CN112308889A - Point cloud registration method and storage medium by utilizing rectangle and oblateness information - Google Patents
Point cloud registration method and storage medium by utilizing rectangle and oblateness information Download PDFInfo
- Publication number
- CN112308889A CN112308889A CN202011146501.2A CN202011146501A CN112308889A CN 112308889 A CN112308889 A CN 112308889A CN 202011146501 A CN202011146501 A CN 202011146501A CN 112308889 A CN112308889 A CN 112308889A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- points
- point
- frame
- value
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
Abstract
The invention discloses a point cloud registration method and a storage medium by utilizing rectangle and oblateness information, wherein the method comprises the following steps: acquiring first frame point cloud data, and clustering the first frame point cloud data to obtain point cloud segments; acquiring a flat rate value of the point cloud segment; performing rectangle fitting on the point cloud segment to obtain a rectangular structure, and acquiring four vertex coordinates of the rectangular structure; and acquiring second frame point cloud data, and registering the first frame point cloud data and the second frame point cloud data according to the four vertex coordinates of the rectangular structure and the oblateness value of the point cloud segment. According to the method, the first frame of point cloud data is segmented by using a clustering method, so that the first frame of point cloud data can be increased according to a certain sequence, and then the flatness information of the segmented point cloud segments is acquired to match the two frames of point cloud data, so that the structure of the point cloud segments is accurately expressed by the flatness information, and the stability and reliability of point cloud registration in scene transformation are improved.
Description
Technical Field
The invention relates to the field of point cloud data, in particular to a point cloud registration method and a storage medium by utilizing rectangle and oblateness information.
Background
Point cloud registration is a fundamental and important topic in three-dimensional computer vision. The method is widely applied to three-dimensional reconstruction, instant positioning and mapping, automatic driving and the like. However, there are many obstacles in practical use. The scholars have done a lot of work on obstacles such as point cloud sparse change, occlusion and partial overlapping, but research on scene transformation is still rare. The existing point cloud registration method for two frames of point cloud data in scene transformation is low in stability and reliability, and a method capable of stably and accurately registering the two frames of point cloud data in the scene transformation is lacked.
Thus, there is still a need for improvement and development of the prior art.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a point cloud registration method and a storage medium using rectangle and flat rate information, aiming at solving the problem that the prior art lacks a method for performing stable and accurate registration on two frames of point cloud data in scene transformation.
The technical scheme adopted by the invention for solving the problems is as follows:
in a first aspect, an embodiment of the present invention provides a point cloud registration method using rectangle and ellipticity information, where the method includes:
acquiring first frame point cloud data, and clustering the first frame point cloud data to obtain point cloud segments;
acquiring a flat rate value of the point cloud segment;
performing rectangle fitting on the point cloud segment to obtain a rectangular structure, and acquiring four vertex coordinates of the rectangular structure;
and acquiring second frame point cloud data, and registering the first frame point cloud data and the second frame point cloud data according to the four vertex coordinates of the rectangular structure and the oblateness value of the point cloud segment.
In one embodiment, the obtaining the first frame of point cloud data and clustering the first frame of point cloud data to obtain the point cloud segment includes:
acquiring first frame point cloud data, and taking a preset radius as a sphere in the first frame point cloud data to obtain a test sphere;
obtaining points in the test sphere to obtain neighborhood points, and taking the neighborhood points as the sphere center and the preset radius as the sphere to obtain a neighborhood sphere;
acquiring the number of points in the neighborhood sphere, the oblateness value of the point set in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and clustering the first frame of point cloud data according to the oblateness value of the point set and the projection distance of the number of the points to obtain a point cloud segment.
In an embodiment, the obtaining the number of points in the neighborhood sphere, the oblateness value of the point set in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and clustering the first frame of point cloud data according to the oblateness value of the point set and the projection distance of the number of the points to obtain the point cloud segment includes:
acquiring the number of points and the coordinates of the points in the neighborhood sphere, and acquiring the oblateness value of the point set in the neighborhood sphere according to the number of the points and the coordinates of the points in the neighborhood sphere;
acquiring the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame point cloud data;
respectively comparing the number of points in the neighborhood sphere, the projection distance, and the oblateness value of the point set in the neighborhood sphere with a point number threshold, a projection distance threshold, and a oblateness value threshold of the point set;
and when the number of the points in the neighborhood sphere, the projection distance and the oblateness value of the point set in the neighborhood sphere are respectively in the point number threshold, the projection distance threshold and the oblateness value threshold of the point set, clustering the neighborhood points and the test points to obtain point cloud segments.
In one embodiment, the obtaining the number of points and the coordinates of the points in the neighborhood sphere, and obtaining the ellipticity value of the point set in the neighborhood sphere according to the number of points and the coordinates of the points in the neighborhood sphere includes:
acquiring the number of points and the coordinates of the points in the neighborhood sphere, and acquiring the arithmetic mean coordinates in the neighborhood sphere according to the number of the points and the coordinates of the points in the neighborhood sphere;
obtaining a first scattering matrix corresponding to the neighborhood sphere according to the number of the points in the neighborhood sphere, the coordinates of the points and the arithmetic mean coordinate;
and acquiring a first eigenvalue and a third eigenvalue of the first scatter matrix, and obtaining the ellipticity value of the point set in the neighborhood sphere according to the first eigenvalue and the third eigenvalue of the first scatter matrix.
In one embodiment, the obtaining the ellipticity value of the point cloud segment includes:
acquiring the number of points and the coordinates of the points in the point cloud segment, and acquiring arithmetic mean coordinates according to the number of the points and the coordinates of the points in the point cloud segment;
obtaining a second scatter matrix corresponding to the point cloud segment according to the number of the points in the point cloud segment, the coordinates of the points and the arithmetic mean value coordinate;
and acquiring a first eigenvalue and a third eigenvalue of the second scatter matrix, and obtaining the flatness value of the point cloud segment according to the first eigenvalue and the third eigenvalue of the second scatter matrix.
