CN113205547A - Point cloud registration method, bone registration method, device, equipment and storage medium - Google Patents

Point cloud registration method, bone registration method, device, equipment and storage medium Download PDF

Info

Publication number
CN113205547A
CN113205547A CN202110292830.6A CN202110292830A CN113205547A CN 113205547 A CN113205547 A CN 113205547A CN 202110292830 A CN202110292830 A CN 202110292830A CN 113205547 A CN113205547 A CN 113205547A
Authority
CN
China
Prior art keywords
point cloud
cloud set
registered
registration
point
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
Application number
CN202110292830.6A
Other languages
Chinese (zh)
Inventor
张逸凌
刘星宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changmugu Medical Technology Qingdao Co ltd
Longwood Valley Medtech Co Ltd
Original Assignee
Changmugu Medical Technology Qingdao Co ltd
Longwood Valley Medtech Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Changmugu Medical Technology Qingdao Co ltd, Longwood Valley Medtech Co Ltd filed Critical Changmugu Medical Technology Qingdao Co ltd
Priority to CN202110292830.6A priority Critical patent/CN113205547A/en
Publication of CN113205547A publication Critical patent/CN113205547A/en
Pending legal-status Critical Current

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
    • G06T7/337—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G06T7/0014—Biomedical image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • 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
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30008—Bone

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to the technical field of three-dimensional point cloud modeling, and discloses a point cloud registration method, a bone registration method, a device, equipment and a storage medium. Wherein, the method comprises the following steps: acquiring a point cloud set to be registered and a reference point cloud set; calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method; based on a first reference coordinate system and a second reference coordinate system, carrying out initial registration on a point cloud set to be registered and a reference point cloud set; based on a multi-dimensional binary search tree algorithm, searching a point closest to the cloud set of points to be registered in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs; respectively calculating the direction vector included angles among the corresponding point pairs; and performing fine registration on the point cloud set to be registered and the reference point cloud set based on a preset included angle threshold and a direction vector included angle. By implementing the method, the registration method is prevented from falling into a local optimal solution, and the registration efficiency and the registration accuracy of the point cloud are improved.

