensional point cloud processing, and particularly relates to a stacked scattered target point cloud segmentation method based on convex region growth.
Background
Along with the continuous popularization of depth cameras and the continuous development of three-dimensional vision technologies, the three-dimensional point cloud processing technology plays an increasingly important role in the fields of intelligent robot sorting, three-dimensional vision detection and the like. For the intelligent sorting task of a robot for randomly stacking workpieces or randomly placing objects in a living scene in an industrial field, the conditions of uneven density, unstructured data and the like possibly exist in the three-dimensional point cloud data of the targets obtained by adopting a depth camera, and the target point clouds are mutually adhered to each other, so that the follow-up treatment is not facilitated, and the task of dividing the point cloud is more difficult due to the random stacking among the targets.
The point cloud segmentation is an important ring in the three-dimensional point cloud processing technology, and the accuracy of links such as subsequent point cloud identification and registration is determined by the quality of a point cloud segmentation algorithm. For the sparse distribution of the target point cloud segmentation, each target point cloud can be easily segmented according to Euclidean distance characteristics among targets, and for the scattered and stacked target point cloud segmentation, single target point cloud segmentation is relied onThe attribute features do not achieve good segmentation results. The traditional point cloud segmentation method based on region growth judges the similarity between adjacent points according to the neighborhood information of the points, clusters the points with similar attributes into a region, so that the points with larger differences are segmented in different regions, the method is insensitive to noise points and abnormal points, and has simple principle and high operation speed, but the method depends on the selection of seed points in the point cloud and the region growth rule, so that the conditions of over segmentation and under segmentation are easy to occur. Stein et al (S.C.Stein, M.Schoeler, J.Papon and F)."Object Partitioning Using Local Convexity,"2014 IEEE Conference on Computer Vision and Pattern Recognition,Columbus,OH,2014,pp.304-311.) improves the idea of super-voxelization of the point cloud, and provides a point cloud segmentation method for clustering by utilizing the concave-convex property between adjacent super-voxels.
Disclosure of Invention
Aiming at the defects and problems of the prior art, the invention provides a stacking scattered target point cloud segmentation method based on the growth of a convex region, which solves the problems of low operation efficiency and non-ideal segmentation effect in the prior stacking scattered target point cloud segmentation method, carries out normal vector and curvature estimation on smoothed point cloud to obtain more accurate normal vector and curvature information, then carries out clustering according to normal vector, curvature and concave-convex characteristics between neighborhood points, increases concave-convex judgment conditions compared with the point cloud segmentation method based on the growth of the region, avoids the over-segmentation and under-segmentation of the point cloud, does not need to carry out super-voxel clustering in advance compared with the point cloud segmentation method based on super-voxel concave-convex, improves the point cloud segmentation efficiency, and can obtain more ideal segmentation effect on the premise of guaranteeing the segmentation efficiency.
The invention is realized at least by one of the following technical schemes.
A stacking scattered target point cloud segmentation method based on convex region growth comprises the following steps:
step 1, fitting point cloud to obtain smoothed three-dimensional point cloud data;
step 2, solving normal vector information and curvature information of the point cloud for the three-dimensional point cloud data smoothed in the step 1;
step 3, ordering the points in the point cloud according to the curvature, and judging whether the seed points and the neighborhood points belong to the same area or not by taking the point with the minimum curvature as the seed point;
and 4, marking the points of the same area as the same target, then selecting a seed point from unlabeled points to perform the operation of the step 3 until all points are marked after traversing the whole point cloud, and finishing the point cloud segmentation.
Preferably, in step 1, the point cloud is re-fitted by using a mobile least squares algorithm.
Preferably, the moving least square algorithm calculates fitting function f (t) fitting stacked scattered target three-dimensional point cloud data according to a moving least square method principle, achieves the effect of point cloud smoothing, and establishes the following objective function to solve f (t):
wherein t= [ x, y ]]Is a two-dimensional vector consisting of x and y coordinates of point cloud data, called a node; n represents the number of nodes used for fitting, i represents the nodes; z is the z coordinate of the point cloud data, called the node value; omega (. Cndot.) is t i Weight function of t-t i Is the distance between the nodes; min represents the minimum of the objective function.