In one embodiment, the performing a rectangular fitting on the point cloud segment to obtain a rectangular structure, and obtaining coordinate data of coordinates of four vertices of the rectangular structure includes:
performing rectangle fitting on the point cloud segment to obtain a rectangular structure;
acquiring a second eigenvalue of the second dispersion matrix;
and obtaining coordinate data of four vertexes of the rectangular structure according to the coordinates of the points in the point cloud segment, the arithmetic mean value coordinates, the first characteristic value and the second characteristic value.
In one embodiment, the acquiring the second frame of point cloud data, and the registering the first frame of point cloud data and the second frame of point cloud data according to the coordinate data of the four vertices of the rectangular structure and the ellipticity value of the point cloud segment includes:
acquiring second frame point cloud data and six-degree-of-freedom transformation quantity in an iteration process;
obtaining coordinate data of a transformation point under the coordinate of the first frame of point cloud data generated based on a point under the coordinate system of the second frame of point cloud data according to the six-degree-of-freedom variation, and determining a point cloud segment corresponding to the transformation point;
obtaining a target pose transformation matrix according to the flatness value of the point cloud segment, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and the coordinate data of the transformation point;
and registering the second frame of point cloud data and the first frame of point cloud data according to the target pose transformation matrix.
In one embodiment, the obtaining, according to the six-degree-of-freedom variation, coordinate data of a transformation point in coordinates of the first frame point cloud data generated based on a point in a coordinate system of the second frame point cloud data, and determining a point cloud segment corresponding to the transformation point includes:
obtaining a pose transformation matrix according to the six-degree-of-freedom variation;
performing rigid body transformation on points under the coordinate system of the second frame of point cloud data according to the pose transformation matrix to obtain transformation points under the coordinates of the first frame of point cloud data;
and acquiring coordinate data of the transformation points, searching a point cloud segment with the minimum distance value with the transformation points according to the coordinate data of the transformation points, and obtaining the point cloud segment corresponding to the transformation points.
In one embodiment, the obtaining a target pose transformation matrix according to the flatness value of the point cloud segment, the coordinate data of four vertices of the rectangular structure corresponding to the point cloud segment, and the coordinate data of the transformation point includes:
obtaining a square distance value according to the coordinate data of the transformation point, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and a distance square function;
obtaining a likelihood function value of a likelihood function generated based on the pose transformation matrix according to the ellipticity value and the square distance value of the point cloud fragment;
and acquiring a pose transformation matrix input into the likelihood function when the likelihood function value is the maximum value to obtain a target pose transformation matrix.
In a second aspect, an embodiment of the present invention further provides a storage medium having a plurality of instructions stored thereon, where the instructions are adapted to be loaded and executed by a processor to implement any one of the above steps of a point cloud registration method using rectangle and ellipticity information.
The invention has the beneficial effects that: according to the method and the device, the first frame of point cloud data is segmented by using a clustering method, so that the first frame of point cloud data can be increased according to a certain sequence, and then the flatness information of the segmented point cloud segments is acquired to match the two frames of point cloud data, so that the structure of the point cloud segments is accurately expressed by the flatness information, and the stability and the reliability of point cloud registration in scene transformation are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a point cloud registration method using rectangle and ellipticity information according to an embodiment of the present invention.
Fig. 2 is a schematic flow chart of acquiring a point cloud segment according to an embodiment of the present invention.
Fig. 3 is a schematic flowchart of acquiring a flatness value of a point cloud segment according to an embodiment of the present invention.
Fig. 4 is a schematic flowchart of acquiring coordinates of four vertices of a rectangular structure according to an embodiment of the present invention.
FIG. 5 is a schematic diagram of a rectangular structure provided by an embodiment of the present invention;
fig. 6 is a schematic flowchart of the registration of the first frame point cloud data and the second frame point cloud data according to the embodiment of the present invention.
Fig. 7 is a reference diagram for calculating a square value of a distance from a transformation point to a rectangular structure according to an embodiment of the present invention.
Fig. 8 is a schematic diagram of two partially overlapped frames of point cloud data according to an embodiment of the present invention.
Fig. 9 is a diagram illustrating an effect of registration of two frames of partially overlapped point cloud data according to the method of the present invention.
Fig. 10 is a schematic diagram of two frames of partially overlapped point cloud data after adding outliers and noise according to an embodiment of the present invention.
Fig. 11 is an effect diagram after registration of two frames of point cloud data after outliers and noise are added according to the embodiment of the present invention.
Fig. 12 is a functional block diagram of a terminal according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
It should be noted that, if directional indications (such as up, down, left, right, front, and back … …) are involved in the embodiment of the present invention, the directional indications are only used to explain the relative positional relationship between the components, the movement situation, and the like in a specific posture (as shown in the drawing), and if the specific posture is changed, the directional indications are changed accordingly.
Point cloud registration is a fundamental and important topic in three-dimensional computer vision. The method is widely applied to three-dimensional reconstruction, instant positioning and mapping, automatic driving and the like. However, there are many obstacles in practical use. The scholars have done a lot of work on obstacles such as point cloud sparse change, occlusion and partial overlapping, but research on scene transformation is still rare. Scene transformation brings great challenges to point cloud registration, and the reasons for registration failure in the prior art include the following two aspects:
(1) line/plane characteristics are scene critical. The pairing of points with dots is often unreliable because there are not exactly identical pairs of points in an actual laser scan and such pairing is susceptible to noise spots. Pairing points with lines may alleviate the problems with point pairs, but only performs well in two-dimensional registration. In order to be popularized to the three-dimensional registration problem, the pairing of the point and the surface takes the plane normal vector and the curvature information of the scanning line into consideration, so that the registration precision is improved. However, the point-surface pairing is very dependent on the precision of the local normal vector, and the precision of the normal vector is limited by the sparsity of the point cloud.
