CN114742868A - Point cloud registration method and device and electronic equipment - Google Patents
Point cloud registration method and device and electronic equipment Download PDFInfo
- Publication number
- CN114742868A CN114742868A CN202210384544.7A CN202210384544A CN114742868A CN 114742868 A CN114742868 A CN 114742868A CN 202210384544 A CN202210384544 A CN 202210384544A CN 114742868 A CN114742868 A CN 114742868A
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- cloud
- pair
- curvature
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 230000009466 transformation Effects 0.000 claims abstract description 45
- 238000012216 screening Methods 0.000 claims abstract description 12
- 230000001131 transforming effect Effects 0.000 claims abstract description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 18
- 238000004364 calculation method Methods 0.000 description 10
- 238000009826 distribution Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 5
- 238000000354 decomposition reaction Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000013519 translation Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000001174 ascending effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 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
- G06T7/30—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
- 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
- G06T7/35—Determination of transform parameters for the alignment of images, i.e. image registration using statistical methods
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The embodiment of the application discloses a point cloud registration method, a point cloud registration device and electronic equipment, wherein the method comprises the following steps: determining curvatures of points in the first point cloud and the second point cloud; determining a curvature threshold of a target point according to the mean value and the variance of the curvatures of the midpoints in the neighborhood of the target point, and determining the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point to obtain the feature point of the first point cloud and the feature point of the second point cloud, wherein the target point is each point in the first point cloud and the second point cloud; screening the characteristic points of the first point cloud and the second point cloud by using a curvature constraint condition to obtain a plurality of matching point pairs, wherein one matching point pair comprises one characteristic point of the first point cloud and one characteristic point of the second point cloud; determining a transformation relationship of the first point cloud and the second point cloud by using at least one of the plurality of matching point pairs; and transforming the first point cloud to a coordinate system of the second point cloud by using a transformation relation to obtain a point cloud registration result so as to improve the accuracy of point cloud registration.
Description
Technical Field
The present invention relates to the field of computers, and in particular, to a method and an apparatus for point cloud registration, and an electronic device.
Background
With the development of scientific technology, people pay attention to data processing and application. Point cloud data is widely used in various fields as a data storage format for three-dimensional information processing. Generally, a target object is measured at different angles to obtain a plurality of sets of point cloud data, and then the obtained plurality of sets of point cloud data are registered to obtain the data of the finished target object.
In the process of point cloud registration, it is usually necessary to determine feature points in the points of the point cloud to be registered to characterize the features of the point cloud to be registered. However, the current feature point determination method results in low accuracy of point cloud registration.
Disclosure of Invention
In view of this, the present application provides a method and an apparatus for point cloud registration, and an electronic device, so as to improve accuracy of point cloud registration.
In a first aspect, the present application provides a method of point cloud registration, the method comprising:
determining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud;
determining a curvature threshold of the target point according to the mean value and the variance of the curvature of the midpoint in the neighborhood of the target point, and determining the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point to obtain the feature point of the first point cloud and the feature point of the second point cloud; the target point is each point in the first point cloud and the second point cloud;
screening the characteristic points of the first point cloud and the second point cloud by using a curvature constraint condition to obtain a plurality of matching point pairs; wherein, a matching point pair comprises a characteristic point of a first point cloud and a characteristic point of a second point cloud;
determining a transformation relationship of the first point cloud and the second point cloud by using at least one of the plurality of matching point pairs;
and transforming the first point cloud to a coordinate system of the second point cloud by using a transformation relation to obtain a point cloud registration result.
In one possible implementation manner, before determining the transformation relationship between the first point cloud and the second point cloud by using at least one matching point pair of the plurality of matching point pairs, the method further includes:
determining a first point pair from the first point cloud, determining a second point pair corresponding to the first point pair from the second point cloud by using a curvature constraint condition, and determining a two-point pair group formed by the first point pair and the second point pair to obtain a two-point pair set containing a plurality of two-point pairs; wherein each point in the second pair of points and each point in the first pair of points satisfy a curvature constraint condition, the first pair of points being each point pair in the first point cloud;
determining a first two-point pair and a second two-point pair from the two-point pair set, and determining a four-point pair formed by the first two-point pair and the second two-point pair to obtain a four-point pair set containing a plurality of four-point pairs; wherein the first two-point pair and the second two-point pair have no coincident point, and the first two-point pair is each two-point pair in the two-point pair set;
determining a first four-point pair and a second four-point pair from the four-point pair set, and determining an eight-point pair formed by the first four-point pair and the second four-point pair to obtain an eight-point pair set containing a plurality of eight-point pairs; wherein the first four-point pair and the second four-point pair have no coincident point, and the first four-point pair is each four-point pair in the two-point pair set;
and determining four point pairs with preset number from the eight point pair set to obtain at least one matching point pair.
In a possible implementation manner, determining the curvature of each point in the first point cloud and the second point cloud specifically includes:
determining a curved surface of a first point, wherein the curved surface of the first point is obtained by fitting each point in the neighborhood of the first point, and the first point is each point in the first point cloud and the second point cloud;
determining a gaussian curvature of the curved surface of the first point and an average curvature of the curved surface of the first point;
the curvature of the first point is determined from the gaussian curvature of the curved surface of the first point and the average curvature of the curved surface of the first point.
In one possible implementation, a curvature threshold of the target point is determined according to a mean and a variance of curvatures of midpoints in a neighborhood of the target point; the target point is a point in the first point cloud and the second point cloud, and the method specifically comprises the following steps:
determining the mean and variance of the curvature of the midpoints in the neighborhood of the target point;
summing the mean value and the target variance to obtain a curvature threshold of the target point; wherein the target variance is a product of the variance and a preset coefficient.
In a possible implementation manner, before the step of screening the feature points of the first point cloud and the feature points of the second point cloud by using the curvature constraint condition to obtain a plurality of matching point pairs, the method further includes:
determining a similarity threshold;
and determining curvature constraint conditions according to the similarity threshold.