Preferably, in the step 2, a principal component analysis method is adopted to estimate the normal vector and curvature information of the point cloud.
Preferably, the step 2 specifically comprises: for any point P in the point cloud, searching a neighboring point set of the point P, constructing a covariance matrix through the neighboring point set, wherein a feature vector corresponding to a minimum feature value of the covariance matrix is a normal vector at the point P, and constructing the covariance matrix as follows:
wherein k is the number of adjacent points; p (P) i Representing the spatial three-dimensional coordinates of the ith adjacent point;representing three-dimensional centroid coordinates of a set of neighboring points;
the calculation formula of the curvature at point P is as follows:
wherein lambda is 1 、λ 2 、λ 3 Respectively represent three eigenvalues of covariance matrix, and lambda 3 Is the minimum eigenvalue.
Preferably, the neighbor point set of the point P is searched by using kd-tree.
Preferably, in the step 3, whether the seed point and the neighborhood point belong to the same region is determined according to the normals, curvature and concave-convex characteristics of the seed point and the neighborhood point.
Preferably, step 3 comprises the steps of:
1) Setting an empty queue Q for storing seed sequences and an empty cluster array L for storing point cloud data of the same area, and selecting a point with the minimum curvature as a seed point to be placed in the Q;
2) Calculating a normal angle theta between the seed point and the neighborhood point:
according to the method of calculating normal vector at any point in the point cloud in the step 2, namely calculating the feature vector corresponding to the minimum feature value of the covariance matrix of the neighborhood point set as the pointThe normal vector is respectively marked as the normal vector of the seed point and one neighborhood pointAnd->Then->And->The normal vector included angle formula between the two is:
3) Judging whether the included angle theta is smaller than a set included angle threshold value theta threshold ;
4) If the included angle theta in the step 3) is smaller than the set included angle threshold value, then calculating whether the curvature sigma of the neighborhood point is smaller than the curvature threshold value sigma threshold If the curvature sigma of the neighborhood point is smaller than the set curvature threshold sigma threshold Adding the neighborhood point into Q as seed point if curvature sigma of the neighborhood point is larger than set curvature threshold sigma threshold Adding the neighborhood point into L;
5) Judging the concave-convex information of the seed point and the neighborhood point if the included angle theta in the step 3) is larger than the set included angle threshold value, and adding the neighborhood point into L if the seed point and the neighborhood point are convex;
6) After the neighborhood point set is checked, deleting the current seed point, repeating the processes from the step 2) to the step 5) until Q is empty, and marking the points in L as the same area.
Preferably, the concave-convex information discrimination formula of the seed point and the neighborhood point is as follows:
wherein:normal vector representing seed point, +.>Normal vector representing neighborhood point, ++>Vector of three-dimensional coordinates representing seed point, < ->Vector beta representing space three-dimensional coordinates of neighborhood point threshold And d represents a unit vector of the space coordinate vector difference between the seed point and the neighborhood point, and beta represents a normal vector included angle between the seed point and the neighborhood point.
Preferably, in the step 4, marking the points of the same area as the same target specifically includes: and setting the same label for the points in the same area, and distinguishing each target after segmentation according to different label values.
Compared with the prior art, the invention has the following advantages:
(1) According to the point cloud segmentation method, a target model does not need to be trained in advance, point cloud segmentation is performed completely according to characteristic information among neighbor point sets in point cloud data, and the method has strong adaptability and robustness to most stacked target point cloud segmentation.
(2) The method carries out normal estimation on the smoothed point cloud data, has a more accurate normal estimation value, adds the concave-convex judging condition between the neighborhood point sets in the point cloud segmentation process, largely avoids the over-segmentation and under-segmentation conditions, and improves the accuracy of the segmentation result on the premise of ensuring the segmentation efficiency.