Description

Point cloud registration method, bone registration method, device, equipment and storage medium
Technical Field
The invention relates to the technical field of three-dimensional point cloud modeling, in particular to a point cloud registration method, a bone registration method, a device, equipment and a storage medium.
Background
In the three-dimensional point cloud modeling process, three-dimensional laser point cloud registration is one of the key problems to be solved. For three-dimensional laser point cloud registration, a classical Iterative closest point algorithm (ICP) is usually adopted to perform optimal rigid body transformation, and rotation parameters and translation parameters between point cloud data to be registered and reference point cloud data are calculated, so that the point cloud data to be registered and the reference point cloud data meet the convergence accuracy of correct registration, and optimal registration of the point cloud data to be registered and the reference point cloud data is realized. However, the existing iterative closest point algorithm performs iterative computation based on the average distance and the given distance threshold until the registration requirement is met, so that the above registration method easily falls into a local optimal solution, resulting in low registration efficiency between the cloud data of the point to be registered and the cloud data of the reference point.
Disclosure of Invention
In view of this, embodiments of the present invention provide a point cloud registration method, a bone registration method, an apparatus, a device, and a storage medium, so as to solve the problem that the existing closest point algorithm is likely to fall into a local optimal solution, which results in low registration efficiency between point cloud data to be registered and reference point cloud data.
According to a first aspect, an embodiment of the present invention provides a point cloud registration method, including the following steps: acquiring a point cloud set to be registered and a reference point cloud set; calculating a first reference coordinate system corresponding to the cloud set of points to be registered and a second reference coordinate system corresponding to the cloud set of reference points based on a principal component analysis method; performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system; based on a multi-dimensional binary search tree algorithm, searching points closest to the cloud set of points to be registered in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs; respectively calculating the direction vector included angles between the multiple groups of corresponding point pairs; and performing fine registration on the point cloud set to be registered and the reference point cloud set based on a preset included angle threshold and the direction vector included angle.
The point cloud registration method provided by the embodiment of the invention is characterized in that a first reference coordinate system corresponding to a cloud set of points to be registered and a second reference coordinate system corresponding to a cloud set of reference points are calculated based on a principal component analysis method, and the initial registration of the cloud set of points to be registered and the cloud set of reference points is realized through the first reference coordinate system and the second reference coordinate system, so that the cloud set of points to be registered and the cloud set of reference points are approximately coincided quickly, and the registration method is prevented from falling into a local optimal solution. Based on the multi-dimensional binary search tree algorithm, the point closest to the cloud set of the points to be registered is searched in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs between the reference point cloud set and the cloud set of the points to be registered, and the search speed of the corresponding point pairs is improved. By calculating the direction vector included angles among the multiple groups of corresponding point pairs and based on the relationship between the direction vector included angles and the preset included angle threshold value, the point cloud set to be registered and the reference point cloud set are subjected to fine registration, so that the registration efficiency and the registration accuracy of the point cloud set to be registered and the reference point cloud set are improved.
With reference to the first aspect, in a first implementation manner of the first aspect, the calculating, based on a principal component analysis method, a first reference coordinate system corresponding to the cloud set of points to be registered and a second reference coordinate system corresponding to the cloud set of reference points includes: acquiring a first point cloud gravity center of the cloud set of points to be registered and a second point cloud gravity center of the cloud set of reference points; respectively calculating a first main direction of the cloud set of points to be registered and a second main direction of the cloud set of reference points by adopting the main component analysis method; the first main direction is 3 first eigenvectors corresponding to the cloud set of points to be registered, and the second main direction is 3 second eigenvectors corresponding to the cloud set of reference points; determining a first reference coordinate system corresponding to the cloud set of points to be registered based on the first point cloud gravity center and the first main direction; and determining a second reference coordinate system corresponding to the reference point cloud set based on the second point cloud gravity center and the second main direction.
The point cloud registration method provided by the embodiment of the invention is characterized in that a first reference coordinate system corresponding to a point cloud set to be registered and a second reference coordinate system corresponding to a reference point cloud set are calculated based on a principal component analysis method, and the initial registration of the point cloud set to be registered and the reference point cloud set is realized through the first reference coordinate system and the second reference coordinate system, so that the approximate coincidence of the point cloud set to be registered and the reference point cloud set is quickly realized. And searching a point closest to the cloud set of the points to be registered in the initially registered reference point cloud set to obtain a plurality of groups of corresponding point pairs between the cloud set of the reference points and the cloud set of the points to be registered, calculating direction vector included angles between the plurality of groups of corresponding point pairs, and performing precise registration on the cloud set of the points to be registered and the cloud set of the reference points through the relation between the direction vector included angles and a preset included angle threshold value, so that the registration method is prevented from falling into a local optimal solution, and the precise registration of the cloud set of the points to be registered and the cloud set of the reference points is improved.
With reference to the first aspect, in a second implementation manner of the first aspect, the initially registering the cloud set to be registered and the cloud set of reference points based on the first reference coordinate system and the second reference coordinate system includes: calculating the similarity between the point cloud set to be registered and the reference point cloud set, and determining a target point cloud set to be registered and a target reference point cloud set with the maximum similarity; adjusting the first reference coordinate system and the second reference coordinate system corresponding to the target point cloud set to be registered and the target reference point cloud set to the same target coordinate system; establishing a first minimum bounding box corresponding to the target cloud set of points to be registered and a second minimum bounding box corresponding to the target cloud set of reference points in the target coordinate system; calculating a coincidence volume of the first minimum bounding box and the second minimum bounding box; judging whether the coincidence volume meets a coincidence threshold value; and when the coincidence volume meets the coincidence threshold, judging that the target cloud set to be registered and the target reference point cloud set are initially registered.
With reference to the second implementation manner of the first aspect, in a third implementation manner of the first aspect, the initially registering the cloud set to be registered and the cloud set of reference points based on the first reference coordinate system and the second reference coordinate system further includes: and when the coincidence volume does not meet the coincidence threshold, rotating the coordinate axis of the target coordinate system by a preset angle until the coincidence volumes of the first minimum bounding box and the second minimum bounding box meet the coincidence threshold.
According to the point cloud registration method provided by the embodiment of the invention, the point cloud set to be registered and the reference point cloud set with the maximum similarity are determined by calculating the similarity between the cloud set to be registered and the reference point cloud set, the first reference coordinate system and the second reference coordinate system corresponding to the point cloud set are adjusted to the same target coordinate system, the registration volume between the first minimum bounding box corresponding to the cloud set to be registered and the second minimum bounding box corresponding to the reference point cloud set is calculated, and the coordinate axes of the target coordinates are rotated until the registration volume meets the registration threshold according to the relation between the registration volume and the registration threshold, so that the initial registration of the cloud set to be registered and the reference point cloud set is judged, and the accuracy of the initial registration of the cloud set to be registered and the reference point cloud set is ensured.
With reference to the first aspect, in a fourth implementation manner of the first aspect, the separately calculating direction vector included angles between the sets of corresponding point pairs includes: calculating a first tangent plane formed by fitting the points in the point cloud set to be registered with the adjacent points of the point cloud set to be registered and a second tangent plane formed by fitting the points in the reference point cloud set with the adjacent points of the point cloud set to be registered; calculating a first normal vector corresponding to a point in the point cloud set to be registered based on the first tangent plane; calculating a second normal vector corresponding to a point in the reference point cloud set based on the second tangent plane; and calculating a direction vector included angle between the first normal vector and the second normal vector.
With reference to the fourth implementation manner of the first aspect, in a fifth implementation manner of the first aspect, the performing fine registration on the point cloud set to be registered and the reference point cloud set based on a preset included angle threshold and the direction vector included angle includes: judging whether the included angle of the direction vector is smaller than a preset included angle threshold value or not; and when the included angle of the direction vector is smaller than the preset included angle threshold value, judging that the point cloud set to be registered and the reference point cloud set are accurately registered.
With reference to the fifth implementation manner of the first aspect, in a sixth implementation manner of the first aspect, the performing fine registration on the point cloud set to be registered and the reference point cloud set based on the direction vector included angle further includes: when the included angle of the direction vector is equal to or larger than the preset included angle threshold value, judging that the corresponding point pair is an error point pair; and rejecting the error point pairs.
According to the point cloud registration method provided by the embodiment of the invention, the point in the point cloud set to be registered and the point in the reference point cloud set are vectorized to obtain the first normal vector corresponding to the point in the point cloud set to be registered and the second normal vector corresponding to the point in the reference point cloud set, so that the direction vector included angle between the first normal vector and the second normal vector is determined. And when the included angle of the direction vector is equal to or larger than a preset included angle threshold value, judging that the corresponding point pair is an error point pair, and rejecting the error point pair until the included angles of the direction vectors of all the corresponding point pairs are smaller than the preset included angle threshold value, judging that the cloud set of the point to be registered and the cloud set of the reference point are accurately registered, thereby improving the registration efficiency of the cloud set of the point to be registered and the cloud set of the reference point.
With reference to the first aspect, in a seventh implementation manner of the first aspect, the method further includes: and adjusting the precisely registered cloud set of points to be registered and the reference point cloud set based on deep learning.
With reference to the seventh implementation manner of the first aspect, in an eighth implementation manner of the first aspect, the adjusting the finely registered cloud set of points to be registered and the cloud set of reference points based on deep learning includes: performing feature extraction on the precisely registered cloud set of points to be registered and the reference point cloud set based on a point cloud feature extraction network to obtain first point cloud features corresponding to the cloud set of points to be registered and second point cloud features corresponding to the reference point cloud set; predicting a matching relationship between the first point cloud feature and the second point cloud feature; and based on the matching relation, carrying out registration adjustment on the precisely registered cloud set of the point to be registered and the reference point cloud set by adopting a singular value decomposition method.
The point cloud registration method provided by the embodiment of the invention further adjusts the accurately registered cloud set of the point to be registered and the reference point cloud set through deep learning, specifically, performs feature extraction on the accurately registered cloud set of the point to be registered and the reference point cloud set based on a point cloud registration network to obtain a first point cloud feature corresponding to the cloud set of the point to be registered and a second point cloud feature corresponding to the reference point cloud set, performs matching relation between the first point cloud feature and the second point cloud feature, and performs registration adjustment on the accurately registered cloud set of the point to be registered and the reference point cloud set by using a singular value decomposition method based on the matching relation, so as to ensure that the registration accuracy between the cloud set of the point to be registered and the reference point cloud set reaches the optimum, and further improves the registration accuracy of the cloud set of the point to be registered and the reference point cloud set.