In one possible implementation, before determining the curvature of each point in the first point cloud and the second point cloud, the method further includes:
removing outliers in the first point cloud and removing outliers in the second point cloud; the outlier in the first point cloud is a point which is in the first point cloud and has a distance with other points in the first point cloud larger than a first distance threshold, and the outlier in the second point cloud is a point which is in the second point cloud and has a distance with other points in the second point cloud larger than a second distance threshold.
In one possible implementation manner, after transforming the first point cloud to the coordinate system of the second point cloud by using the transformation relationship to obtain the point cloud registration result, the method further includes:
and processing the point cloud registration result by using an iterative nearest neighbor algorithm to obtain a processed registration result.
In a second aspect, the present application provides an apparatus for point cloud registration, the apparatus comprising:
a curvature determining unit for determining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud;
the characteristic point determining unit is used for determining a curvature threshold of the target point according to the mean value and the variance of the curvatures of the midpoints in the neighborhood of the target point, and determining the target point as the characteristic point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point, so as to obtain the characteristic point of the first point cloud and the characteristic point of the second point cloud; the target point is each point in the first point cloud and the second point cloud;
the matching point determining unit is used for screening the characteristic points of the first point cloud and the second point cloud by utilizing the curvature constraint condition to obtain a plurality of matching point pairs; wherein, one matching point pair comprises a characteristic point of the first point cloud and a characteristic point of the second point cloud;
a transformation relation determining unit that determines a transformation relation between the first point cloud and the second point cloud using at least one of the plurality of matching point pairs;
and the point cloud registration unit is used for transforming the first point cloud to the coordinate system of the second point cloud by utilizing the transformation relation to obtain a point cloud registration result.
In a third aspect, the present application provides an electronic device, which includes a processor and a memory, where the memory stores codes, and the processor is configured to call the codes stored in the memory to execute any one of the methods described above.
In a fourth aspect, the present application provides a computer readable storage medium for storing a computer program for performing the method of any one of the above.
Drawings
Fig. 1 is a flowchart of a point cloud matching method according to an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of outliers in a point cloud provided in an embodiment of the present application;
FIG. 3A is a schematic diagram of a point cloud distribution of a first pose provided by an embodiment of the present application;
FIG. 3B is a schematic diagram of a point cloud distribution at a second pose according to an embodiment of the present disclosure;
fig. 4A is a schematic diagram of a first point cloud registration result provided in an embodiment of the present application;
fig. 4B is a schematic diagram of a second point cloud registration result provided in the embodiment of the present application;
FIG. 5A is a schematic diagram of a point cloud distribution of a third pose provided by an embodiment of the present application;
FIG. 5B is a schematic diagram of a point cloud distribution of a fourth pose provided by an embodiment of the present application;
fig. 6A is a schematic diagram of a third point cloud registration result provided in an embodiment of the present application;
fig. 6B is a schematic diagram of a fourth point cloud registration result provided in the embodiment of the present application;
fig. 7 is a schematic structural diagram of an apparatus for point cloud registration provided in an embodiment of the present application;
fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In the embodiments of the present application, words such as "exemplary" or "for example" are used to mean serving as examples, illustrations or descriptions. Any embodiment or design described herein as "exemplary" or "e.g.," is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word "exemplary" or "such as" is intended to present concepts related in a concrete fashion.
The terms "first", "second", and the like in the description of the present application are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated.
The terminology used in the description of the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application, which will be described in detail below with reference to the accompanying drawings.
With the continuous improvement of social informatization degree, the traditional two-dimensional image is difficult to meet the actual production and living needs of people, and the development and maturity of information acquisition equipment such as laser radars, binocular cameras and the like can enable operators to conveniently acquire three-dimensional data, namely point cloud data, of a target object. However, since the target object may be self-shielded during measurement and the field range of the device itself is limited, all data of the target object cannot be obtained by one measurement, which causes difficulty in subsequent applications. Therefore, in practice, a group of point cloud data is obtained by measuring a target object for multiple times at different angles, and then the point cloud data are combined and spliced according to a certain rule to construct complete target object data, and the process is called point cloud registration. The point cloud data registration technology can be applied to the fields of face recognition, automatic driving, three-dimensional reconstruction, smart cities, virtual reality and the like.
In the process of point cloud registration, feature points are determined in points of a point cloud to be registered to represent features of the point cloud to be registered. However, the current feature point determination method causes the accuracy of point cloud registration to be low.
Based on this, in the embodiments of the present application provided by the applicant, the curvature of each point in the first point cloud and the curvature of each point in the second point cloud are first determined; determining a curvature threshold of the target point according to the mean value and the variance of the curvature of the midpoint in the neighborhood of the target point, and determining the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point to obtain the feature point of the first point cloud and the feature point of the second point cloud; the target point is each point in the first point cloud and the second point cloud; screening the characteristic points of the first point cloud and the second point cloud by using a curvature constraint condition to obtain a plurality of matching point pairs; wherein, one matching point pair comprises a characteristic point of the first point cloud and a characteristic point of the second point cloud; determining a transformation relationship of the first point cloud and the second point cloud by using at least one of the plurality of matching point pairs; and transforming the first point cloud to a coordinate system of the second point cloud by using a transformation relation to obtain a point cloud registration result.
For the mode of determining the feature points by fixing the threshold, the reference standard of each point in the point cloud is fixed, the small threshold can cause too many feature points so as to increase the calculation amount, the large threshold can cause too few feature points so as to cause the too small search range of the matching points, and more accurate matching points are difficult to obtain, so that the accuracy of point cloud matching is low; further, the distribution range of the feature points extracted by the fixed threshold method is narrow, and the point cloud data generally has points with high curvature concentrated at positions where the overall fluctuation is large, so that the feature points extracted by the fixed threshold method are generally concentrated at these positions, and the fixed threshold method is difficult to extract for the feature points at positions where the overall fluctuation is small but the local fluctuation is present.