(2) A feature expression based on the point distribution may be trapped in false assumptions. For example, the prior art cuts the three-dimensional space into a grid structure, assuming that the set of points within the grid is bell-shaped, fitting to a normal distribution. But such distribution assumptions are often inaccurate and susceptible to grid size. Too small a mesh reduces the convergence domain of the algorithm, while too large a mesh blurs the details, reducing the registration accuracy.
In view of the lack of a point cloud registration method in application and scene transformation in the prior art, the invention provides a point cloud registration method using rectangle and ellipticity information, which can truly express the structure of a point cloud through the rectangle and ellipticity information, thereby accurately realizing the registration between two frames of point cloud data.
As shown in fig. 1, the method comprises the steps of:
s100, acquiring first frame point cloud data, and clustering the first frame point cloud data to obtain a point cloud segment.
Specifically, in this embodiment, first frame point cloud data that needs to be registered is obtained, and the first frame point cloud data is clustered according to a preset clustering algorithm, where the clustering algorithm includes a specific constraint condition, so that the first frame point cloud data is cut into different point cloud segments.
In one implementation, as shown in fig. 2, the step S100 specifically includes the following steps:
step S110, acquiring first frame point cloud data, and taking a preset radius as a sphere in the first frame point cloud data to obtain a test sphere;
step S120, obtaining points in the test sphere to obtain neighborhood points, and taking the neighborhood points as the sphere center and the preset radius as the sphere to obtain a neighborhood sphere;
step S130, acquiring the number of points in the neighborhood sphere, the oblateness value of the point set in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and clustering the first frame of point cloud data according to the oblateness value of the point set and the projection distance of the number of the points to obtain a point cloud segment.
First, in this embodiment, a test point is randomly selected from the first frame point cloud data for registration, and the test point is used as a sphere center and a preset radius is used as a sphere to obtain a test sphere. And then obtaining points in the test sphere to obtain neighborhood points. In order to determine which of the neighborhood points should be clustered with the test point into the same type of point set, in this embodiment, each neighborhood point is taken as a sphere center, and the preset radius is taken as a sphere to obtain a neighborhood sphere. And then acquiring the number of points in the neighborhood sphere, the oblateness value of a point set in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and determining the neighborhood point which should be clustered with the test point into a point set of the same type according to the oblateness value of the point set and the projection distance according to the number of the points, thereby segmenting the first frame of point cloud data. It should be noted that all the points in the test sphere region are neighborhood points, that is, there are a plurality of neighborhood points rather than only one, in one mode, for each neighborhood point, the neighborhood point is taken as the sphere center, the preset radius is taken as the sphere, a new neighborhood sphere is obtained again, and clustering is performed according to the above steps, that is, the step is an iterative process, and clustering is not performed on the point set which has been subjected to clustering until all the points in the first frame of point cloud data are traversed.
In an implementation manner, in order to implement accurate segmentation of the first frame point cloud data, in this embodiment, first, the number of points and coordinates of the points in the neighborhood sphere are obtained, and according to the number of points and the coordinates of the points in the neighborhood sphere, an oblateness value of a point set in the neighborhood sphere is obtained. And simultaneously acquiring the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame point cloud data. And then respectively comparing the number of points in the neighborhood sphere, the projection distance, the oblateness value of the point set in the neighborhood sphere with a point threshold, a projection distance threshold and a oblateness value threshold of the point set, and clustering the neighborhood points and the test points to obtain point cloud segments when the number of points in the neighborhood sphere, the projection distance and the oblateness value of the point set in the neighborhood sphere are respectively in the point threshold, the projection distance threshold and the oblateness value threshold of the point set.
In short, the point cloud area can be increased according to the wired sequence of the surface structure, the line structure and other structures through three constraint conditions, so that the first frame of point cloud data is correctly segmented. The first constraint condition is the number of points in the neighborhood sphere, the number of points in the neighborhood sphere is the point cloud density in the neighborhood sphere region, and the point cloud separated in space can be correctly segmented according to the constraint condition, including outliers. However, the spatially continuous turning lines cannot be correctly segmented only by using the constraint condition of the point cloud density, so that the implementation adds a second constraint condition, that is, the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame point cloud data represents the offset degree of the projection on the normal vector to a certain extent, so that the projection distance is limited, that is, the segmentation direction of the point cloud is constrained, and after the constraint condition is added, the spatially continuous turning lines can be well separated. However, only by using density constraint and direction constraint, the separation point and the turning point may not be truly corresponding, so the third constraint condition is added in this embodiment, that is, the oblateness value of the point set in the neighborhood sphere represents the oblateness of the point cloud in the neighborhood sphere region. After the constraint condition is added, the separation point and the turning point can be correctly corresponded, so that an ideal segmentation effect is obtained.
In one implementation, to obtain the oblateness values of the point sets in the neighborhood sphere, the embodiment obtains the arithmetic mean coordinates in the neighborhood sphere according to the number of points in the neighborhood sphere and the coordinates of the points by first obtaining the number of points and the coordinates of the points in the neighborhood sphere. And then obtaining a first scatter matrix corresponding to the neighborhood sphere according to the number of the points in the neighborhood sphere, the coordinates of the points and the arithmetic mean coordinate. And finally, acquiring a first eigenvalue and a third eigenvalue of the first scattering matrix, and obtaining the ellipticity value of the point set in the neighborhood sphere according to the first eigenvalue and the third eigenvalue of the first scattering matrix. The eigenvalues of the scatter matrix are an inherent property of the matrix and, in one implementation, may be obtained using SVD decomposition. Generally, when processing a three-dimensional point cloud, the scatter matrix is a 3X3 matrix, which includes three eigenvalues, namely a first eigenvalue, a second eigenvalue, and a third eigenvalue, in order of magnitude.