According to a second aspect, an embodiment of the present invention provides a bone registration method based on point cloud registration, including: acquiring a first point cloud set corresponding to a target registration bone and a second point cloud set corresponding to a reference registration bone; and registering the first point cloud set and the second point cloud set by using the point cloud registration method of the first aspect or any embodiment of the first aspect.
According to the bone registration method based on point cloud registration provided by the embodiment of the invention, the first point cloud set and the second point cloud set are initially registered based on the point cloud registration method, so that the first point cloud set and the reference point cloud set are approximately overlapped quickly, and the accurate registration between a target registration bone and a reference registration bone is completed. And searching points closest to the first point cloud set in the approximately coincident second point cloud set to obtain a plurality of groups of corresponding point pairs between the first point cloud set and the second point cloud set, and performing fine registration on the first point cloud set and the second point cloud set by calculating direction vector included angles between the plurality of groups of corresponding point pairs, so that the registration efficiency and the registration accuracy between the target registration bone and the reference registration bone are improved.
According to a third aspect, an embodiment of the present invention provides a point cloud registration apparatus, including: the first acquisition module is used for acquiring a cloud set of points to be registered and a cloud set of reference points; the first calculation module is used for calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method; the primary registration module is used for carrying out primary registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system; the corresponding module is used for searching a point which is closest to the cloud set of points to be registered in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs; the second calculation module is used for calculating the direction vector included angles among the multiple groups of corresponding point pairs respectively; and the fine registration module is used for performing fine registration on the point cloud set to be registered and the reference point cloud set based on the direction vector included angle.
The point cloud registration device provided by the embodiment of the invention calculates the first reference coordinate system corresponding to the cloud set of the point to be registered and the second reference coordinate system corresponding to the cloud set of the reference point based on the principal component analysis method, and realizes the initial registration of the cloud set of the point to be registered and the cloud set of the reference point through the first reference coordinate system and the second reference coordinate system so as to quickly realize the approximate coincidence of the cloud set of the point to be registered and the cloud set of the reference point, thereby avoiding the registration method from falling into the local optimal solution. Based on the multi-dimensional binary search tree algorithm, the point closest to the cloud set of the points to be registered is searched in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs between the reference point cloud set and the cloud set of the points to be registered, and the search speed of the corresponding point pairs is improved. By calculating the direction vector included angles among the multiple groups of corresponding point pairs and based on the relationship between the direction vector included angles and the preset included angle threshold value, the point cloud set to be registered and the reference point cloud set are precisely registered, so that the precise registration of the point cloud set to be registered and the reference point cloud set is improved.
According to a fourth aspect, an embodiment of the present invention provides a bone registration apparatus based on point cloud registration, the apparatus including: the second acquisition module is used for acquiring a first point cloud set corresponding to the target registration bone and a second point cloud set corresponding to the reference registration bone; a registration module, configured to register the first point cloud set and the second point cloud set by using the point cloud registration method described in the first aspect or any embodiment of the first aspect.
According to the bone registration device based on point cloud registration provided by the embodiment of the invention, the first point cloud set and the second point cloud set are initially registered based on a point cloud registration method, so that the first point cloud set and the reference point cloud set are approximately overlapped quickly, and the accurate registration between a target registration bone and a reference registration bone is completed. And searching points closest to the first point cloud set in the approximately coincident second point cloud set to obtain a plurality of groups of corresponding point pairs between the first point cloud set and the second point cloud set, and performing fine registration on the first point cloud set and the second point cloud set by calculating direction vector included angles between the plurality of groups of corresponding point pairs, so that the registration efficiency and the registration accuracy between the target registration bone and the reference registration bone are improved.
According to a fifth aspect, an embodiment of the present invention provides an electronic device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, and the processor executing the computer instructions to perform the point cloud registration method according to the first aspect or any embodiment of the first aspect, or to perform the bone registration method based on point cloud registration according to any embodiment of the second aspect or any embodiment of the second aspect.
According to a sixth aspect, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the point cloud registration method according to the first aspect or any embodiment of the first aspect, or execute the bone registration method based on point cloud registration according to any embodiment of the second aspect or the second aspect.
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, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a point cloud registration method according to an embodiment of the invention;
FIG. 2 is another flow chart of a point cloud registration method according to an embodiment of the invention;
FIG. 3 is another flow chart of a point cloud registration method according to an embodiment of the invention;
FIG. 4 is another flow chart of a point cloud registration method according to an embodiment of the invention;
FIG. 5 is a search diagram of a multidimensional binary search tree, according to an embodiment of the present invention;
FIG. 6 is a comparison diagram of different point cloud registration methods according to an embodiment of the invention;
FIG. 7 is a flow chart of a method of bone registration based on point cloud registration according to an embodiment of the present invention;
fig. 8 is a schematic view of registration between an acetabulum and a femoral head according to an embodiment of the invention;
fig. 9 is a block diagram of a point cloud registration apparatus according to an embodiment of the present invention;
FIG. 10 is a block diagram of a bone registration apparatus based on point cloud registration according to an embodiment of the present invention;
fig. 11 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For three-dimensional laser point cloud registration, a classical Iterative closest point algorithm (ICP) is usually adopted to perform optimal rigid body transformation, and rotation parameters and translation parameters between the cloud set data of the point to be registered and the cloud set data of the reference point are calculated, so that the cloud set data of the point to be registered and the cloud set data of the reference point meet the convergence precision of correct registration, and the optimal registration of the cloud set data of the point to be registered and the cloud set data of the reference point is realized. However, the existing iterative closest point algorithm performs iterative computation based on the average distance and the given distance threshold until the registration requirement is met, so that the above registration method is easily trapped in a local optimal solution, resulting in low registration efficiency between the cloud set data of the point to be registered and the cloud set data of the reference point.
Based on the above, aiming at the problem that the iterative closest point algorithm is easy to fall into the local optimal solution, a principal component analysis method is introduced in the initial registration process so as to quickly realize the approximate coincidence of the cloud set of the point to be registered and the cloud set of the reference point; and the accurate registration of the cloud set of the points to be registered and the cloud set of the reference points is carried out based on the multidimensional binary search tree and the direction vector included angle, the quick search is carried out through the multidimensional binary search tree, the search speed of corresponding point pairs between the cloud set of the reference points and the cloud set of the points to be registered is improved, the error point pairs are removed through the direction vector included angle, and the registration efficiency of the cloud set of the points to be registered and the cloud set of the reference points is improved.
In accordance with an embodiment of the present invention, there is provided an embodiment of a point cloud registration method, it is noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.
In this embodiment, a point cloud registration method is provided, which may be used in electronic devices, such as scanning devices, medical detection devices, and the like, fig. 1 is a flowchart of the point cloud registration method according to an embodiment of the present invention, and as shown in fig. 1, the flowchart includes the following steps:
and S11, acquiring a point cloud set to be registered and a reference point cloud set.
The cloud set of points to be registered is a three-dimensional laser point cloud data set acquired by the electronic equipment, and the cloud set of reference points is a point cloud data set with unchanged positions. The cloud set of points to be registered is acquired by scanning an object to be detected from different positions or angles by the electronic equipment. The reference point cloud set is obtained by simulating a virtual object corresponding to the object to be detected.
And S12, calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method.
And simplifying the point cloud set to be registered by adopting a Principal Component Analysis (PCA), namely performing dimensionality reduction on the point cloud set to be registered so as to retain important data in the point cloud set to be registered. And extracting principal components corresponding to the cloud set of the points to be registered from the cloud set of the points to be registered by adopting a principal component analysis method, and extracting principal components corresponding to the cloud set of the reference points from the reference point cloud set. The electronic device may generate a first reference coordinate system based on the gravity center of the cloud set of points to be registered and the principal component corresponding to the cloud set of points to be registered, and generate a second reference coordinate system based on the gravity center of the cloud set of reference points and the principal component corresponding to the cloud set of reference points.
And S13, performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system.
The initial registration is used to characterize the approximate coincidence of the cloud set of points to be registered and the cloud set of reference points. The electronic device may perform coordinate transformation on a point in the cloud set of points to be registered located in the first reference coordinate system and a point in the reference point cloud set located in the second reference coordinate system, so as to adjust the cloud set of points to be registered and the cloud set of reference points to the same coordinate system, and determine a coincidence volume of the cloud set of points to be registered and the cloud set of reference points. And adjusting the same coordinate system where the cloud set of the point to be registered and the cloud set of the reference point are located based on the coincidence volume of the cloud set of the point to be registered and the cloud set of the reference point to ensure that the cloud set of the point to be registered and the cloud set of the reference point can be approximately coincided, thereby realizing the initial registration of the cloud set of the point to be registered and the cloud set of the reference point.
And S14, based on the multi-dimensional binary search tree algorithm, searching the point closest to the point cloud set to be registered in the reference point cloud set after initial registration to obtain multiple groups of corresponding point pairs.
The multi-dimensional binary search tree algorithm is a k-d tree algorithm, after initial registration of the cloud set of the point to be registered and the cloud set of the reference point, approximate registration of the cloud set of the point to be registered and the cloud set of the reference point is achieved, but registration accuracy of the cloud set of the point to be registered and the cloud set of the reference point is still low, and the point cloud registration requirement is not met. At the moment, a point closest to the cloud set of the points to be registered can be quickly searched in the reference point cloud set by adopting a k-d tree algorithm, and a plurality of groups of corresponding point pairs between the cloud set of the points to be registered and the reference point cloud set are obtained. The time efficiency of searching the closest point by using the k-d tree algorithm is O (nlogn) level, so that the searching speed of the closest point of the point cloud can be greatly improved.
Specifically, firstly, the dividing lines are searched one by one along the X-axis direction of the coordinate axis, the average value of the X coordinate values of all searched points is calculated, the point closest to the average value is selected from the searched points, and the space is divided into two parts along the Y-axis direction of the coordinate axis at the point. Then, in the previously divided subspace, the partition lines are searched one by one along the Y-axis direction of the coordinate axis, and the subspace is divided into two parts. And then, the subspace divided in the previous link is divided along the X-axis direction of the coordinate axis, and the process is repeated in such a way, and the division is stopped when only one point is left in the divided subspace. As shown in FIG. 5, the points in FIG. 5 represent input points, circles represent query ranges, and numbers represent traversal order when querying.
And S15, respectively calculating the direction vector included angles between the corresponding point pairs.