By adopting the technical scheme of the embodiment of the application, the characteristic points in the point cloud to be registered are determined by means of the curvature of the point cloud to be registered and the self-adaptive threshold, and compared with the method of determining the characteristic points by means of the fixed threshold, the self-adaptability of the curvature threshold can be utilized to obtain a proper number of characteristic points, so that the accuracy of point cloud matching is improved, and meanwhile, the calculation amount is balanced; by adopting the technical scheme of the embodiment of the application, the characteristic points of the positions with large overall undulation and the characteristic points of the positions with small overall undulation but local undulation can be extracted.
In order to facilitate understanding of the technical solutions provided in the embodiments of the present application, a method, an apparatus, and an electronic device for point cloud registration provided in the embodiments of the present application are described below with reference to the accompanying drawings.
While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be embodied in various forms and should not be limited to the embodiments set forth herein. Other embodiments, which can be derived by those skilled in the art from the embodiments given herein without any inventive contribution, are also within the scope of the present application.
In the claims and specification of the present application and in the drawings accompanying the description, the terms "comprise" and "have" and any variations thereof, are intended to cover non-exclusive inclusions.
The application provides a point cloud matching method.
Referring to fig. 1, fig. 1 is a flowchart illustrating a point cloud matching method according to an embodiment of the present disclosure.
As shown in fig. 1, the method of point cloud matching in the embodiment of the present application includes S101-S105.
S101, determining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud.
In differential geometry, the two principal curvatures (principal curvatures) at a given point of a surface measure how differently a surface curves in different directions at that point.
And obtaining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud.
S102, determining a curvature threshold of the target point according to the mean value and the variance of the curvatures of the midpoints in the neighborhood of the target point, and determining the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point to obtain the feature point of the first point cloud and the feature point of the second point cloud, wherein the target point is each point in the first point cloud and the second point cloud.
The Neighborhood of the target point (neighborwood) may refer to any open interval centered at the target point, and the Neighborhood midpoint of the target point refers to the point located within the interval.
At least one point may be included in the neighborhood of the target point.
And obtaining the mean value of the curvatures of the midpoints in the neighborhood of the target point and the variance of the curvatures of the midpoints in the neighborhood of the target point.
When the curvature of the target point is greater than the curvature threshold of the target point, determining the target point as a feature point of the point cloud to which the target point belongs, namely:
for example, when the target point is a point in the first point cloud, the target point is determined to be a feature point of the first point cloud when the curvature of the target point is greater than the curvature threshold of the target point.
The feature points of the first point cloud may include one or more points in the first point cloud.
The feature points of the second point cloud may include one or more points in the second point cloud.
The target point is each point in the first point cloud and the second point cloud, and the target point is processed for each point in the first point cloud and each point in the second point cloud.
S103, screening the characteristic points of the first point cloud and the second point cloud by using the curvature constraint conditions to obtain a plurality of matching point pairs.
A matching point pair includes a feature point of a first point cloud and a feature point of a second point cloud.
For each matching point pair, the feature points of the first point cloud and the feature points of the second point cloud are corresponding, and the feature points of the first point cloud and the feature points of the second point cloud satisfy the curvature constraint condition.
And S104, determining a transformation relation between the first point cloud and the second point cloud by using at least one matching point pair in the plurality of matching point pairs.
The transformation relation of the point clouds to be registered is used for converting coordinate systems of two point clouds to be matched, and the transformation relation comprises a group of rotation matrixes R and translation vectors T.
And S105, transforming the first point cloud to a coordinate system of the second point cloud by using a transformation relation to obtain a point cloud registration result.
By adopting the technical scheme of the embodiment of the application, the characteristic points in the point cloud to be registered are determined by utilizing the curvature of the middle point of the point cloud to be registered and the self-adaptive threshold, and compared with the method of determining the characteristic points by utilizing the fixed threshold, the self-adaptability of the curvature threshold can be utilized to obtain a proper number of characteristic points, so that the accuracy of point cloud matching is improved, and meanwhile, the calculation amount is balanced; by adopting the technical scheme of the embodiment of the application, the characteristic points of the positions with large overall undulation and the characteristic points of the positions with small overall undulation but local undulation can be extracted.
The application also provides another point cloud matching method.
The method for point cloud matching in the embodiment of the application comprises S201-S208.
S201, removing outliers in the first point cloud and the second point cloud.
Outliers of a point cloud refer to points that are off-centered from the main portion of the point cloud, typically due to equipment and environmental noise.
Referring to fig. 2, fig. 2 is a schematic diagram of outliers in a point cloud according to an embodiment of the present disclosure.
Due to the noise of the equipment and the environment, the point cloud usually contains outliers, as shown in fig. 2. Outliers may cause large errors in parameters of a fitted surface of some points, and curvature estimation is inaccurate, so that subsequent feature point extraction is affected. Therefore, outliers in the point cloud need to be removed.
The outliers in the first point cloud may include one or more points in the first point cloud.
For example, the point a is an outlier in the first point cloud, and if there are other points in the first point cloud besides the point a, the distance between the point a and the other points in the first point cloud is larger, that is, the point a deviates from the main body of the first point cloud.
In some possible cases, the distance between point a and the other points is greater than the first distance threshold.
Outliers in the second point cloud may include one or more points in the second point cloud.
For example, point B is an outlier in the second point cloud, and there are other points in the second point cloud besides point B, and then the distance between point B and other points in the second point cloud is larger, that is, point B deviates from the main body of the second point cloud.
In some possible cases, the distance between point B and the other points is greater than the second distance threshold.
In some possible cases, the first point cloud and the second point cloud may not have outliers.
The point P is an arbitrary point of the first point cloud, and how to remove outliers will be described below by taking the point P as an example.
Firstly, a k neighborhood of a point P and k points in the k neighborhood of the point P are determined, wherein the k neighborhood refers to k points which are closest to a certain central point in a point cloud.
The k neighborhood of point P refers to the k points in the point cloud that are closest in distance to point P.
And according to the distances between the point P and the k points, determining the sum of the distances between the point P and the k points, namely the sum of the distances from the point P to the k field P0.
Then, for k points in the k neighborhood of the point P, the sum of the distances from the k points to the k neighborhood is calculated in the manner described above, and the sum is sequentially denoted as P1, P2, …, and pk.