The normal calculation procedure for the flatness value is as follows: it is assumed that all points within a point cloud region fall within an ellipsoid, i.e. every point (x, y, z) within the region satisfies the following formula (1):
wherein, a, b, c are the lengths of three half axes of the ellipsoid respectively, and the ellipticity f of the point cloud segment is defined as the following formula (2):
the following formula (3) can be obtained by expressing the formula 1 in a matrix form as shown below, and the formulas (4), (5) can be obtained from the formula (3):
where Σ is the scatter matrix for the point cloud segment, m represents the number of points, and μ represents the arithmetic mean coordinate of all points. In order to save the computation time, in this embodiment, a dispersion matrix of the point set in the neighborhood sphere is obtained instead of a conventional manner of computing the oblateness value, and the oblateness value of the region is obtained according to the first eigenvalue and the third eigenvalue of the dispersion matrix, that is, the oblateness value is obtained according to formula 6. The equation 6 is as follows:
wherein λ isL,λSThe maximum and minimum eigenvalues of the scatter matrix, i.e. the first eigenvalue and the third eigenvalue of the scatter matrix, respectively.
In an implementation manner, in order to ensure the reliability of the test point, after the test point is selected, the number of points in the test sphere and the ellipticity value of the point set need to be obtained. And comparing the number of the points with a preset number of points, judging the test point as an invalid test point when the number of the points is less than the preset number of points, and reselecting the test point from the first frame of point cloud data. And comparing the flatness value of the point set with a preset flatness value of the point set, and when the flatness value of the point set is smaller than the preset flatness value of the point set, similarly determining the test point as an invalid test point, and reselecting the test point from the first frame point cloud data. And the test point can be used as an effective test point only when the number of the points is more than or equal to the preset number of points and the oblateness value of the point set is more than or equal to the preset oblateness value of the point set.
After the first frame of point cloud data is aggregated to obtain a point cloud segment, the method further comprises the following steps:
and S200, acquiring a flatness value of the point cloud segment.
Specifically, in order to obtain the flatness value of the point cloud segment, in an implementation manner, as shown in fig. 3, the step S200 specifically includes the following steps:
step S210, acquiring the number of points and the coordinates of the points in the point cloud segment, and obtaining arithmetic mean coordinates according to the number of the points and the coordinates of the points in the point cloud segment;
step S220, obtaining a second scatter matrix corresponding to the point cloud segment according to the number of the points in the point cloud segment, the coordinates of the points and the arithmetic mean value coordinate;
step S230, obtaining the first eigenvalue and the third eigenvalue of the second scattering matrix, and obtaining the flatness value of the point cloud segment according to the first eigenvalue and the third eigenvalue of the second scattering matrix.
In this embodiment, the number of points and the coordinates of the points in the point cloud segment are first obtained, and an arithmetic mean coordinate is obtained according to the number of the points and the coordinates of the points in the point cloud segment. And then obtaining a second scatter matrix corresponding to the point cloud segment according to the number of the points in the point cloud segment, the coordinates of the points and the arithmetic mean value coordinate. And finally, acquiring a first eigenvalue and a third eigenvalue of the second scattering matrix, and obtaining the flatness value of the rectangular structure according to the first eigenvalue and the third eigenvalue of the second scattering matrix. This step is similar to the above-described process of obtaining the ellipticity value of the point set in the neighborhood sphere region, and will not be described again here.
In addition, a rectangle fitting needs to be performed on the segmented point cloud segment, as shown in fig. 1, the method further includes the following steps:
and S300, performing rectangle fitting on the point cloud segment to obtain a rectangular structure, and acquiring four vertex coordinates of the rectangular structure.
In this embodiment, after the first frame of point cloud data is divided into a plurality of point cloud segments, each point cloud segment needs to be subjected to rectangle fitting to obtain a rectangular structure. In order to obtain coordinates of four vertices of the rectangular structure, in an implementation manner, as shown in fig. 4, the step S300 specifically includes the following steps:
s310, performing rectangle fitting on the point cloud segment to obtain a rectangular structure;
step S320, obtaining a second eigenvalue of the second dispersion matrix;
and S330, obtaining coordinate data of four vertexes of the rectangular structure according to the coordinates of the points in the point cloud segment, the arithmetic mean coordinates, the first characteristic value and the second characteristic value.
Specifically, rectangular fitting is performed on the point cloud segment, that is, the coordinate of each marked point in fig. 5 is calculated, and this embodiment mainly obtains the coordinate of each marked point in fig. 5 through the following formulas (7) to (14):
m＝μ+min{(piμ)T·v1}·v1 (7)
n＝μ+max{(pi-μ)T·v1}·v1 (8)
e＝μ+min{(pi-μ)T·v2}·v2 (9)
f＝μ+max{(pi-μ)T·v2}·v2 (10)
a＝m+f-μ (11)
b＝m+e-μ (12)
c＝n+e-μ (13)
d＝m+f-μ (14)
wherein p isiRepresents the ith point, mu is the arithmetic mean coordinate of the point cloud segment, v1，v2The eigenvectors corresponding to the first eigenvalue and the second eigenvalue of the dispersion matrix sigma, respectively, and a, b, c, d are the coordinates of the four vertices of the rectangular structure.
In order to achieve accurate registration of two frames of point cloud data, as shown in fig. 1, the method further includes the following steps:
and S400, acquiring second frame point cloud data, and registering the first frame point cloud data and the second frame point cloud data according to the four vertex coordinates of the rectangular structure and the oblateness value of the point cloud segment.
The method comprises the steps of firstly obtaining a second frame of point cloud needing point cloud registration, and then obtaining the optimal condition for registering two frames of point cloud data according to four vertex coordinates of the rectangular structure and the oblateness value of the point cloud segment so as to register the two frames of point cloud data.
In one implementation, as shown in fig. 6, the step S400 specifically includes the following steps:
s410, acquiring second frame point cloud data and six-degree-of-freedom transformation quantity in an iteration process;
step S420, obtaining coordinate data of transformation points under the coordinates of the first frame of point cloud data generated based on points under the coordinate system of the second frame of point cloud data according to the six-degree-of-freedom variation, and determining point cloud segments corresponding to the transformation points;
step S430, obtaining a target pose transformation matrix according to the flatness value of the point cloud segment, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and the coordinate data of the transformation point;
and step S340, registering the second frame of point cloud data and the first frame of point cloud data according to the target pose transformation matrix.