And searching to obtain a plurality of groups of corresponding point pairs between the cloud sets of the points to be registered and the cloud sets of the reference points by using a k-d tree algorithm, wherein the corresponding point pairs still contain certain error points, and the registration accuracy of the cloud sets of the points to be registered and the cloud sets of the reference points is influenced. Vectorizing each point in the cloud set of points to be registered and the reference point cloud set respectively, determining a normal vector corresponding to each corresponding point pair, and calculating a direction vector included angle corresponding to the normal vector according to the normal vector of the corresponding point pair.
Specifically, the cosine of the included angle between the two normal vectors can be calculated based on the cosine formula of the included angle between the two normal vectors corresponding to the corresponding point pair, and the included angle of the direction vector corresponding to the two normal vectors can be determined according to the cosine value of the included angle. Of course, other calculation methods may be adopted to determine the direction vector included angle, which is not specifically limited in this application.
And S16, performing fine registration on the point cloud set to be registered and the reference point cloud set based on the preset included angle threshold and the direction vector included angle.
After initial registration, the cloud set of points to be registered and the cloud set of reference points are approximately overlapped, if the corresponding point pair is correct, the condition that the included angle of the direction vector is smaller than a given preset included angle threshold value is met, if the included angle of the direction vector is larger than the given preset included angle threshold value, the corresponding point pair can be judged to be an error point pair, the error point pair needs to be removed, and when all the corresponding point pairs are correct corresponding point pairs, the cloud set of points to be registered and the cloud set of reference points are judged to finish accurate registration.
The point cloud registration method provided in this embodiment calculates a first reference coordinate system corresponding to a cloud set of points to be registered and a second reference coordinate system corresponding to a cloud set of reference points based on a principal component analysis method, and realizes initial registration of the cloud set of points to be registered and the cloud set of reference points through the first reference coordinate system and the second reference coordinate system, so as to quickly realize approximate coincidence of the cloud set of points to be registered and the cloud set of reference points, thereby avoiding the registration method from falling into a local optimal solution. Based on the multi-dimensional binary search tree algorithm, the point closest to the cloud set of the points to be registered is searched in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs between the reference point cloud set and the cloud set of the points to be registered, and the search speed of the corresponding point pairs is improved. By calculating the direction vector included angles among the multiple groups of corresponding point pairs and based on the relationship between the direction vector included angles and the preset included angle threshold value, the point cloud set to be registered and the reference point cloud set are accurately registered, so that the registration efficiency of the point cloud set to be registered and the reference point cloud set is improved.
In this embodiment, a point cloud registration method is provided, which may be used in electronic devices, such as scanning devices, medical detection devices, and the like, fig. 2 is a flowchart of the point cloud registration method according to an embodiment of the present invention, and as shown in fig. 2, the flowchart includes the following steps:
and S21, acquiring a point cloud set to be registered and a reference point cloud set. For a detailed description, refer to the related description of step S11 corresponding to the above method embodiment, and are not repeated herein.
And S22, calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method.
Specifically, the step S22 may include the following steps:
s221, a first point cloud gravity center of the point cloud set to be registered and a second point cloud gravity center of the reference point cloud set are obtained.
The first point cloud gravity center is a gravity center point coordinate of the cloud set of the points to be registered, and the coordinate is obtained by calculating an average value of values of all points in the cloud set of the points to be registered, namely gravity center coordinate values of the XYZ three axes of the cloud set of the points to be registered are respectively calculated.
Similarly, the gravity center of the second point cloud is the gravity center point coordinate of the reference point cloud set, and can be obtained by calculating the average value of all the points in the reference point cloud set, that is, the gravity center coordinate values of the XYZ axes of the reference point cloud set are calculated respectively.
And S222, respectively calculating a first main direction of the cloud set of the point to be registered and a second main direction of the cloud set of the reference point by adopting a principal component analysis method, wherein the first main direction is 3 first eigenvectors corresponding to the cloud set of the point to be registered, and the second main direction is 3 second eigenvectors corresponding to the cloud set of the reference point.
Carrying out mean value removing processing on the cloud set of the point to be registered, and calculating a covariance matrix corresponding to the cloud set of the point to be registered and a characteristic value and a characteristic vector corresponding to the covariance matrix; calculating the number of eigenvalues of the covariance matrix which are larger than the threshold of the eigenvalue, and performing descending order arrangement on the obtained eigenvalues; removing the characteristic values in the back of the sequence, namely selecting the maximum 3 characteristic values; and combining the 3 eigenvalues, selecting a whitening matrix consisting of eigenvectors corresponding to the 3 eigenvalues, extracting three principal components, namely 3 first eigenvectors, from the whitening matrix, and taking the 3 first eigenvectors as a first principal direction of the point cloud set to be registered.
Similarly, 3 second eigenvectors corresponding to the reference point cloud set can be calculated, and the 3 second eigenvectors are used as the second main direction of the reference point cloud set.
S223, determining a first reference coordinate system corresponding to the cloud set of the point to be registered based on the gravity center of the first point cloud and the first main direction; and determining a second reference coordinate system corresponding to the reference point cloud set based on the second point cloud gravity center and the second main direction.
And taking the gravity center of the first point cloud as an origin, and taking the 3 first feature vectors corresponding to the first main direction as three coordinate axes of a coordinate system respectively, thereby establishing a first reference coordinate system corresponding to the cloud set of the points to be registered. Similarly, the gravity center of the second point cloud is used as an origin, and the 3 second feature vectors corresponding to the second principal direction are respectively used as three coordinate axes of a coordinate system, so that a first reference coordinate system corresponding to the cloud set of the point to be registered can be established.
And S23, performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system.
Specifically, the step S23 may include the following steps:
and S231, calculating the similarity between the cloud set of the point to be registered and the cloud set of the reference point, and determining the target cloud set of the point to be registered and the target cloud set of the reference point with the maximum similarity.
The principal component analysis method reflects the characteristic vector of the cloud set of the point to be registered which contributes most to the variance and the characteristic vector of the reference point cloud set which contributes most to the variance, so that the similarity between the cloud set of the point to be registered and the reference point cloud set can be calculated based on the characteristic vectors, and the target cloud set of the point to be registered and the target reference point cloud set which have the maximum similarity can be determined from the cloud set of the point to be registered and the reference point cloud set.
And S232, adjusting the first reference coordinate system and the second reference coordinate system corresponding to the target cloud set to be registered and the target reference cloud set to the same target coordinate system.
The target coordinate system is a coordinate system corresponding to the first reference coordinate system and the second reference coordinate system after coordinate conversion. Performing coordinate conversion on all points in a target point cloud set to be registered in a first reference coordinate system, and adjusting the points to a target coordinate system; and performing coordinate conversion on all points in the target reference point cloud set positioned in the second reference coordinate system, and adjusting the points to the target coordinate system. Thereby, the first reference coordinate system and the second reference coordinate system are adjusted to the same target coordinate system.
And S233, establishing a first minimum bounding box corresponding to the target cloud set of points to be registered and a second minimum bounding box corresponding to the target cloud set of reference points in the target coordinate system.
Because the first principal direction and the second principal direction calculated by the PCA algorithm may have a 180 ° difference, a first minimum bounding box corresponding to the target cloud set of points to be registered and a second minimum bounding box corresponding to the target cloud set of reference points need to be established. The person skilled in the art may determine the method for establishing the minimum bounding box empirically, as long as it is ensured that the first minimum bounding box can surround the target cloud set of points to be registered, and the second minimum bounding box can surround the target cloud set of reference points, which is not limited specifically herein.
And S234, calculating the coincidence volume of the first minimum bounding box and the second minimum bounding box.
And calculating the coincidence volume of a first minimum bounding box corresponding to the target cloud set to be registered and a second minimum bounding box corresponding to the target reference point cloud set after coordinate transformation so as to test whether the target cloud set to be registered and the target reference point cloud set are approximately coincident or not. The first minimum bounding box and the second minimum bounding box are both in the same target coordinate system, and therefore the coincident positions enclosed by the first minimum bounding box and the second minimum bounding box can be determined. And calculating the coincidence volume of the first minimum bounding box and the second minimum bounding box according to the coordinate values corresponding to the coincidence positions.
And S235, judging whether the coincidence volume meets a coincidence threshold value.
And comparing the coincidence volume with the coincidence threshold value, and determining whether the coincidence volume meets the coincidence threshold value. When the coincidence volume satisfies the coincidence threshold, step S236 is performed, otherwise step S237 is performed.
And S236, judging that the target cloud set to be registered and the target reference point cloud set are initially registered.
When the coincidence volume meets the coincidence threshold, the target cloud set of the point to be registered and the target cloud set of the reference point are approximately coincided, and it can be judged that the target cloud set of the point to be registered and the target cloud set of the reference point have completed initial registration.
And S237, rotating the coordinate axis of the target coordinate system by a preset angle until the coincidence volume of the first minimum bounding box and the second minimum bounding box meets the coincidence threshold.
The preset angle of rotation may be 180 °, and may be other angles, and those skilled in the art may rotate according to actual needs, and is not limited specifically herein. When the coincidence volume does not meet the coincidence threshold, the electronic device may rotate the coordinate axes of the target coordinate system until the target to-be-registered point cloud set and the target reference point cloud set substantially coincide. Specifically, the rotation may be 180 ° clockwise or 180 ° counterclockwise, and the present application is not limited to this.
And S24, based on the multi-dimensional binary search tree algorithm, searching the point closest to the point cloud set to be registered in the reference point cloud set after initial registration to obtain multiple groups of corresponding point pairs. For a detailed description, refer to the related description of step S14 corresponding to the above method embodiment, and are not repeated herein.
And S25, respectively calculating the direction vector included angles between the corresponding point pairs. For a detailed description, refer to the related description of step S15 corresponding to the above method embodiment, and are not repeated herein.
And S26, performing fine registration on the point cloud set to be registered and the reference point cloud set based on the preset included angle threshold and the direction vector included angle. For a detailed description, refer to the related description of step S16 corresponding to the above method embodiment, and are not repeated herein.
The point cloud registration method provided in this embodiment calculates a first reference coordinate system corresponding to the cloud set of the point to be registered and a second reference coordinate system corresponding to the cloud set of the reference point based on a principal component analysis method, and realizes initial registration of the cloud set of the point to be registered and the cloud set of the reference point through the first reference coordinate system and the second reference coordinate system, so as to quickly realize approximate coincidence of the cloud set of the point to be registered and the cloud set of the reference point. And searching a point closest to the cloud set of the points to be registered in the initially registered reference point cloud set to obtain a plurality of groups of corresponding point pairs between the cloud set of the reference points and the cloud set of the points to be registered, calculating direction vector included angles between the plurality of groups of corresponding point pairs, and performing precise registration on the cloud set of the points to be registered and the cloud set of the reference points through the relation between the direction vector included angles and a preset included angle threshold value, so that the registration method is prevented from falling into a local optimal solution, and the precise registration of the cloud set of the points to be registered and the cloud set of the reference points is improved.