Calculating the mean μ of the distance sums of the k points:
calculating the variance σ of the distance sum of the k points:
when the sum of the distances P0 from the point P to the k-field is greater than the sum of the mean and variance described above, i.e., P0> μ + σ, the point P is determined to be an outlier of the first point cloud.
The manner of determining outliers of the second point cloud is similar and will not be described herein.
In some possible cases, outliers in the first and second point clouds may also be determined using other means.
S202, curvature of each point in the first point cloud and the second point cloud is determined.
And obtaining the curvature of each point in the first point cloud and each point in the second point cloud.
The curvature of each point may be referred to as a curvature, and is used to describe the amount of calculation of characteristics such as the degree of curvature and the direction of the curved surface corresponding to the point.
Points with a large value of curvature contain a higher amount of information and are suitable as feature points, and therefore the features of each point are quantitatively measured using curvature.
The following description will take the example of determining the curvature of each point in the first point cloud.
The point P is an arbitrary point of the first point cloud, and how to determine the curvature will be described below by taking the point P as an example.
First, a k ' neighborhood of point P is determined, as well as k ' points in the k ' neighborhood of point P.
In one possible implementation, k' is greater than or equal to 4.
The k' neighborhood used to determine the curvature and the k neighborhood used to determine the discrete points described above may be different.
Fitting a quadratic surface by using K ' points in the K ' neighborhood of the point P, and applying a least square method and a theory related to surface theory to the three-dimensional coordinates of (K ' +1) points (the point P and the K ' points in the K ' neighborhood of the point P) to obtain the Gaussian curvature K and the average curvature H of the surface.
Two principal curvatures k1 and k2 of point P are then determined:
curvature f0 of point P is obtained:
the determination method is similar for the curvatures of other points of the first point cloud, and details are not repeated here.
Similar ways of determining the curvature of each point in the second point cloud are not described here.
In some possible cases, the curvatures of the points in the first point cloud and the second point cloud may also be determined in other ways.
S203, determining the characteristic points of the first point cloud and the second point cloud.
The embodiment of the application provides a curvature adaptive threshold to screen points with curvatures meeting threshold requirements as feature points.
The following description will be given taking an example of determining the feature points of the first point cloud.
The point P is an arbitrary point of the first point cloud, and the determination of the feature point of the first point cloud will be described below by taking the point P as an example.
First, the k "neighborhood of point P is determined, as well as the k" points in the k "neighborhood of point P.
The k "neighborhood used to determine the feature points of the first point cloud, and the k' neighborhood used to determine the curvature, and the k neighborhood used to determine the discrete points, may be different.
The curvature of point P is f0, and the curvatures of k "points in the k" neighborhood of point P are f1, f2, …, fk ", respectively.
The mean μ "and variance σ" of the curvature of k "points in the k" neighborhood of point P are calculated:
when f0> μ ″ + α ×, σ ″, the curvature of the point P is considered to be significantly higher than the local curvature, and the point P is considered to be a suddenness point, and the point P can be extracted as a feature point of the first point cloud to reflect the local feature; wherein, alpha is a self-adaptive coefficient and can be adjusted according to the number of required characteristic points.
The manner of determining whether the points are feature points is similar for other points in the first point cloud and each point in the second point cloud, and details are not repeated here.
In some possible implementation manners, the processing is performed on each point in the first point cloud and each point in the second point cloud to obtain feature points of the first point cloud and feature points of the second point cloud.
S204, screening the characteristic points of the first point cloud and the second point cloud by using the curvature constraint condition to obtain a plurality of matching point pairs.
A matching point pair includes a feature point of a first point cloud and a feature point of a second point cloud.
For each matching point pair, the feature points of the first point cloud and the feature points of the second point cloud are corresponding, and the feature points of the first point cloud and the feature points of the second point cloud satisfy the curvature constraint condition.
And the transformation relation of the point clouds to be registered is used for converting the coordinate systems of the two point clouds to be matched, and the transformation relation comprises a group of rotation matrixes R and translation vectors T.
The transformation relation can be obtained by solving the coordinates of the matching point pairs in the two point clouds to be registered.
The matching point pair refers to points corresponding to the same position of the target object in two point clouds to be matched, and the two points in the matching point pair can be overlapped with each other in theory after point cloud registration.
The higher the matching degree of two points in the matching point pair is, the more accurate the transformation relation of the cloud of the point to be registered is, and the more accurate registration result can be obtained.
In the embodiment of the application, the matching point pairs are screened on the basis of feature point extraction.
Firstly, the characteristic points of the first point cloud and the characteristic points of the second point cloud are screened by using curvature constraint conditions to obtain a plurality of matching point pairs.
For a matching point pair in two point clouds to be registered, for example, the point s in the first point cloud and the point d in the second point cloud, their local features in the respective point clouds are usually similar, i.e., the type of local curved surface to be fitted needs to be consistent and the degree and direction of curvature of the curved surface should also be consistent.
Curvature constraints may include:
wherein s and d are respectively matching points in two point clouds to be registered (a first point cloud and a second point cloud), Ks and Kd are respectively gaussian curvatures of the two points, Hs and Hd are respectively average curvatures of the two points, k1(s), k1(d), k2(s), and k2(d) are respectively two main curvatures of the two points, and epsilon is a similarity threshold (which may be 0.1, for example).
In some possible cases, there may be multiple correspondences between points in the point cloud after being screened by the curvature constraint.
And S205, determining at least one matching point pair in the obtained multiple matching point pairs.
And determining at least one matching point pair in a plurality of matching point pairs obtained by screening by using the curvature constraint condition.
Each matching point pair includes one point in the first point cloud and one point in the second point cloud.
The determination of the at least one matching point pair is described below.
The point cloud to be matched comprises a first point cloud and a second point cloud, wherein points in the first point cloud are { s1, s2, …, sm }, and points in the second point cloud are { d1, d2, …, dn }.
First, two points are randomly selected from the first point cloud to form a first point pair { si, sj }, where i is 1,2, …, m, j is 1,2, …, m, and i is not equal to j.