Drawings
FIG. 1 is a flow chart of a stacked scattered target point cloud segmentation method based on convex region growth according to the invention;
FIG. 2 is a stacked scattered target point cloud to be segmented in an example;
FIG. 3 is a point cloud diagram containing normal vectors after the normal vectors are obtained by adopting a principal component analysis method in an example;
FIG. 4 is a flow chart of point cloud segmentation based on normal vector, curvature, and asperity features of the point cloud;
FIG. 5 is a graph of the effect of cloud segmentation of stacked scattered target points using convex region-based growth in this example;
FIG. 6 is a graph of the effect of point cloud segmentation based on region growth in the prior art;
fig. 7 is a graph of the effect of point cloud segmentation based on super-voxel convexity in the prior art method.
Detailed Description
The invention will be further described with reference to the accompanying drawings and examples.
The invention discloses a stacking scattered target point cloud segmentation method based on convex region growth, which segments the stacking scattered target point cloud according to normal vector, curvature and concave-convex information of a three-dimensional point cloud to obtain point cloud data of each target, and is convenient for subsequent implementation of the steps of point cloud identification, registration and the like.
Fig. 1 is a flowchart of an implementation of a stacked scattered target point cloud segmentation method based on convex region growth according to the present invention, and fig. 2 is a stacked scattered target point cloud image to be segmented in this example, and the implementation steps are as follows:
in step 1, some noise points and invalid points may exist in three-dimensional point cloud data acquired by a depth camera, in order to ensure accuracy of normal vector and curvature estimation of each point in subsequent point clouds, influence of the noise points and the invalid points is reduced, a moving least square method is adopted to re-fit and stack scattered target point clouds, and smooth point cloud data are obtained.
Specifically, fitting function f (t) fitting stacking scattered target three-dimensional point cloud data is obtained according to a moving least square method principle, the effect of point cloud smoothing is achieved, and the following objective function is established to solve f (t):
wherein t= [ x, y ]]Is a two-dimensional vector consisting of x and y coordinates of point cloud data, called a node; n represents the number of nodes used for fitting, i represents the nodes; z is the z coordinate of the point cloud data, called the node value; omega (. Cndot.) is t i Is not unique, and can be selected according to requirements, such as cubic spline weight function), t-t i Is the distance between the nodes; min represents the minimum of the objective function.
As another embodiment, the point cloud data acquired by the depth camera may be acquired instead of other devices, such as by stereo matching with a binocular camera.
As another embodiment, the moving least square method may be replaced by a least square method for point cloud smoothing.
And 2, estimating normal vector and curvature information of the point cloud by adopting a principal component analysis method for the smoothed data obtained in the step 1, wherein the method comprises the following specific steps of: for any point P in the point cloud, a kdtree is adopted to search a neighbor point set of the point P, a covariance matrix is constructed through the neighbor point set, a feature vector corresponding to the minimum feature value of the covariance matrix is a normal vector at the point P, curvature at the point P is estimated according to three feature values of the covariance matrix, normal vectors and curvature information of all points in the point cloud can be obtained in the same way, and a point cloud diagram containing the normal vector is shown in figure 3.
The covariance matrix is constructed as follows:
wherein k is the number of adjacent points; p (P) i Representing the spatial three-dimensional coordinates of the ith adjacent point;representing three-dimensional centroid coordinates of a set of neighboring points;
the calculation formula of the curvature at point P is as follows:
wherein lambda is 1 、λ 2 、λ 3 Respectively represent three eigenvalues of covariance matrix, and lambda 3 Is the minimum eigenvalue.
As another embodiment, the principal component analysis method may be replaced with a Delaunay/Voronoi-based method or a robust statistics-based method for point cloud normal vector estimation.
As another example, kdtree searching may be replaced with octree searching for neighbor points.
And 3, performing point cloud segmentation according to the normal vector, curvature and concave-convex characteristics of the point cloud, wherein a segmentation flow chart is shown in fig. 4, and the specific implementation steps are as follows:
1) Ordering all points in the point cloud from small to large according to curvature, setting an empty seed sequence Q and an empty cluster array L, adding points with the minimum curvature as seed points into the Q, and searching neighborhood points of the seed points through a kd tree;
2) If the normal angle between the seed point and the neighborhood point is smaller than the set angle threshold value theta threshold Continuing to judge the curvature of the neighborhood point, if the curvature of the neighborhood point is smaller than the set curvature threshold sigma threshold And adding the neighborhood point as a new seed point into Q, otherwise adding the neighborhood point into L.