The method comprises the steps of determining a point cloud set to be registered and a reference point cloud set with the maximum similarity by calculating the similarity between the point cloud set to be registered and the reference point cloud set, adjusting a first reference coordinate system and a second reference coordinate system corresponding to the point cloud set to the same target coordinate system, calculating the coincidence volume between a first minimum bounding box corresponding to the point cloud set to be registered and a second minimum bounding box corresponding to the reference point cloud set, and judging that the point cloud set to be registered and the reference point cloud set are initially registered when the coincidence volume meets the coincidence threshold by rotating coordinate axes of the target coordinate according to the relation between the coincidence volume and the coincidence threshold, so that the accuracy of initial registration of the point cloud set to be registered and the reference point cloud set is ensured.
In this embodiment, a point cloud registration method is provided, which may be used in electronic devices, such as scanning devices, medical detection devices, and the like, fig. 3 is a flowchart of the point cloud registration method according to an embodiment of the present invention, and as shown in fig. 3, the flowchart includes the following steps:
and S31, acquiring a point cloud set to be registered and a reference point cloud set. For a detailed description, refer to the related description of step S21 corresponding to the above embodiment, and the detailed description is omitted here.
And S32, calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method. For a detailed description, refer to the related description of step S22 corresponding to the above embodiment, and the detailed description is omitted here.
And S33, performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system. For a detailed description, refer to the related description of step S23 corresponding to the above embodiment, and the detailed description is omitted here.
And S34, based on the multi-dimensional binary search tree algorithm, searching the point closest to the point cloud set to be registered in the reference point cloud set after initial registration to obtain multiple groups of corresponding point pairs. For a detailed description, refer to the related description of step S24 corresponding to the above embodiment, and the detailed description is omitted here.
And S35, respectively calculating the direction vector included angles between the corresponding point pairs.
Specifically, the step S35 may include the following steps:
s351, calculating a first tangent plane formed by fitting the points in the point cloud set to be registered with the adjacent points of the point cloud set to be registered and a second tangent plane formed by fitting the points in the reference point cloud set with the adjacent points of the point cloud set to be registered.
For any point P in the point cloud set to be registered, a plurality of adjacent points corresponding to the point P exist, and a first tangent plane corresponding to the point P can be fit by the point P and the corresponding adjacent points. Similarly, for any point Q in the reference point cloud set, there are several adjacent points corresponding to the point Q, and the point Q and the several adjacent points corresponding to the point Q can be fit into a second tangent plane corresponding to the point Q. Wherein, the point in the point cloud set to be registered and the point in the reference point cloud set can form a corresponding point pair.
And S352, calculating a first normal vector corresponding to the point in the point cloud set to be registered based on the first tangent plane.
And on the first tangent plane, solving a normal vector of a point in the point cloud set to be registered by adopting a neighborhood covariance analysis method. Specifically, a covariance matrix can be obtained for any point P in the point cloud set to be registered and its adjacent N points:
Figure BDA0002983026280000151
wherein p is0The feature vector corresponding to the minimum feature value of the covariance matrix CV is the first normal vector of the point P.
And S353, calculating a second normal vector corresponding to the point in the reference point cloud set based on the second tangent plane.
And on the second tangent plane, solving the normal vector of the point in the reference point cloud set by adopting a neighborhood covariance analysis method. Specifically, for any point Q in the reference point cloud set and M points adjacent to the point Q, a covariance matrix can be obtained:
Figure BDA0002983026280000152
wherein q is0The feature vector corresponding to the minimum feature value of the covariance matrix CV is the second normal vector of the point Q.
It should be noted that the first normal vector and the second normal vector calculated by the above method may have two directions, and therefore, the first normal vector corresponding to all the points in the point cloud set to be registered needs to be adjusted so as to point to the same side of the cloud curved surface of the point to be registered, and the second normal vector corresponding to all the points in the reference point cloud set needs to be adjusted so as to point to the same side of the cloud curved surface of the reference point. Because the point cloud set has dense sampling points, the correct orientation of an initial point can be determined, then the normal vector orientation of a point close to the initial point is adjusted, and then the orientation adjustment of a first normal vector of the whole point cloud set to be registered and the orientation adjustment of a second normal vector of the reference point cloud set are completed through continuous neighborhood diffusion. In particular, since the point cloud is dense enough and the sampling plane is smooth everywhere, the normal vectors of two adjacent points will be close to parallel. E.g. niAnd njIs the normal vector of two adjacent points, if the directions of the normal vectors are consistent, then ni·nj1, if this inner product is negative, it indicates that the normal vector of a certain point needs to be inverted. Therefore, it is first necessary to set a normal vector orientation for a certain point in the point cloud, then go through all other points, if the normal vector of the current point is set to niThen n isjFor the next point to traverse, if n isi·njIf < 0, n isjAnd turning over, otherwise, keeping unchanged.
S354, calculating a direction vector included angle between the first normal vector and the second normal vector.
Based on a first normal vector corresponding to a point in the point cloud set to be registered and a second normal vector corresponding to a point in the reference point cloud set, an included angle cosine value between the first normal vector and the second normal vector can be calculated, and a direction vector included angle between the first normal vector and the second normal vector can be determined based on the included angle cosine value. For example, the first normal vector of a point correspondence in the point cloud set to be registered is
Figure BDA0002983026280000161
The second normal vector corresponding to a point in the reference point cloud set is
Figure BDA0002983026280000162
The cosine value of the included angle between the first normal vector and the second normal vector is:
Figure BDA0002983026280000163
and theta is a direction vector included angle between the first normal vector and the second normal vector.
And S36, performing fine registration on the point cloud set to be registered and the reference point cloud set based on the preset included angle threshold and the direction vector included angle.
Specifically, the step S36 may include the following steps:
s361, judging whether the included angle of the direction vector is smaller than a preset included angle threshold value.
The preset included angle threshold is an included angle value representing that the point in the point cloud set to be registered and the point in the reference point cloud set are correct corresponding point pairs. And comparing the direction vector included angle between the point in the point cloud set to be registered and the point in the reference point cloud set obtained by calculation with a preset included angle threshold value, and determining whether the direction vector included angle is smaller than the preset included angle threshold value. And executing step S362 when the direction vector included angle is smaller than the preset included angle threshold, otherwise executing step S363.
And S362, judging that the cloud set of the point to be registered and the cloud set of the reference point are accurately registered.
When the included angle of the direction vector is smaller than the preset included angle threshold value, the corresponding point pair formed by the point in the point cloud set to be registered and the point in the reference point cloud set is indicated to be a correct corresponding point pair, and the fact that the cloud set of the point to be registered and the cloud set of the reference point are accurately registered can be judged.
And S363, judging the corresponding point pair to be an error point pair, and rejecting the error point pair.
When the included angle of the direction vector is equal to or larger than a preset included angle threshold value, the corresponding point pair formed by the points in the point cloud set to be registered and the points in the reference point cloud set is represented as an error point pair, and the error point pair can be removed at the moment so as to ensure the accurate registration of the point cloud set to be registered and the reference point cloud set.
In the point cloud registration method provided by this embodiment, a first normal vector corresponding to a point in a point cloud set to be registered and a second normal vector corresponding to a point in a reference point cloud set are obtained by vectorizing the point in the point cloud set to be registered and the point in the reference point cloud set, so that a direction vector included angle between the first normal vector and the second normal vector is determined. And when the included angle of the direction vector is equal to or larger than a preset included angle threshold value, judging that the corresponding point pair is an error point pair, and rejecting the error point pair until the included angles of the direction vectors of all the corresponding point pairs are smaller than the preset included angle threshold value, judging that the cloud set of the point to be registered and the cloud set of the reference point are accurately registered, thereby improving the registration efficiency of the cloud set of the point to be registered and the cloud set of the reference point.
In this embodiment, a point cloud registration method is provided, which may be used in electronic devices, such as scanning devices, medical detection devices, and the like, fig. 4 is a flowchart of the point cloud registration method according to an embodiment of the present invention, and as shown in fig. 4, the flowchart includes the following steps:
and S41, acquiring a point cloud set to be registered and a reference point cloud set. For a detailed description, refer to the related description of step S31 corresponding to the above method embodiment, and are not repeated herein.
And S42, calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method. For a detailed description, refer to the related description of step S32 corresponding to the above method embodiment, and are not repeated herein.
And S43, performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system. For a detailed description, refer to the related description of step S33 corresponding to the above method embodiment, and are not repeated herein.
And S44, based on the multi-dimensional binary search tree algorithm, searching the point closest to the point cloud set to be registered in the reference point cloud set after initial registration to obtain multiple groups of corresponding point pairs. For a detailed description, refer to the related description of step S34 corresponding to the above method embodiment, and are not repeated herein.
And S45, respectively calculating the direction vector included angles between the corresponding point pairs. For a detailed description, refer to the related description of step S35 corresponding to the above method embodiment, and are not repeated herein.
And S46, performing fine registration on the point cloud set to be registered and the reference point cloud set based on the preset included angle threshold and the direction vector included angle. For a detailed description, refer to the related description of step S36 corresponding to the above method embodiment, and are not repeated herein.
And S47, adjusting the finely registered cloud set of the point to be registered and the cloud set of the reference point based on deep learning.
The Deep learning may be a Point cloud registration network (DCP), or may be other registration algorithms, and the Deep learning DCP is used to perform fine adjustment on the Point cloud set to be registered and the reference Point cloud set subjected to fine registration, so as to further improve the registration accuracy of the Point cloud set to be registered and the reference Point cloud set. As shown in fig. 6, the effect contrast maps are obtained by using different point cloud registration methods, and compared with the registration method using the ICP algorithm and the DCP algorithm, the point cloud registration method of the present embodiment significantly improves the registration accuracy between the target registration bone and the reference registration bone.
The point cloud registration network can be divided into: initial Features, Attention, Pointer Generation, and SVD Module. Specifically, the step S47 may include the following steps:
and S471, performing feature extraction on the precisely registered cloud set of the point to be registered and the reference point cloud set based on the point cloud feature extraction network to obtain a first point cloud feature corresponding to the cloud set of the point to be registered and a second point cloud feature corresponding to the reference point cloud set.
The point cloud feature extraction network can be an expansion gate convolutional neural network DGCNN and a feature extractor PointNet. Assuming that the point cloud set to be registered contains N points, the reference point cloud set contains M points, and the point cloud set to be registered is X ═ X1,x2,…,xi,…,xNY as reference point cloud set1,y2,…,yi,…,yMAnd M ═ N. The local features extracted before the last layer (L-th layer) of the point cloud feature extraction network are defined as
Figure BDA0002983026280000181
And
Figure BDA0002983026280000182
firstly, performing independent feature extraction on each point in a point cloud set to be registered and each point in a reference point cloud set by adopting PointNet, wherein the PointNet does not consider the relationship between the current point and the adjacent point, and then adopting a nearest node algorithm KNN of DGCNN to include neighborhood information of the current point. And respectively extracting a first point cloud characteristic corresponding to the point cloud set to be registered and a second point cloud characteristic corresponding to the reference point cloud set based on the point cloud characteristic extraction network.
S472, predicting a matching relation between the first point cloud characteristic and the second point cloud characteristic.
Based on the fact that the corresponding relation between the first point cloud feature and the second point cloud feature is similar to a sequence-to-sequence problem in Natural Language Processing (NLP), a transformer can be used as an attention function phi, and attention output is used as a residual item to correct an original feature FXAnd FYAnd obtaining the final characteristics:
Figure BDA0002983026280000183
and
Figure BDA0002983026280000184
and then calculating the similarity by using the dot product to predict the matching relationship between the first point cloud characteristic and the second point cloud characteristic as follows:
Figure BDA0002983026280000185
and S473, based on the matching relation, performing registration adjustment on the precisely registered cloud set of the point to be registered and the reference point cloud set by adopting a singular value decomposition method.
After the matching relation between the first point cloud characteristic and the second point cloud characteristic is obtained, the corresponding relation between the points in the cloud set of the points to be registered and the points in the reference point cloud set is determined. For point xiUsing the above matchingThe relationship can calculate the matching probability between the reference point cloud set Y and each point in the reference point cloud set Y, and the weighted summation of the matching probabilities can calculate an average point:
Figure BDA0002983026280000186
wherein, YTThe coordinates of M points in the cloud set Y of reference points are represented.
Figure BDA0002983026280000187
Is xiThe adjusted target point. According to the same method, the target point of each point in the cloud set X to be registered can be calculated. Finally, Singular Value Decomposition (SVD) method is used for calculating
Figure BDA0002983026280000188
Of (3) a rotation matrix RxyAnd a translation matrix txy。
The whole point cloud registration network is equivalent to inputting a point cloud set X to be registered and a reference point cloud set Y, and outputting a rotation matrix RxyAnd a translation matrix txy. With Rxy、txyAnd constructing a Loss function Loss with the ground truth value:
Figure BDA0002983026280000191
ideally, RxyIs an orthonormal matrix, so its transpose and inverse should be the same, and the following equation should hold:
Figure BDA0002983026280000192
and performing registration adjustment on the cloud set of the point to be registered and the cloud set of the reference point which are subjected to fine registration based on the loss function so as to quickly acquire correct corresponding point pairs.
The point cloud registration method provided in this embodiment further adjusts the to-be-registered point cloud set and the reference point cloud set through deep learning, specifically, performs feature extraction on the to-be-registered point cloud set and the reference point cloud set based on a point cloud registration network to obtain a first point cloud feature corresponding to the to-be-registered point cloud set and a second point cloud feature corresponding to the reference point cloud set, further performs matching prediction on the to-be-registered point cloud set and the reference point cloud set based on a correspondence between the first point cloud feature and the second point cloud feature, calculates a registration matrix between the to-be-registered point cloud set and the reference point cloud set through singular value decomposition, performs registration adjustment on the to-be-registered point cloud set and the reference point cloud set through the registration matrix to ensure that registration accuracy between the to-be-registered point cloud set and the reference point cloud set is optimal, and the registration precision of the cloud set of the point to be registered and the cloud set of the reference point is further improved.
In accordance with an embodiment of the present invention, there is provided an embodiment of a bone registration method based on point cloud registration, it is noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.
In this embodiment, a bone registration method based on point cloud registration is provided, which may be used in medical devices in the field of medical instruments, and the like, fig. 7 is a flowchart of the bone registration method based on point cloud registration according to an embodiment of the present invention, and as shown in fig. 7, the flowchart includes the following steps:
and S51, acquiring a first point cloud set corresponding to the target registration bone and a second point cloud set corresponding to the reference registration bone.
The reference registered bone is a fixed-position bone, and the second point cloud set corresponding to the reference registered bone is obtained by scanning the medical device from different angles. The target registration bone is a bone correspondingly connected with the reference registration bone, and the second point cloud set corresponding to the target registration bone can be obtained by scanning from different positions or different angles through the medical equipment. Specifically, the registration of the reference registration bone and the target registration bone may be registration between the acetabulum and the femoral head, registration between the femur and the tibia, registration between the femur and the pelvis, and the like, such as the registration between the acetabulum and the femoral head shown in fig. 8.
And S52, registering the first point cloud set and the second point cloud set by using the point cloud registration method in the embodiment.
For the detailed description of the point cloud registration method, the related description corresponding to the above method embodiment is referred to, and is not repeated herein. And performing initial registration and fine registration on the first point cloud set and the second point cloud set according to the point cloud registration method to complete accurate registration between the target registration bones and the reference registration bones, and further finely adjusting the accurately registered target registration bones and the reference registration bones by a deep learning (point cloud feature extraction network and the like) method, thereby improving the registration accuracy of the target registration bones and the reference registration bones.
In the bone registration method based on point cloud registration provided by this embodiment, the first point cloud set and the second point cloud set are initially registered based on the point cloud registration method, so that approximate coincidence between the first point cloud set and the reference point cloud set is rapidly achieved, and accurate registration between a target registration bone and a reference registration bone is completed. And searching points closest to the first point cloud set in the approximately coincident second point cloud set to obtain a plurality of groups of corresponding point pairs between the first point cloud set and the second point cloud set, and performing fine registration on the first point cloud set and the second point cloud set by calculating direction vector included angles between the plurality of groups of corresponding point pairs, so that the registration efficiency and the registration accuracy between the target registration bone and the reference registration bone are improved.
The present embodiment further provides a point cloud registration apparatus, which is used to implement the foregoing embodiments and preferred embodiments, and the description of the point cloud registration apparatus is omitted for brevity. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
The present embodiment provides a point cloud registration apparatus, as shown in fig. 9, including:
the first obtaining module 61 is configured to obtain a cloud set of points to be registered and a cloud set of reference points. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
The first calculating module 62 is configured to calculate, based on a principal component analysis method, a first reference coordinate system corresponding to the cloud set of the point to be registered and a second reference coordinate system corresponding to the cloud set of the reference point. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
And the primary registration module 63 is configured to perform primary registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
And the corresponding module 64 is configured to find a point closest to the cloud set of points to be registered in the reference point cloud set after the initial registration to obtain a plurality of groups of corresponding point pairs. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
And a second calculating module 65, configured to calculate direction vector included angles between the corresponding pairs of points. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
And the fine registration module 66 is configured to perform fine registration on the point cloud set to be registered and the reference point cloud set based on the direction vector included angle. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
The point cloud registration device provided in this embodiment calculates a first reference coordinate system corresponding to the cloud set of the point to be registered and a second reference coordinate system corresponding to the cloud set of the reference point based on a principal component analysis method, and realizes initial registration of the cloud set of the point to be registered and the cloud set of the reference point through the first reference coordinate system and the second reference coordinate system, so as to quickly realize approximate coincidence of the cloud set of the point to be registered and the cloud set of the reference point, thereby avoiding a registration method from falling into a local optimal solution. Based on the multi-dimensional binary search tree algorithm, the point closest to the cloud set of the points to be registered is searched in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs between the reference point cloud set and the cloud set of the points to be registered, and the search speed of the corresponding point pairs is improved. By calculating the direction vector included angles among the multiple groups of corresponding point pairs and based on the relationship between the direction vector included angles and the preset included angle threshold value, the point cloud set to be registered and the reference point cloud set are precisely registered, so that the precise registration of the point cloud set to be registered and the reference point cloud set is improved.
The point cloud registration apparatus in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC circuit, a processor and memory executing one or more software or fixed programs, and/or other devices that can provide the above-described functions.
Further functional descriptions of the modules are the same as those of the corresponding embodiments, and are not repeated herein.
In this embodiment, a bone registration apparatus based on point cloud registration is further provided, and the apparatus is used to implement the foregoing embodiments and preferred embodiments, which have already been described and are not described again. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
The present embodiment provides a bone registration apparatus based on point cloud registration, as shown in fig. 10, including:
a second obtaining module 71, configured to obtain a first point cloud set corresponding to the target registration bone and a second point cloud set corresponding to the reference registration bone. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
The registration module 72 is configured to be a registration module, and configured to register the first point cloud set and the second point cloud set by using the point cloud registration method described in the foregoing embodiment. For a detailed description, reference is made to the corresponding related description of the above method embodiments, which is not repeated herein.
The bone registration device based on point cloud registration provided by this embodiment performs initial registration on the first point cloud set and the second point cloud set based on the point cloud registration method, thereby quickly achieving approximate coincidence between the first point cloud set and the reference point cloud set, and completing accurate registration between a target registration bone and a reference registration bone. And searching points closest to the first point cloud set in the approximately coincident second point cloud set to obtain a plurality of groups of corresponding point pairs between the first point cloud set and the second point cloud set, and performing fine registration on the first point cloud set and the second point cloud set by calculating direction vector included angles between the plurality of groups of corresponding point pairs, so that the registration efficiency and the registration accuracy between the target registration bone and the reference registration bone are improved.
An embodiment of the present invention further provides an electronic device, which has the point cloud registration apparatus shown in fig. 9 and the bone registration apparatus based on point cloud registration shown in fig. 10.
Referring to fig. 11, fig. 11 is a schematic structural diagram of an electronic device according to an alternative embodiment of the present invention, and as shown in fig. 11, the electronic device may include: at least one processor 601, such as a CPU (Central Processing Unit), at least one communication interface 603, memory 604, and at least one communication bus 602. Wherein a communication bus 602 is used to enable the connection communication between these components. The communication interface 603 may include a Display (Display) and a Keyboard (Keyboard), and the optional communication interface 603 may also include a standard wired interface and a standard wireless interface. The Memory 604 may be a high-speed RAM (Random Access Memory) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The memory 604 may optionally be at least one storage device located remotely from the processor 601. Wherein the processor 601 may be in connection with the apparatus described in fig. 9 and 10, the memory 604 stores an application program, and the processor 601 calls the program code stored in the memory 604 for performing any of the above-mentioned method steps.
The communication bus 602 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 602 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 11, but this is not intended to represent only one bus or type of bus.
The memory 604 may include a volatile memory (RAM), such as a random-access memory (RAM); the memory may also include a non-volatile memory (english: non-volatile memory), such as a flash memory (english: flash memory), a hard disk (english: hard disk drive, abbreviated: HDD) or a solid-state drive (english: SSD); the memory 604 may also comprise a combination of the above types of memory.
The processor 601 may be a Central Processing Unit (CPU), a Network Processor (NP), or a combination of a CPU and an NP.
The processor 601 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.
Optionally, the memory 604 is also used for storing program instructions. The processor 601 may call program instructions to implement a point cloud registration method as shown in the embodiments of fig. 1 to 6 of the present application, and a bone registration method based on point cloud registration as shown in the embodiments of fig. 7 and 8 of the present application.
The embodiment of the invention also provides a non-transitory computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions can execute the point cloud registration method in any method embodiment and the processing method of the bone registration method based on point cloud registration. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD), a Solid State Drive (SSD), or the like; the storage medium may also comprise a combination of memories of the kind described above.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims (14)