In order to realize the registration between the first frame point cloud data and the second frame point cloud data, the embodiment first needs to acquire the second frame point cloud data and the six-degree-of-freedom transformation amount in the iterative process. In particular, the spatial complex motion of any object can be simplified into translation and rotation in space, and the translation and rotation in space can be specifically expressed as translation in the direction of X, Y, Z and rotation around the axis X, Y, Z, namely, transformation of six degrees of freedom. Therefore, when the transformation amount of a certain point under the coordinate system of a certain frame of point cloud data on six degrees of freedom in the iteration process is determined, the corresponding point of the point under the coordinate system of another frame of point cloud data after iteration, namely the transformation point, can be obtained. Therefore, the point under the coordinate system of the second frame of point cloud data is transformed according to the six-degree-of-freedom variable quantity to obtain a transformation point under the coordinate of the first frame of point cloud data. And then searching the point cloud segment corresponding to the transformation point according to a preset condition to obtain a target point cloud segment.
In one implementation, the step of obtaining, according to the six-degree-of-freedom variation, coordinate data of a transformation point in the coordinate of the first frame point cloud data generated based on a point in the coordinate system of the second frame point cloud data, and determining a point cloud segment corresponding to the transformation point specifically includes: obtaining a pose transformation matrix according to the six-degree-of-freedom variation; and performing rigid body transformation on the point under the coordinate system of the second frame of point cloud data according to the pose transformation matrix to obtain a transformation point under the coordinate of the first frame of point cloud data. And then acquiring coordinate data of the transformation points, and searching a point cloud segment with the minimum distance value with the transformation points according to the coordinate data of the transformation points to obtain the point cloud segment corresponding to the transformation points.
For example, assume that a point p ═ p in the second frame point cloud data is knownx,py,pz]TAnd six-freedom transformation quantity xi ═ t in iterative processx,ty,tz,α,β,γ]TThe transformed point p' is calculated by the rigid transformation T (ξ) and is considered to be under the coordinate system of the reference point cloud. The specific process is shown in formula (15):
and then obtaining a target pose transformation matrix according to the flatness value of the point cloud segment, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and the coordinate data of the transformation point. In short, the final objective of this embodiment is to optimize the pose transformation matrix of two frames of point cloud data, and the pose transformation matrix obtained after optimization is the target pose transformation matrix, which can realize accurate registration between two frames of point cloud data.
In order to obtain the target pose transformation matrix, in an implementation manner, in this embodiment, a squared distance value is first obtained according to the coordinate data of the transformation point, the coordinate data of four vertices of the rectangular structure corresponding to the point cloud segment, and a distance squared function, then a likelihood function value of a likelihood function generated based on the pose transformation matrix is obtained according to a flat rate value of the point cloud segment and the squared distance value, and finally, a pose transformation matrix input to the likelihood function is obtained when the likelihood function value is a maximum value, so as to obtain a target pose transformation matrix.
For example, as shown in fig. 7, assuming that a rectangular structure with four coordinate points a, b, c, and d as vertices is closest to the transformation point, a rectangle is obtained according to the four coordinate points a, b, c, and d, and four sides of the rectangle are extended to obtain a dotted line in the drawing, and the dotted line is extended into a plane in the direction of a normal vector of the plane where the rectangle is located, so that the space is finally divided into 9 mutually non-overlapping subspaces. The dots in the figure represent the transformed points p 'that fall in different subspaces, the arrow points to the point on the rectangle closest to p', and the length of the arrow represents the point-to-rectangle distance. Specifically, the distances from a point to a rectangle can be classified into the following three categories:
1) point-to-point distance: in the subspace 2,3,4,5, the closest point of p 'is the vertex a, b, c, d of the rectangle, so the distance from point p' to vertex is taken as the point-to-rectangle distance.
2) Distance from point to line: in subspaces 6,7,8,9, the closest point of p 'becomes the foot of the foot from its closest rectangular side, so the distance of point p' to the closest rectangular side line is taken as the point-to-rectangular distance.
3) Distance from point to plane: in subspace 1, the closest point of p 'becomes its projected point on the plane of the rectangle along the normal vector, so the distance of point p' to the plane of the rectangle is taken as the point-to-rectangle distance.
Therefore, the squared distance value D can be solved by substituting the coordinates of the transformed points into a distance square function D (p '), where the following equation 16 is the distance square function D (p'):
as shown in the following equation (16), the distance squared function D (p ') is obtained by substituting the point p' obtained by the rigid body transformation:
wherein, a, b, c and d are four vertexes of the rectangle in the figure, e1、e2Is a unit direction vector of adjacent sides of a rectangle in the figure, r1、r2Representing the division plane in fig. 7. And then obtaining a likelihood function value of a likelihood function generated based on the pose transformation matrix according to the ellipticity value of the point cloud segment and the square distance value.
For example, the likelihood value of the point is obtained by substituting the squared distance d into a medium likelihood function l (d), i.e., the following equation (17):
wherein the content of the first and second substances,
is the flatness value. Likelihood function of all pointsThe values are accumulated to obtain the final likelihood value as shown in the following equation (18):
wherein, Pk+1For the second frame of point cloud data, RkIndicating that the ellipticity information is extracted from the first frame of point cloud data. And finally, acquiring a pose transformation matrix input into the likelihood function when the likelihood function value is the maximum value, and acquiring a target pose transformation matrix.