A second point pair { di, dj } corresponding to the first point pair is determined in the second point cloud using a curvature constraint, wherein points di and dj in the second point pair and point si in the first point pair satisfy the curvature constraint, and points di and dj in the second point pair and point sj in the first point pair also satisfy the curvature constraint.
The first point pair { si, sj } and the second point pair { di, dj } form a four point pair { si, sj, di, dj }.
The first point pair { si, sj } is each point pair in the first point cloud, that is, all the point pairs in the first point cloud are traversed to obtain corresponding point pairs in the second point cloud, so as to obtain a two-point pair set E2.
The two-point pair set E2 includes a plurality of two-point pairs, each of which is composed of four points, including two points in the first point cloud and two points in the second point cloud.
And randomly determining a two-point pair in the obtained two-point pair set E2, and marking as a first two-point pair { si, sj, di, dj }, and then determining a second two-point pair { si ', sj', di ', dj' } in the two-point pair set E2, wherein the first two-point pair and the second two-point pair have no coincident points.
The first two point pairs { si, sj, di, dj } and the second two point pairs { si ', sj', di ', dj' } form a four point pair { si, sj, di, dj, si ', sj', di ', dj' }.
The first two-point pair { si, sj, di, dj } is each two-point pair in the first point cloud, that is, all two-point pairs in the two-point pair set E2 are traversed and corresponding two-point pairs are obtained, so as to obtain a four-point pair set E4.
The set of four-point pairs E4 includes a plurality of four-point pairs, each of which is composed of eight points, including four points in the first point cloud and four points in the second point cloud.
Randomly determining a four-point pair in the obtained four-point pair set E4, marking as a first four-point pair { si, sj, di, dj, si ', sj', di ', dj' }, and then determining a second four-point pair { si ', sj', di ', dj', si ', sj', di ', dj' } in the four-point pair set E4, wherein points which are not coincident in the first four-point pair and the second four-point pair are not coincident.
The first four point pairs { si, sj, di, dj, si ', sj', di ', dj' } and the second four point pairs { si ", sj", di ", dj", si '", sj'", di '", dj'" } form an eight point pair { si, sj, di, dj, si ', sj', di ', dj', si ", sj", di ", dj", si '", sj'", di '", dj'" }.
The first four point pairs { si, sj, di, dj, si ', sj', di ', dj' } are each four point pairs in the first point cloud, that is, all four point pairs in the four point pair set E4 are traversed and corresponding four point pairs are obtained, so as to obtain an eight point pair set E8.
The eight-point pair set E8 includes eight point pairs, each of which consists of sixteen points, including eight points in the first point cloud and eight points in the second point cloud.
After the eight-point pair set E8 is obtained, the root mean square distance is calculated for each of the eight point pairs included in the eight-point pair set E8.
For eight point pairs { s1, s2, d1, d2, s3, s4, d3, d4, s5, s6, d5, d6, s7, s8, d7, d8} contained in the eight point pair set E8, the root mean square distance is:
the root mean square distance of each eight point pair in the eight point pair set E8 is obtained according to the formula of the root mean square distance, and each eight point pair in the eight point pair set E8 is arranged in ascending order according to the size of the root mean square distance.
A certain number of eight point pairs are determined from the eight point pair set E8, and for example, after the above-described ascending order, the first 10% of the eight point pairs are selected and subjected to the subsequent processing.
The top 10% of the eight point pairs may not be one, and the best eight point pairs need to be determined, resulting in at least one matching point pair.
And the transformation relation of the point clouds to be registered is used for converting the coordinate systems of the two point clouds to be matched, and the transformation relation comprises a group of rotation matrixes R and translation vectors T.
The following processing is performed for each of the eight point pairs of the first 10%:
and calculating a transformation relation { R, t } by using a singular value decomposition method, transforming the first point cloud into a coordinate system of the second point cloud by using the transformation relation { R, t }, and then calculating the overall error between the first point cloud and the second point cloud in the coordinate system of the second point cloud.
Singular value decomposition (singular value decomposition): and obtaining the transformation relation of the two data sets through singular value decomposition.
The errors of each point of the second point cloud in the coordinate system of the second point cloud are as follows: in the coordinate system of the second point cloud, the distance from the point of the second point cloud to the nearest point of the point in the first point cloud; the overall error is the sum of the errors of the points in the second point cloud.
And determining the eight point pairs with the minimum overall error as the optimal eight point pairs.
The optimal eight point pairs include sixteen points, including eight points in the first point cloud and eight points in the second point cloud, including at least one matching point pair.
The matching point pair in the optimal eight point pairs is the at least one matching point pair and is used for determining the transformation relation of the two point clouds to be matched.
In the technical scheme of the embodiment of the application, possible corresponding relations of each point of the point cloud are obtained preliminarily by using curvature constraint conditions, all matching relations are determined by combining the point pairs, then the distance root mean square error of each point pair is calculated and screened, the subsequent calculation amount is reduced, and finally the optimal eight matching point pairs are selected by the point cloud overall error.
S206, determining the transformation relation between the first point cloud and the second point cloud by using the at least one matching point pair.
And the transformation relation of the point clouds to be registered is used for converting the coordinate systems of the two point clouds to be matched, and the transformation relation comprises a group of rotation matrixes R and translation vectors T.
And S207, transforming the first point cloud to a coordinate system of the second point cloud by using the obtained transformation relation to obtain a point cloud registration result.
The point cloud registration process can be divided into two links of coarse registration and fine registration, wherein the coarse registration is the registration of the point cloud under the condition that the relative position of the point cloud is completely unknown, and good initial conditions can be provided for subsequent registration. The point cloud registration result can be regarded as a result of coarse registration.
And S208, processing the obtained point cloud registration result by using an iterative nearest neighbor algorithm to obtain a final point cloud registration result.
And obtaining a final point cloud registration result by using an iterative nearest neighbor algorithm on the coarse registration result.