1. A point cloud registration method is characterized by comprising the following steps:
acquiring a point cloud set to be registered and a reference point cloud set;
calculating a first reference coordinate system corresponding to the cloud set of points to be registered and a second reference coordinate system corresponding to the cloud set of reference points based on a principal component analysis method;
performing initial registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system;
based on a multi-dimensional binary search tree algorithm, searching points closest to the cloud set of points to be registered in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs;
respectively calculating the direction vector included angles between the multiple groups of corresponding point pairs;
and performing fine registration on the point cloud set to be registered and the reference point cloud set based on a preset included angle threshold and the direction vector included angle.
2. The method according to claim 1, wherein the calculating a first reference coordinate system corresponding to the cloud set of points to be registered and a second reference coordinate system corresponding to the cloud set of reference points based on a principal component analysis method comprises:
acquiring a first point cloud gravity center of the cloud set of points to be registered and a second point cloud gravity center of the cloud set of reference points;
respectively calculating a first main direction of the cloud set of points to be registered and a second main direction of the cloud set of reference points by adopting the main component analysis method; the first main direction is 3 first eigenvectors corresponding to the cloud set of points to be registered, and the second main direction is 3 second eigenvectors corresponding to the cloud set of reference points;
determining a first reference coordinate system corresponding to the cloud set of points to be registered based on the first point cloud gravity center and the first main direction;
and determining a second reference coordinate system corresponding to the reference point cloud set based on the second point cloud gravity center and the second main direction.
3. The method according to claim 1, wherein the initially registering the cloud set of points to be registered and the cloud set of reference points based on the first reference coordinate system and the second reference coordinate system comprises:
calculating the similarity between the point cloud set to be registered and the reference point cloud set, and determining a target point cloud set to be registered and a target reference point cloud set with the maximum similarity;
adjusting the first reference coordinate system and the second reference coordinate system corresponding to the target point cloud set to be registered and the target reference point cloud set to the same target coordinate system;
establishing a first minimum bounding box corresponding to the target cloud set of points to be registered and a second minimum bounding box corresponding to the target cloud set of reference points in the target coordinate system;
calculating a coincidence volume of the first minimum bounding box and the second minimum bounding box;
judging whether the coincidence volume meets a coincidence threshold value;
and when the coincidence volume meets the coincidence threshold, judging that the target cloud set to be registered and the target reference point cloud set are initially registered.
4. The method of claim 3, wherein the initially registering the cloud set of points to be registered and the cloud set of reference points based on the first reference coordinate system and the second reference coordinate system, further comprises:
and when the coincidence volume does not meet the coincidence threshold, rotating the coordinate axis of the target coordinate system by a preset angle until the coincidence volumes of the first minimum bounding box and the second minimum bounding box meet the coincidence threshold.
5. The method according to claim 1, wherein said calculating direction vector angles between the corresponding pairs of points comprises:
calculating a first tangent plane formed by fitting the points in the point cloud set to be registered with the adjacent points of the point cloud set to be registered and a second tangent plane formed by fitting the points in the reference point cloud set with the adjacent points of the point cloud set to be registered;
calculating a first normal vector corresponding to a point in the point cloud set to be registered based on the first tangent plane;
calculating a second normal vector corresponding to a point in the reference point cloud set based on the second tangent plane;
and calculating a direction vector included angle between the first normal vector and the second normal vector.
6. The method according to claim 5, wherein the pre-set angle threshold and the fine registration of the point cloud set to be registered and the reference point cloud set based on the direction vector angle comprise:
judging whether the included angle of the direction vector is smaller than a preset included angle threshold value or not;
and when the included angle of the direction vector is smaller than the preset included angle threshold value, judging that the point cloud set to be registered and the reference point cloud set are accurately registered.
7. The method of claim 6, wherein the fine registration of the point cloud set to be registered and the reference point cloud set based on the direction vector angle further comprises:
and when the included angle of the direction vector is equal to or larger than the preset included angle threshold value, judging that the corresponding point pair is an error point pair, and rejecting the error point pair.
8. The method of claim 1, further comprising:
and adjusting the precisely registered cloud set of points to be registered and the reference point cloud set based on deep learning.
9. The method according to claim 8, wherein the adjusting the finely registered cloud set of points to be registered and the reference point cloud set based on deep learning comprises:
performing feature extraction on the precisely registered cloud set of points to be registered and the reference point cloud set based on a point cloud feature extraction network to obtain first point cloud features corresponding to the cloud set of points to be registered and second point cloud features corresponding to the reference point cloud set;
predicting a matching relationship between the first point cloud feature and the second point cloud feature;
and based on the matching relation, carrying out registration adjustment on the precisely registered cloud set of the point to be registered and the reference point cloud set by adopting a singular value decomposition method.
10. A bone registration method based on point cloud registration is characterized by comprising the following steps:
acquiring a first point cloud set corresponding to a target registration bone and a second point cloud set corresponding to a reference registration bone;
registering the first point cloud set and the second point cloud set using the point cloud registration method of any of claims 1 to 9.
11. A point cloud registration apparatus, comprising:
the first acquisition module is used for acquiring a cloud set of points to be registered and a cloud set of reference points;
the first calculation module is used for calculating a first reference coordinate system corresponding to the cloud set of the points to be registered and a second reference coordinate system corresponding to the cloud set of the reference points based on a principal component analysis method;
the primary registration module is used for carrying out primary registration on the point cloud set to be registered and the reference point cloud set based on the first reference coordinate system and the second reference coordinate system;
the corresponding module is used for searching a point which is closest to the cloud set of points to be registered in the reference point cloud set after initial registration to obtain a plurality of groups of corresponding point pairs;
the second calculation module is used for calculating the direction vector included angles among the multiple groups of corresponding point pairs respectively;
and the fine registration module is used for performing fine registration on the point cloud set to be registered and the reference point cloud set based on the direction vector included angle.
12. A bone registration apparatus based on point cloud registration, comprising:
the second acquisition module is used for acquiring a first point cloud set corresponding to the target registration bone and a second point cloud set corresponding to the reference registration bone;
a registration module for registering the first point cloud set and the second point cloud set using the point cloud registration method of any one of claims 1 to 9.
13. An electronic device, comprising:
a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the point cloud registration method of any one of claims 1 to 9, or to perform the bone registration method based on point cloud registration of claim 10.
14. A computer-readable storage medium storing computer instructions for causing a computer to perform the point cloud registration method of any one of claims 1 to 9 or perform the bone registration method based on point cloud registration of claim 10.
CN202110292830.6A 2021-03-18 2021-03-18 Point cloud registration method, bone registration method, device, equipment and storage medium Pending CN113205547A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110292830.6A CN113205547A (en) 2021-03-18 2021-03-18 Point cloud registration method, bone registration method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110292830.6A CN113205547A (en) 2021-03-18 2021-03-18 Point cloud registration method, bone registration method, device, equipment and storage medium