Let the normal vectors of the seed point and a neighborhood point be respectively recorded asAnd->Then->And->The normal vector included angle formula between the two is:
3) If the normal angle between the seed point and the neighborhood point is larger than the set angle threshold, judging the concave-convex information of the seed point and the neighborhood point, and if the seed point and the neighborhood point are convex, adding the neighborhood point into L.
The concave-convex information discrimination formula of the seed point and the neighborhood point is as follows:
wherein:normal vector representing seed point, +.>Normal vector representing neighborhood point, ++>Vector representing the spatial three-dimensional coordinates of seed points,/>Vector beta representing space three-dimensional coordinates of neighborhood point threshold And d represents a unit vector of the space coordinate vector difference between the seed point and the neighborhood point, and beta represents a normal vector included angle between the seed point and the neighborhood point.
4) After the neighborhood point set is checked, deleting the current seed point, repeating the processes from the step 2) to the step 3) until Q is empty, and marking the points in L as the same area.
And 4, marking the points in the same area as the same target, then selecting a seed point from unlabeled points to perform the operation of the step 3 until all points are marked after traversing the whole point cloud, and finishing the point cloud segmentation, wherein the segmentation effect is shown in fig. 5.
The method provided by the invention belongs to a point cloud segmentation method in the technical field of three-dimensional point cloud processing, can be applied to intelligent sorting of robots for stacking scattered targets in an industrial scene, and can be used for obtaining the point cloud of a single target by segmenting the stacked scattered target point cloud, so that the subsequent identification and pose estimation of the single target are facilitated, and the targets are sorted unordered by the robots. The conventional point cloud segmentation method has the defects of low segmentation precision, low operation efficiency and the like for the stacked scattered target point cloud segmentation, the method obviously improves the defects, and an ideal point cloud segmentation effect can be obtained under higher operation efficiency.
Comparing the method of the invention with the existing point cloud segmentation method, fig. 6 and 7 are respectively a point cloud segmentation effect diagram based on region growth and a point cloud segmentation effect diagram based on super-voxel convexity. The comparison of the three methods shows that the point cloud segmentation algorithm based on the region growth has obvious under segmentation and over segmentation conditions, but the operation efficiency is highest; some underspection targets exist in the point cloud segmentation algorithm based on the super-voxel convexity, the overall segmentation effect is better than that of the point cloud segmentation algorithm based on the region growth, but the operation efficiency is the lowest; the point cloud segmentation effect of the method is superior to that of the first two methods, the operation efficiency is centered, and the operation efficiency pairs of the three methods are shown in a table 1.
Table 1 comparative table of the operating efficiency of the process of the invention versus the prior art process
In order to better understand the aspects of the present invention, the present invention will be described in further detail with reference to the accompanying drawings and detailed description. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Claims (1)
1. The stacked scattered target point cloud segmentation method based on the convex region growth is characterized by comprising the following steps of:
step 1, re-fitting the point cloud by adopting a mobile least square algorithm to obtain smoothed three-dimensional point cloud data;
the moving least square algorithm obtains fitting function f (t) fitting stacking scattered target three-dimensional point cloud data according to a moving least square method principle, achieves the effect of point cloud smoothing, and establishes the following objective function to solve f (t):
wherein t= [ x, y ]]Is a two-dimensional vector consisting of x and y coordinates of point cloud data, called a node; n represents the number of nodes used for fitting, i represents the nodes; z is the z coordinate of the point cloud data, called the node value; omega (. Cndot.) is t i Weight function of t-t i Is the distance between the nodes; min represents the minimum value of the objective function;
step 2, solving normal vector information and curvature information of the point cloud for the three-dimensional point cloud data smoothed in the step 1;