In mathematical statistics, a likelihood function is a function of parameters in a statistical model, representing the likelihood in the model parameters. Likelihood functions play a significant role in statistical inference, where "likelihood" is synonymous with "likelihood" or "probability", and refers to the likelihood of an event occurring, but in statistics, there is a clear distinction between "likelihood" and "likelihood" or "probability". Probabilities are used to predict the results of subsequent observations given some parameters, while likelihoods are used to estimate parameters about the nature of things given some observations. In short, in this embodiment, the likelihood function is used to determine the optimal pose transformation matrix when the first frame point cloud data and the second frame point cloud data are transformed, and when the function value of the likelihood function is larger, the probability that the transformation point obtained after the point in the second frame point cloud data passes through the pose transformation matrix is correct is larger, and vice versa, so that the input pose transformation matrix when the likelihood function value is maximum is used as the target pose transformation matrix to achieve accurate registration between the two frames of point cloud data.
In order to obtain the pose transformation matrix input when the likelihood function value is maximum, this embodiment provides an analytic expression of the final first derivative of the likelihood function, and converts the problem of solving the maximum value of the likelihood function into solving an unconstrained minimization problem, as shown in the following equation (19):
from the chain rule of the functional differentiation, the first derivative of the likelihood function can be solved for equation (20):
more specifically, each sub-function in the composite function
The derivative and derivation process of (c) is as follows:
finally, the first derivative of the total likelihood function is the sum of the likelihood function derivatives for each point, as shown in equation (21) below:
in combination with the above formula, an analytical expression of the likelihood function and its first derivative can be obtained, which can prove to be continuously derivable. Therefore, the common mathematical optimization method can be adopted to solve
Is equivalent to solving a likelihood function
Is measured. In one implementation, to obtain a faster convergence rate, the solution may continue
And the minimum value is solved by newton's method. Although the second derivative is continuously derivable only in each subspace and jumps exist on the spatial interface, the probability of the point falling on the interface can be ignored due to the dispersion of the point cloud, and therefore the optimization effect is not influenced. This embodiment also provides solving by newton's method, as shown in the following equation (22)
The solution method of the Hessian matrix required in the second derivative of (1) is as follows:
as shown in fig. 8, the graph is a diagram of two partially overlapped point clouds of two frames of a selected city street scene, and fig. 9 is an effect diagram obtained by the method provided by the present invention, so that it can be seen that the two frames of point clouds are effectively registered. In addition, as shown in fig. 10, images after outliers and noise are added to two partially overlapped frame point clouds, and fig. 11 shows that the two frame point clouds after outliers and noise are added are registered by using the method provided by the present invention, and it can be seen that successful registration can still be performed, so the method also has higher robustness to noise in point cloud registration.
Based on the above embodiments, the present invention further provides an intelligent terminal, and a schematic block diagram thereof may be as shown in fig. 12. The intelligent terminal comprises a processor, a memory, a network interface and a display screen which are connected through a system bus. Wherein, the processor of the intelligent terminal is used for providing calculation and control capability. The memory of the intelligent terminal comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the intelligent terminal is used for being connected and communicated with an external terminal through a network. The computer program, when executed by a processor, implements a method of point cloud registration using rectangle and ellipticity information. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen.
It will be understood by those skilled in the art that the block diagram of fig. 12 is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation to the intelligent terminal to which the solution of the present invention is applied, and a specific intelligent terminal may include more or less components than those shown in the figure, or combine some components, or have different arrangements of components.
In one implementation, one or more programs are stored in a memory of the smart terminal and configured to be executed by one or more processors include instructions for performing a method of point cloud registration utilizing rectangle and ellipticity information.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, databases, or other media used in embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
In summary, the present invention provides a point cloud registration method using rectangle and ellipticity information, that is, a storage medium, the method includes: acquiring first frame point cloud data, and clustering the first frame point cloud data to obtain point cloud segments; acquiring a flat rate value of the point cloud segment; performing rectangle fitting on the point cloud segment to obtain a rectangular structure, and acquiring four vertex coordinates of the rectangular structure; and acquiring second frame point cloud data, and registering the first frame point cloud data and the second frame point cloud data according to the four vertex coordinates of the rectangular structure and the oblateness value of the point cloud segment. According to the method, the first frame of point cloud data is segmented by using a clustering method, so that the first frame of point cloud data can be increased according to a certain sequence, and then the flatness information of the segmented point cloud segments is acquired to match the two frames of point cloud data, so that the structure of the point cloud segments is accurately expressed by the flatness information, and the stability and reliability of point cloud registration in scene transformation are improved.
It is to be understood that the invention is not limited to the examples described above, but that modifications and variations may be effected thereto by those of ordinary skill in the art in light of the foregoing description, and that all such modifications and variations are intended to be within the scope of the invention as defined by the appended claims.
Claims (10)
1. A point cloud registration method using rectangle and ellipticity information, the method comprising:
acquiring first frame point cloud data, and clustering the first frame point cloud data to obtain point cloud segments;
acquiring a flat rate value of the point cloud segment;
performing rectangle fitting on the point cloud segment to obtain a rectangular structure, and acquiring coordinate data of four vertexes of the rectangular structure;
and acquiring second frame point cloud data, and registering the first frame point cloud data and the second frame point cloud data according to the coordinate data of the four vertexes of the rectangular structure and the oblateness value of the point cloud segment.
2. The point cloud registration method using rectangle and ellipticity information according to claim 1, wherein the acquiring a first frame of point cloud data, clustering the first frame of point cloud data to obtain point cloud segments:
acquiring first frame point cloud data, and taking a preset radius as a sphere in the first frame point cloud data to obtain a test sphere;
obtaining points in the test sphere to obtain neighborhood points, and taking the neighborhood points as the sphere center and the preset radius as the sphere to obtain a neighborhood sphere;
acquiring the number of points in the neighborhood sphere, the oblateness value of the point set in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and clustering the first frame of point cloud data according to the number of the points, the oblateness value of the point set and the projection distance to obtain a point cloud segment.