Iterative Closest Points Algorithm, ICPA/ICP): and calculating a transformation relation by taking the nearest point of each point in the point cloud on the other point cloud as a corresponding point set, and continuously iterating to perform coordinate transformation and transformation relation calculation until the registration result of the two point clouds is optimal.
And (3) using an iterative nearest neighbor algorithm to realize fine registration after coarse registration so as to minimize the spatial position difference between the point clouds and optimize the registration effect.
The embodiment of the application also provides a registration result of different point cloud registration modes.
Referring to fig. 3A and 3B, fig. 3A is a schematic view of a point cloud distribution in a first posture according to an embodiment of the present disclosure, and fig. 3B is a schematic view of a point cloud distribution in a second posture according to an embodiment of the present disclosure.
The first posture and the second posture are point clouds obtained by measuring the target object at different angles respectively.
The first pose is distinguished from the second pose, the first pose point cloud comprising 35947 points and the second pose point cloud comprising 29786 points.
And respectively carrying out point cloud registration on the first attitude point cloud and the second attitude point cloud by using two modes.
Firstly, feature points are obtained by using a fixed threshold value mode, and point cloud registration is performed on the first attitude point cloud and the second attitude point cloud by using a method of randomly combining and screening matching point pairs, so that a first point cloud registration result is obtained, as shown in fig. 4A.
Referring to fig. 4A, fig. 4A is a schematic diagram of a first point cloud registration result according to an embodiment of the present disclosure.
By using the point cloud registration method provided by the embodiment of the application, the point cloud registration is performed on the first attitude point cloud and the second attitude point cloud, and a second point cloud registration result is obtained, as shown in fig. 4B.
Referring to fig. 4B, fig. 4B is a schematic diagram of a second point cloud registration result according to an embodiment of the present disclosure.
Please refer to table 1, where table 1 is an index of the registration result of the two point clouds.
TABLE 1 Point cloud registration result index
| Registration results | Root mean square error | Time of registration(s) |
| First point cloud registration result | 0.0029 | 4.9797 |
| Second point cloud registration result | 0.000361 | 4.5320 |
As can be seen from table 1, the point cloud registration method provided in the embodiment of the present application has better results in two important indicators, namely, the registration accuracy (root mean square error) and the registration time.
At present, a coarse registration algorithm based on feature matching calculates the feature searching transformation relation of whole or partial point clouds, and a good coarse registration effect is difficult to achieve under the scenes that some point clouds are sparse or partially missing.
The embodiment of the application also provides a registration result of different point cloud registration modes on the sparse point cloud.
Referring to fig. 5A and 5B, fig. 5A is a schematic view of a point cloud distribution in a third posture according to an embodiment of the present disclosure, and fig. 5B is a schematic view of a point cloud distribution in a fourth posture according to an embodiment of the present disclosure.
And the point cloud of the third posture and the point cloud of the fourth posture are both sparse point clouds, and the registration is carried out by utilizing the two methods respectively.
Sparse point clouds refer to point clouds with a small number of points.
The third posture and the fourth posture are point clouds obtained by measuring the target object at different angles respectively.
The third pose is distinguished from the fourth pose, the third pose point cloud comprising 3595 points and the fourth pose point cloud comprising 2397 points.
And respectively carrying out point cloud registration on the third attitude point cloud and the fourth attitude point cloud by using the two modes.
Referring to fig. 6A and 6B, fig. 6A is a schematic diagram of a third point cloud registration result provided in the embodiment of the present application, and fig. 6B is a schematic diagram of a fourth point cloud registration result provided in the embodiment of the present application.
The third point cloud registration result is obtained by performing point cloud registration on the third attitude point cloud and the fourth attitude point cloud by using the fixed threshold and the method for screening the random combined matching point pairs, as shown in fig. 6A.
The fourth point cloud registration result is obtained by performing point cloud registration on the third attitude point cloud and the fourth attitude point cloud by using the point cloud registration method provided by the embodiment of the present application, and is shown in fig. 6B.
Please refer to table 2, table 2 is an index of the registration result of the two point clouds.
TABLE 2 Point cloud registration result index
| Registration results | Root mean square error | Time of registration(s) |
| Third point cloud registration result | 0.0121 | 0.9856 |
| Fourth point cloud registration result | 0.0023 | 0.6911 |
As can be seen from table 2, under the sparse point cloud condition, the point cloud registration method provided in the embodiment of the present application has a better result in two important indexes, namely, registration accuracy (root mean square error) and registration time.
By adopting the technical scheme of the embodiment of the application, the characteristic points in the point cloud to be registered are determined by utilizing the curvature of the middle point of the point cloud to be registered and the self-adaptive threshold, and compared with the method of determining the characteristic points by utilizing the fixed threshold, the self-adaptability of the curvature threshold can be utilized to obtain a proper number of characteristic points, so that the accuracy of point cloud matching is improved, and meanwhile, the calculation amount is balanced; by adopting the technical scheme of the embodiment of the application, the characteristic points of the positions with large overall undulation and the characteristic points of the positions with small overall undulation but local undulation can be extracted.
The technical scheme of the embodiment of the application has the advantage that the distribution range of the extracted characteristic points is wide. Some points have a low curvature when viewed as a whole in the cloud of points, but may have a peak locally, and these points may well reflect the local features of the cloud of points. According to the technical scheme of the embodiment of the application, the points can be well reserved as the characteristic points in the self-adaptive threshold mode, so that the characteristic points can be extracted from the point cloud at each part of the point cloud.
At present, a rough registration algorithm based on exhaustive search of matching point pairs exists, and specifically, a transformation relation which minimizes an error function is selected or a transformation relation which can generate the most corresponding point pairs is selected for rough registration by traversing the whole point cloud combination. However, the method has a long calculation link and high time complexity. According to the technical scheme, the curvature condition constraint, the matching point pair combination and the root mean square error of the point pairs are used, the subsequent overall error calculation amount can be reduced, the matching point pair calculation is efficient, the accuracy of point cloud registration cannot be greatly lost, and a better point cloud registration result can be obtained in a shorter time.
The embodiment of the application also provides a point cloud registration device.