in the step 2, the principal component analysis method is adopted to estimate the normal vector and curvature information of the point cloud, and the method specifically comprises the following steps: for any point P in the point cloud, searching a neighboring point set of the point P, constructing a covariance matrix through the neighboring point set, wherein a feature vector corresponding to a minimum feature value of the covariance matrix is a normal vector at the point P, and constructing the covariance matrix as follows:
wherein k is the number of adjacent points; p (P) i Representing the spatial three-dimensional coordinates of the ith adjacent point;representing three-dimensional centroid coordinates of a set of neighboring points;
the calculation formula of the curvature at point P is as follows:
wherein lambda is 1 、λ 2 、λ 3 Respectively represent three eigenvalues of covariance matrix, and lambda 3 Is the minimum characteristic value;
the adjacent point set of the point P is searched by using a kd tree;
step 3, ordering the points in the point cloud according to the curvature, judging whether the seed point and the neighborhood point belong to the same area according to the normals of the seed point and the neighborhood point, the curvature and the concave-convex characteristic, and judging whether the seed point and the neighborhood point belong to the same area by taking the point with the minimum curvature as the seed point; the method comprises the following steps:
1) Setting an empty queue Q for storing seed sequences and an empty cluster array L for storing point cloud data of the same area, and selecting a point with the minimum curvature as a seed point to be placed in the Q;
2) Calculating a normal angle theta between the seed point and the neighborhood point:
according to the method for calculating normal vector at any point in the point cloud in the step 2, namely calculating the feature vector corresponding to the minimum feature value of the covariance matrix of the neighborhood point set, taking the feature vector as the normal vector at the point, and respectively marking the normal vector of the seed point and one neighborhood point asAnd->Then->And->The normal vector included angle formula between the two is:
3) Judging whether the included angle theta is smaller than a set included angle threshold value theta threshold ;
4) If the included angle theta in the step 3) is smaller than the set included angle threshold value, then calculating whether the curvature sigma of the neighborhood point is smaller than the curvature threshold value sigma threshold If the curvature sigma of the neighborhood point is smaller than the set curvature threshold sigma threshold Adding the neighborhood point into Q as seed point if curvature sigma of the neighborhood point is larger than set curvature threshold v threshold Adding the neighborhood point into L;
5) Judging the concave-convex information of the seed point and the neighborhood point if the included angle theta in the step 3) is larger than the set included angle threshold value, and adding the neighborhood point into L if the seed point and the neighborhood point are convex;
6) After the neighborhood point set is checked, deleting the current seed point, repeating the processes from the step 2) to the step 5) until Q is empty, and marking the points in L as the same area;
the concave-convex information discrimination formula of the seed point and the neighborhood point is as follows:
wherein:normal vector representing seed point, +.>Normal vector representing neighborhood point, ++>Vector of three-dimensional coordinates representing seed point, < ->Vector beta representing space three-dimensional coordinates of neighborhood point threshold For a set vector angle threshold value,a unit vector representing the difference of the space coordinate vector of the seed point and the neighborhood point, and beta represents the normal vector included angle of the seed point and the neighborhood point;
and 4, marking the points of the same area as the same target, then selecting a next seed point from unlabeled points to perform the operation of the step 3 until all points are marked after traversing the whole point cloud, and marking the points of the same area as the same target specifically comprises the following steps: and setting the same label for the points in the same area, and distinguishing each target after segmentation according to different label values.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110247203.0A CN113012161B (en) | 2021-03-05 | 2021-03-05 | Stacked scattered target point cloud segmentation method based on convex region growth |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110247203.0A CN113012161B (en) | 2021-03-05 | 2021-03-05 | Stacked scattered target point cloud segmentation method based on convex region growth |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113012161A CN113012161A (en) | 2021-06-22 |