3. The method of claim 2, wherein the obtaining the number of points in the neighborhood sphere, the oblateness value of the set of points in the neighborhood sphere, and the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame of point cloud data, and the clustering the first frame of point cloud data according to the number of points, the oblateness value of the set of points, and the projection distance to obtain the point cloud segment comprises:
acquiring the number of points and the coordinates of the points in the neighborhood sphere, and acquiring the oblateness value of the point set in the neighborhood sphere according to the number of the points and the coordinates of the points in the neighborhood sphere;
acquiring the projection distance of the distance between the neighborhood point and the test point on the normal vector of the plane corresponding to the first frame point cloud data;
respectively comparing the number of points in the neighborhood sphere, the projection distance, and the oblateness value of the point set in the neighborhood sphere with a point number threshold, a projection distance threshold, and a oblateness value threshold of the point set;
and when the number of the points in the neighborhood sphere, the projection distance and the oblateness value of the point set in the neighborhood sphere are respectively in the point number threshold, the projection distance threshold and the oblateness value threshold of the point set, clustering the neighborhood points and the test points to obtain point cloud segments.
4. The method of claim 3, wherein the obtaining the number of points and coordinates of the points in the neighborhood sphere, and obtaining the ellipticity value of the set of points in the neighborhood sphere according to the number of points and coordinates of the points in the neighborhood sphere comprises:
acquiring the number of points and the coordinates of the points in the neighborhood sphere, and acquiring the arithmetic mean coordinates in the neighborhood sphere according to the number of the points and the coordinates of the points in the neighborhood sphere;
obtaining a first scattering matrix corresponding to the neighborhood sphere according to the number of the points in the neighborhood sphere, the coordinates of the points and the arithmetic mean coordinate;
and acquiring a first eigenvalue and a third eigenvalue of the first scatter matrix, and obtaining the ellipticity value of the point set in the neighborhood sphere according to the first eigenvalue and the third eigenvalue of the first scatter matrix.
5. The method of claim 1, wherein the obtaining the ellipticity values of the point cloud segments comprises:
acquiring the number of points and the coordinates of the points in the point cloud segment, and acquiring arithmetic mean coordinates according to the number of the points and the coordinates of the points in the point cloud segment;
obtaining a second scatter matrix corresponding to the point cloud segment according to the number of the points in the point cloud segment, the coordinates of the points and the arithmetic mean value coordinate;
and acquiring a first eigenvalue and a third eigenvalue of the second scatter matrix, and obtaining the flatness value of the point cloud segment according to the first eigenvalue and the third eigenvalue of the second scatter matrix.
6. The point cloud registration method using rectangle and ellipticity information according to claim 5, wherein the performing rectangle fitting on the point cloud segments to obtain a rectangular structure, and obtaining coordinate data of coordinates of four vertices of the rectangular structure comprises:
performing rectangle fitting on the point cloud segment to obtain a rectangular structure;
acquiring a second eigenvalue of the second dispersion matrix;
and obtaining coordinate data of four vertexes of the rectangular structure according to the coordinates of the points in the point cloud segment, the arithmetic mean value coordinates, the first characteristic value and the second characteristic value.
7. The method of claim 1, wherein the obtaining the second frame of point cloud data, and the registering the first frame of point cloud data and the second frame of point cloud data according to the coordinate data of the four vertices of the rectangular structure and the ellipticity of the point cloud segment comprises:
acquiring second frame point cloud data and six-degree-of-freedom transformation quantity in an iteration process;
obtaining coordinate data of a transformation point under the coordinate of the first frame of point cloud data generated based on a point under the coordinate system of the second frame of point cloud data according to the six-degree-of-freedom variation, and determining a point cloud segment corresponding to the transformation point;
obtaining a target pose transformation matrix according to the flatness value of the point cloud segment, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and the coordinate data of the transformation point;
and registering the second frame of point cloud data and the first frame of point cloud data according to the target pose transformation matrix.
8. The method of claim 7, wherein obtaining coordinate data of transformation points in coordinates of the first frame of point cloud data generated based on points in a coordinate system of the second frame of point cloud data according to the six-degree-of-freedom variance and determining point cloud segments corresponding to the transformation points comprises:
obtaining a pose transformation matrix according to the six-degree-of-freedom variation;
performing rigid body transformation on points under the coordinate system of the second frame of point cloud data according to the pose transformation matrix to obtain transformation points under the coordinates of the first frame of point cloud data;
and acquiring coordinate data of the transformation points, searching a point cloud segment with the minimum distance value with the transformation points according to the coordinate data of the transformation points, and obtaining the point cloud segment corresponding to the transformation points.
9. The point cloud registration method using rectangle and flat rate information according to claim 7, wherein the obtaining a target pose transformation matrix according to the flat rate values of the point cloud segments, the coordinate data of the four vertices of the rectangular structure corresponding to the point cloud segments, and the coordinate data of the transformation points comprises:
obtaining a square distance value according to the coordinate data of the transformation point, the coordinate data of four vertexes of the rectangular structure corresponding to the point cloud segment and a distance square function;
obtaining a likelihood function value of a likelihood function generated based on the pose transformation matrix according to the ellipticity value and the square distance value of the point cloud fragment;
and acquiring a pose transformation matrix input into the likelihood function when the likelihood function value is the maximum value to obtain a target pose transformation matrix.