Referring to fig. 7, fig. 7 is a schematic structural diagram of a point cloud registration apparatus according to an embodiment of the present disclosure.
As shown in fig. 7, the apparatus 200 for point cloud registration of the embodiment of the present application includes the following units:
a curvature determining unit 201 for determining curvatures of points in the first point cloud and points in the second point cloud;
the feature point determining unit 202 is configured to determine a curvature threshold of the target point according to a mean and a variance of curvatures of midpoints in a neighborhood of the target point, and determine the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point, so as to obtain a feature point of the first point cloud and a feature point of the second point cloud; the target point is each point in the first point cloud and the second point cloud;
the matching point determining unit 203 is configured to screen feature points of the first point cloud and feature points of the second point cloud by using curvature constraint conditions to obtain a plurality of matching point pairs; wherein, a matching point pair comprises a characteristic point of a first point cloud and a characteristic point of a second point cloud;
a transformation relation determining unit 204 that determines a transformation relation between the first point cloud and the second point cloud using at least one of the plurality of matching point pairs;
the point cloud registration unit 205 transforms the first point cloud to the coordinate system of the second point cloud by using a transformation relation, and obtains a point cloud registration result.
The units included in the device 200 for point cloud registration can achieve the same technical effects as the method for point cloud registration in the above embodiments, and are not described here again to avoid repetition.
The embodiment of the application also provides the electronic equipment.
Please refer to fig. 8, fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.
As shown in fig. 8, an electronic device 300 according to an embodiment of the present application includes a processor 301 and a memory 302, where the memory 302 stores codes, and the processor 301 is configured to call the codes stored in the memory 302 to perform any one of the above methods for point cloud registration.
The units included in the electronic device 300 can achieve the same technical effects as the point cloud registration method in the above embodiments, and are not described herein again to avoid repetition.
In an embodiment of the present application, a computer-readable storage medium is further provided, where the computer-readable storage medium is used to store a computer program, and the computer program is used to execute the above method for point cloud registration, and can achieve the same technical effect, and in order to avoid repetition, details are not repeated here. The computer-readable storage medium may be a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method of point cloud registration, the method comprising:
determining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud;
determining a curvature threshold of a target point according to a mean value and a variance of curvatures of midpoints in a neighborhood of the target point, and determining the target point as a feature point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point, so as to obtain a feature point of the first point cloud and a feature point of the second point cloud; wherein the target point is each point in the first point cloud and the second point cloud;
screening the characteristic points of the first point cloud and the characteristic points of the second point cloud by using curvature constraint conditions to obtain a plurality of matching point pairs; wherein one of the matching point pairs comprises a feature point of the first point cloud and a feature point of the second point cloud;
determining a transformation relationship of the first point cloud and the second point cloud using at least one of the plurality of pairs of matching points;
and transforming the first point cloud to a coordinate system of the second point cloud by using the transformation relation to obtain a point cloud registration result.
2. The method of claim 1, further comprising, prior to said determining a transformation relationship for the first point cloud and the second point cloud using at least one of the plurality of pairs of matching points:
determining a first point pair from the first point cloud, determining a second point pair corresponding to the first point pair from the second point cloud by using the curvature constraint condition, and determining a two-point pair group formed by the first point pair and the second point pair to obtain a two-point pair set comprising a plurality of two point pairs; wherein each point of the second pair of points and each point of the first pair of points satisfies the curvature constraint, the first pair of points being each pair of points in the first point cloud;
determining a first two-point pair and a second two-point pair from the two-point pair set, and determining a four-point pair formed by the first two-point pair and the second two-point pair to obtain a four-point pair set containing a plurality of four-point pairs; wherein the first two-point pair is each point pair in the set of two-point pairs that does not coincide with a point in the second two-point pair;
determining a first four-point pair and a second four-point pair from the four-point pair set, and determining an eight-point pair formed by the first four-point pair and the second four-point pair to obtain an eight-point pair set containing a plurality of eight-point pairs; wherein the first pair of four points and the second pair of four points have no coincident points, the first pair of four points being each pair of four points in the set of two pairs of points;
and determining four point pairs with preset number from the eight point pair set to obtain the at least one matching point pair.
3. The method of claim 1, wherein determining the curvature of each point in the first point cloud and the second point cloud comprises:
determining a curved surface of a first point, wherein the curved surface of the first point is obtained by fitting each point in the neighborhood of the first point, and the first point is each point in the first point cloud and the second point cloud;
determining a gaussian curvature of the curved surface of the first point and an average curvature of the curved surface of the first point;
and determining the curvature of the first point according to the Gaussian curvature of the curved surface of the first point and the average curvature of the curved surface of the first point.
4. The method of claim 1, wherein the curvature threshold of the target point is determined from a mean and variance of curvatures of midpoints in a neighborhood of the target point; the target point is a point in the first point cloud and the second point cloud, and the method specifically comprises the following steps:
determining a mean and a variance of curvatures of points in a neighborhood of the target point;
summing the mean value and the target variance to obtain a curvature threshold of the target point; wherein the target variance is a product of the variance and a preset coefficient.
5. The method of claim 1, further comprising, before said screening the feature points of the first point cloud and the feature points of the second point cloud using curvature constraints to obtain a plurality of pairs of matching points:
determining a similarity threshold;
and determining the curvature constraint condition according to the similarity threshold.
6. The method of claim 1, further comprising, prior to said determining the curvature of each point in the first and second point clouds:
removing outliers in the first point cloud and removing outliers in the second point cloud; wherein an outlier in the first point cloud is a point in the first point cloud that is a distance greater than a first distance threshold from other points in the first point cloud and an outlier in the second point cloud is a point in the second point cloud that is a distance greater than a second distance threshold from other points in the second point cloud.
7. The method of claim 1, further comprising, after said transforming the first point cloud to the coordinate system of the second point cloud using the transformation relationship, obtaining a point cloud registration result:
and processing the point cloud registration result by using an iterative nearest neighbor algorithm to obtain a processed registration result.