CN113012161B true CN113012161B (en) | 2023-04-28 |
Family
ID=76407389
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110247203.0A Active CN113012161B (en) | 2021-03-05 | 2021-03-05 | Stacked scattered target point cloud segmentation method based on convex region growth |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113012161B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113469968B (en) * | 2021-06-29 | 2022-11-04 | 湖北中烟工业有限责任公司 | Method and device for detecting height difference of box seams of cigarette packaging box and electronic equipment |
CN115330819B (en) * | 2022-10-12 | 2023-02-10 | 杭州蓝芯科技有限公司 | Soft package segmentation positioning method, industrial personal computer and robot grabbing system |
CN115908425B (en) * | 2023-02-14 | 2023-06-30 | 四川大学 | Edge detection-based rock-fill grading information detection method |
CN117291936B (en) * | 2023-11-27 | 2024-02-20 | 法奥意威(苏州)机器人系统有限公司 | Point cloud segmentation method, device, equipment and medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107369161A (en) * | 2017-07-19 | 2017-11-21 | 无锡信捷电气股份有限公司 | A kind of workpiece point cloud segmentation method at random based on the European cluster of improvement |
CN109409437A (en) * | 2018-11-06 | 2019-03-01 | 安徽农业大学 | A kind of point cloud segmentation method, apparatus, computer readable storage medium and terminal |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108510516A (en) * | 2018-03-30 | 2018-09-07 | 深圳积木易搭科技技术有限公司 | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud |
CN109859208A (en) * | 2019-01-03 | 2019-06-07 | 北京化工大学 | Scene cut and Target Modeling method based on concavity and convexity and RSD feature |
CN111091062B (en) * | 2019-11-21 | 2023-07-04 | 东南大学 | Robot out-of-order target sorting method based on 3D visual clustering and matching |
-
2021
- 2021-03-05 CN CN202110247203.0A patent/CN113012161B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107369161A (en) * | 2017-07-19 | 2017-11-21 | 无锡信捷电气股份有限公司 | A kind of workpiece point cloud segmentation method at random based on the European cluster of improvement |
CN109409437A (en) * | 2018-11-06 | 2019-03-01 | 安徽农业大学 | A kind of point cloud segmentation method, apparatus, computer readable storage medium and terminal |
Also Published As
Publication number | Publication date |
---|---|
CN113012161A (en) | 2021-06-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113012161B (en) | Stacked scattered target point cloud segmentation method based on convex region growth | |
CN112070818B (en) | Robot disordered grabbing method and system based on machine vision and storage medium | |
CN111091062B (en) | Robot out-of-order target sorting method based on 3D visual clustering and matching | |
CN107369161B (en) | Scattered workpiece point cloud segmentation method based on improved Euclidean clustering | |
CN108648233B (en) | Target identification and capture positioning method based on deep learning | |
CN111299815B (en) | Visual detection and laser cutting trajectory planning method for low-gray rubber pad | |
CN105046197B (en) | Multi-template pedestrian detection method based on cluster | |
JP6216508B2 (en) | Method for recognition and pose determination of 3D objects in 3D scenes | |
CN111553409B (en) | Point cloud identification method based on voxel shape descriptor | |
CN108537239B (en) | Method for detecting image saliency target | |
CN109034065B (en) | Indoor scene object extraction method based on point cloud | |
Hagelskjær et al. | Pointvotenet: Accurate object detection and 6 dof pose estimation in point clouds | |
CN109035207B (en) | Density self-adaptive laser point cloud characteristic detection method | |
CN114972377A (en) | 3D point cloud segmentation method and device based on moving least square method and hyper-voxels | |
CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
CN104715251A (en) | Salient object detection method based on histogram linear fitting | |
Czajewski et al. | 3D object detection and recognition for robotic grasping based on RGB-D images and global features | |
CN104915951B (en) | A kind of stippled formula DPM two-dimension code area localization methods | |
CN114492619A (en) | Point cloud data set construction method and device based on statistics and concave-convex property | |
Ecins et al. | Cluttered scene segmentation using the symmetry constraint | |
Pla et al. | Matching feature points in image sequences through a region-based method | |
CN113409332B (en) | Building plane segmentation method based on three-dimensional point cloud | |
CN116416305B (en) | Multi-instance pose estimation method based on optimized sampling five-dimensional point pair characteristics | |
Liu et al. | Deep learning of directional truncated signed distance function for robust 3D object recognition | |
Borisagar et al. | A novel segment-based stereo matching algorithm for disparity map generation |
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 |