10. A storage medium having stored thereon instructions adapted to be loaded and executed by a processor to perform the steps of a method for point cloud registration using rectangle and ellipticity information according to any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011146501.2A CN112308889B (en) | 2020-10-23 | 2020-10-23 | Point cloud registration method and storage medium by utilizing rectangle and oblateness information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011146501.2A CN112308889B (en) | 2020-10-23 | 2020-10-23 | Point cloud registration method and storage medium by utilizing rectangle and oblateness information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112308889A true CN112308889A (en) | 2021-02-02 |
| CN112308889B CN112308889B (en) | 2021-08-31 |
Family
ID=74327393
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011146501.2A Active CN112308889B (en) | 2020-10-23 | 2020-10-23 | Point cloud registration method and storage medium by utilizing rectangle and oblateness information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112308889B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581481A (en) * | 2022-03-07 | 2022-06-03 | 广州小鹏自动驾驶科技有限公司 | Target object speed estimation method and device, vehicle and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160350979A1 (en) * | 2015-05-28 | 2016-12-01 | The Florida International University Board Of Trustees | Systems and methods for shape analysis using landmark-driven quasiconformal mapping |
| CN109116321A (en) * | 2018-07-16 | 2019-01-01 | 中国科学院国家空间科学中心 | A kind of phase filtering method and height measurement method of spaceborne interference imaging altimeter |
| CN109493407A (en) * | 2018-11-19 | 2019-03-19 | 腾讯科技（深圳）有限公司 | Realize the method, apparatus and computer equipment of laser point cloud denseization |
| WO2020055272A1 (en) * | 2018-09-12 | 2020-03-19 | Auckland Uniservices Limited | Methods and systems for ocular imaging, diagnosis and prognosis |
| US10665035B1 (en) * | 2017-07-11 | 2020-05-26 | B+T Group Holdings, LLC | System and process of using photogrammetry for digital as-built site surveys and asset tracking |
| CN111524174A (en) * | 2020-04-16 | 2020-08-11 | 上海航天控制技术研究所 | Binocular vision three-dimensional construction method for moving target of moving platform |
| CN111539432A (en) * | 2020-03-11 | 2020-08-14 | 中南大学 | Method for extracting urban road by using multi-source data to assist remote sensing image |
-
2020
- 2020-10-23 CN CN202011146501.2A patent/CN112308889B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160350979A1 (en) * | 2015-05-28 | 2016-12-01 | The Florida International University Board Of Trustees | Systems and methods for shape analysis using landmark-driven quasiconformal mapping |
| US10665035B1 (en) * | 2017-07-11 | 2020-05-26 | B+T Group Holdings, LLC | System and process of using photogrammetry for digital as-built site surveys and asset tracking |
| CN109116321A (en) * | 2018-07-16 | 2019-01-01 | 中国科学院国家空间科学中心 | A kind of phase filtering method and height measurement method of spaceborne interference imaging altimeter |
| WO2020055272A1 (en) * | 2018-09-12 | 2020-03-19 | Auckland Uniservices Limited | Methods and systems for ocular imaging, diagnosis and prognosis |
| CN109493407A (en) * | 2018-11-19 | 2019-03-19 | 腾讯科技（深圳）有限公司 | Realize the method, apparatus and computer equipment of laser point cloud denseization |
| CN111539432A (en) * | 2020-03-11 | 2020-08-14 | 中南大学 | Method for extracting urban road by using multi-source data to assist remote sensing image |
| CN111524174A (en) * | 2020-04-16 | 2020-08-11 | 上海航天控制技术研究所 | Binocular vision three-dimensional construction method for moving target of moving platform |
Non-Patent Citations (4)
| Title |
|---|
| HAO MING 等: "Robust Multisource Remote Sensing Image Registration Method Based on Scene Shape Similarity", 《PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING》 * |
| XIJIANG CHEN 等: "Evaluating point cloud accuracy of static three-dimensional laser scanning based on point cloud error ellipsoid model", 《JOURNAL OF APPLIED REMOTE SENSING》 * |
| 付怡然: "集成ICP和NDT的地面激光扫描点云渐进配准方法", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 孙文潇 等: "法线特征约束的激光点云精确配准", 《武汉大学学报信息科学版》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581481A (en) * | 2022-03-07 | 2022-06-03 | 广州小鹏自动驾驶科技有限公司 | Target object speed estimation method and device, vehicle and storage medium |
| CN114581481B (en) * | 2022-03-07 | 2023-08-25 | 广州小鹏自动驾驶科技有限公司 | Target speed estimation method and device, vehicle and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112308889B (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111780763B (en) | Visual positioning method and device based on visual map | |
| US20210082181A1 (en) | Method and apparatus for object detection, intelligent driving method and device, and storage medium | |
| US11417011B2 (en) | 3D human body pose estimation using a model trained from unlabeled multi-view data | |
| CN108830931B (en) | Laser point cloud simplification method based on dynamic grid k neighborhood search | |
| CN106780442B (en) | Stereo matching method and system | |
| Nurunnabi et al. | Robust statistical approaches for local planar surface fitting in 3D laser scanning data | |
| US20200394506A1 (en) | Machine learning system | |
| WO2022041598A1 (en) | Remote sensing image segmentation method and system, terminal, and storage medium | |
| CN111127631B (en) | Three-dimensional shape and texture reconstruction method, system and storage medium based on single image | |
| CN109886279B (en) | Image processing method, device, computer equipment and storage medium | |
| Du et al. | New iterative closest point algorithm for isotropic scaling registration of point sets with noise | |
| CN111814905A (en) | Target detection method, target detection device, computer equipment and storage medium | |
| CN115546519B (en) | Matching method of image and millimeter wave radar target for extracting pseudo-image features | |
| WO2023116632A1 (en) | Video instance segmentation method and apparatus based on spatio-temporal memory information | |
| CN115457492A (en) | Target detection method and device, computer equipment and storage medium | |
| Masuda et al. | Toward unsupervised 3d point cloud anomaly detection using variational autoencoder | |
| CN112308889B (en) | Point cloud registration method and storage medium by utilizing rectangle and oblateness information | |
| CN113436223B (en) | Point cloud data segmentation method and device, computer equipment and storage medium | |
| CN115526850A (en) | Training method and device for real-space decoder of refrigeration electron microscope and electronic equipment | |
| Chuang et al. | Learning-guided point cloud vectorization for building component modeling | |
| CN111765883A (en) | Monte Carlo positioning method and equipment for robot and storage medium | |
| CN109166138B (en) | Target tracking method and device based on high-order cumulant and storage medium | |
| US20220351533A1 (en) | Methods and systems for the automated quality assurance of annotated images | |
| López-Rubio et al. | Robust fitting of ellipsoids by separating interior and exterior points during optimization | |
| WO2023164933A1 (en) | Building modeling method and related apparatus |
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 |