8. An apparatus for point cloud registration, the apparatus comprising:
a curvature determining unit for determining the curvature of each point in the first point cloud and the curvature of each point in the second point cloud;
the characteristic point determining unit is used for determining a curvature threshold of the target point according to the mean value and the variance of the curvature of the midpoint in the neighborhood of the target point, and determining the target point as the characteristic point of the point cloud to which the target point belongs when the curvature of the target point is greater than the curvature threshold of the target point, so as to obtain the characteristic point of the first point cloud and the characteristic point of the second point cloud; wherein the target point is each of the first point cloud and the second point cloud;
a matching point determining unit, configured to screen feature points of the first point cloud and feature points of the second point cloud by using a curvature constraint condition to obtain a plurality of matching point pairs; wherein one of the matching point pairs comprises one of the feature points of the first point cloud and one of the feature points of the second point cloud;
a transformation relationship determination unit that determines a transformation relationship between the first point cloud and the second point cloud using at least one of the plurality of matching point pairs;
and the point cloud registration unit is used for transforming the first point cloud to the coordinate system of the second point cloud by utilizing the transformation relation to obtain a point cloud registration result.
9. An electronic device comprising a processor and a memory, wherein the memory stores code and the processor is configured to invoke the code stored in the memory to perform the method of any of claims 1 to 7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium is used to store a computer program for performing the method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210384544.7A CN114742868A (en) | 2022-04-13 | 2022-04-13 | Point cloud registration method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210384544.7A CN114742868A (en) | 2022-04-13 | 2022-04-13 | Point cloud registration method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114742868A true CN114742868A (en) | 2022-07-12 |
Family
ID=82282077
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210384544.7A Pending CN114742868A (en) | 2022-04-13 | 2022-04-13 | Point cloud registration method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114742868A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115100258A (en) * | 2022-08-29 | 2022-09-23 | 杭州三坛医疗科技有限公司 | Hip joint image registration method, device, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645170A (en) * | 2009-09-03 | 2010-02-10 | 北京信息科技大学 | Precise registration method of multilook point cloud |
| CN110473239A (en) * | 2019-08-08 | 2019-11-19 | 刘秀萍 | A kind of high-precision point cloud registration method of 3 D laser scanning |
| CN112529945A (en) * | 2020-11-17 | 2021-03-19 | 西安电子科技大学 | Registration method for multi-view three-dimensional ISAR scattering point set |
-
2022
- 2022-04-13 CN CN202210384544.7A patent/CN114742868A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645170A (en) * | 2009-09-03 | 2010-02-10 | 北京信息科技大学 | Precise registration method of multilook point cloud |
| CN110473239A (en) * | 2019-08-08 | 2019-11-19 | 刘秀萍 | A kind of high-precision point cloud registration method of 3 D laser scanning |
| CN112529945A (en) * | 2020-11-17 | 2021-03-19 | 西安电子科技大学 | Registration method for multi-view three-dimensional ISAR scattering point set |
Non-Patent Citations (4)
| Title |
|---|
| 朱德海: "点云库PCL学习教程", 31 October 2012, 北京航空航天大学出版社, pages: 189 * |
| 王晋疆: "一种基于点签名的散乱点云特征点检测方法", 计算机工程, 31 July 2014 (2014-07-31), pages 174 - 178 * |
| 薛耀红;梁学章;马婷;梁英;车翔玖;: "扫描点云的一种自动配准方法", 计算机辅助设计与图形学学报, no. 02, 15 February 2011 (2011-02-15), pages 223 - 231 * |
| 韩磊;陈宇;王春阳;于艳鑫;: "一种自适应点云特征点提取算法", 现代计算机(专业版), no. 11, 15 April 2018 (2018-04-15), pages 48 - 52 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115100258A (en) * | 2022-08-29 | 2022-09-23 | 杭州三坛医疗科技有限公司 | Hip joint image registration method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105184801B (en) | It is a kind of based on multi-level tactful optics and SAR image high-precision method for registering | |
| CN111160229B (en) | SSD network-based video target detection method and device | |
| CN108550166B (en) | Spatial target image matching method | |
| CN112347550A (en) | Coupling type indoor three-dimensional semantic graph building and modeling method | |
| Soltanpour et al. | Multimodal 2D–3D face recognition using local descriptors: pyramidal shape map and structural context | |
| CN110047133A (en) | A kind of train boundary extraction method towards point cloud data | |
| CN111340862A (en) | Point cloud registration method and device based on multi-feature fusion and storage medium | |
| CN112884820A (en) | Method, device and equipment for training initial image registration and neural network | |
| CN117274072A (en) | Point cloud denoising method and device based on two-dimensional multi-modal range image | |
| CN107194364B (en) | Huffman-L BP multi-pose face recognition method based on divide and conquer strategy | |
| CN114742868A (en) | Point cloud registration method and device and electronic equipment | |
| CN114332172B (en) | Laser point cloud registration method based on covariance matrix improvement | |
| CN113723294B (en) | Data processing method and device and object recognition method and device | |
| CN107194994B (en) | Method and device for reconstructing cylindrical surface by using point cloud data without calibration curved surface | |
| CN110264562B (en) | Automatic calibration method for feature points of skull model | |
| CN111080649A (en) | Image segmentation processing method and system based on Riemann manifold space | |
| CN113610971B (en) | Fine-grained three-dimensional model construction method and device and electronic equipment | |
| CN116630662A (en) | Feature point mismatching eliminating method applied to visual SLAM | |
| CN113033270B (en) | 3D object local surface description method and device adopting auxiliary axis and storage medium | |
| CN111401385B (en) | Similarity calculation method for image local topological structure feature descriptors | |
| CN115631370A (en) | Identification method and device of MRI (magnetic resonance imaging) sequence category based on convolutional neural network | |
| CN108304862A (en) | A kind of map building polygon mode identification method based on wavelet transformation | |
| CN118365802B (en) | Three-dimensional point cloud reconstruction method and device and computer readable storage medium | |
| CN113256693A (en) | Multi-view registration method based on K-means and normal distribution transformation | |
| CN118675213B (en) | Automatic identification system and method for AI (advanced technology attachment) face key points |
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 |