Publications (1)

Publication Number Publication Date
CN113205547A true CN113205547A (en) 2021-08-03

Family

ID=77025449

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110292830.6A Pending CN113205547A (en) 2021-03-18 2021-03-18 Point cloud registration method, bone registration method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN113205547A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066989A (en) * 2021-11-16 2022-02-18 杭州云象商用机器有限公司 Robot positioning method, device, equipment and readable storage medium
CN114241018A (en) * 2021-12-09 2022-03-25 电子科技大学 Tooth point cloud registration method and system and readable storage medium
CN114305685A (en) * 2021-12-17 2022-04-12 杭州键嘉机器人有限公司 Hip bone registration method used in hip joint replacement surgery
CN114792326A (en) * 2022-03-30 2022-07-26 河海大学 Surgical navigation point cloud segmentation and registration method based on structured light
WO2024002360A1 (en) * 2022-06-30 2024-01-04 武汉联影智融医疗科技有限公司 Registration method and apparatus, and computer device and readable storage medium
CN117798933A (en) * 2024-01-30 2024-04-02 广州创之力智能科技有限公司 Dynamic following method and device for machine tail end based on intelligent robot

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109872352A (en) * 2018-12-29 2019-06-11 中国科学院遥感与数字地球研究所 Power-line patrolling LiDAR data autoegistration method based on shaft tower characteristic point
CN110415342A (en) * 2019-08-02 2019-11-05 深圳市唯特视科技有限公司 A kind of three-dimensional point cloud reconstructing device and method based on more merge sensors
EP3574473A1 (en) * 2017-01-27 2019-12-04 UCL Business PLC Apparatus, method, and system for alignment of 3d datasets
CN112017220A (en) * 2020-08-27 2020-12-01 南京工业大学 Point cloud accurate registration method based on robust constraint least square algorithm

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3574473A1 (en) * 2017-01-27 2019-12-04 UCL Business PLC Apparatus, method, and system for alignment of 3d datasets
CN109872352A (en) * 2018-12-29 2019-06-11 中国科学院遥感与数字地球研究所 Power-line patrolling LiDAR data autoegistration method based on shaft tower characteristic point
CN110415342A (en) * 2019-08-02 2019-11-05 深圳市唯特视科技有限公司 A kind of three-dimensional point cloud reconstructing device and method based on more merge sensors
CN112017220A (en) * 2020-08-27 2020-12-01 南京工业大学 Point cloud accurate registration method based on robust constraint least square algorithm

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
赵夫群: "《点云模型的优化配准方法研究》", 31 July 2018, 哈尔滨工业大学出版社 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066989A (en) * 2021-11-16 2022-02-18 杭州云象商用机器有限公司 Robot positioning method, device, equipment and readable storage medium
CN114066989B (en) * 2021-11-16 2022-05-27 杭州云象商用机器有限公司 Robot positioning method, device, equipment and readable storage medium
CN114241018A (en) * 2021-12-09 2022-03-25 电子科技大学 Tooth point cloud registration method and system and readable storage medium
CN114241018B (en) * 2021-12-09 2023-05-09 电子科技大学 Tooth point cloud registration method, system and readable storage medium
CN114305685A (en) * 2021-12-17 2022-04-12 杭州键嘉机器人有限公司 Hip bone registration method used in hip joint replacement surgery
CN114305685B (en) * 2021-12-17 2024-05-24 杭州键嘉医疗科技股份有限公司 Hip bone registration method for hip joint replacement operation
CN114792326A (en) * 2022-03-30 2022-07-26 河海大学 Surgical navigation point cloud segmentation and registration method based on structured light
WO2024002360A1 (en) * 2022-06-30 2024-01-04 武汉联影智融医疗科技有限公司 Registration method and apparatus, and computer device and readable storage medium
CN117798933A (en) * 2024-01-30 2024-04-02 广州创之力智能科技有限公司 Dynamic following method and device for machine tail end based on intelligent robot
CN117798933B (en) * 2024-01-30 2024-06-07 广州创之力智能科技有限公司 Dynamic following method and device for machine tail end based on intelligent robot

Similar Documents

Publication Publication Date Title
CN113205547A (en) Point cloud registration method, bone registration method, device, equipment and storage medium
CN107944020B (en) Face image searching method and device, computer device and storage medium
Caelli et al. An eigenspace projection clustering method for inexact graph matching
CN112328715B (en) Visual positioning method, training method of related model, related device and equipment
CN111897975A (en) Local training method for learning training facing knowledge graph representation
CN109359534B (en) Method and system for extracting geometric features of three-dimensional object
WO2018010101A1 (en) Method, apparatus and system for 3d face tracking
US9817845B2 (en) Three-dimensional image file searching method and three-dimensional image file searching system
US20210117648A1 (en) 3-dimensional model identification
Polsterer et al. Automatic galaxy classification via machine learning techniques: Parallelized rotation/flipping INvariant Kohonen maps (PINK)
CN116580257A (en) Feature fusion model training and sample retrieval method and device and computer equipment
CN116452648B (en) Point cloud registration method and system based on normal vector constraint correction
Yu et al. Toward faster and simpler matrix normalization via rank-1 update
CN111223128A (en) Target tracking method, device, equipment and storage medium
CN113095333A (en) Unsupervised feature point detection method and unsupervised feature point detection device
CN104751463A (en) Three-dimensional model optimal visual angle selection method based on sketch outline features
CN115982144A (en) Similar text duplicate removal method and device, storage medium and electronic device
CN114663770A (en) Hyperspectral image classification method and system based on integrated clustering waveband selection
CN111597367B (en) Three-dimensional model retrieval method based on view and hash algorithm
Streib et al. Using Ripley's K-function to improve graph-based clustering techniques
CN117216309A (en) Model training method, image retrieval method, device, equipment and medium
CN111860054A (en) Convolutional network training method and device
CN108876830A (en) Point cloud registration method, device, electronic equipment and readable storage medium storing program for executing
Divya Lakshmi et al. Helly hypergraph based matching framework using deterministic sampling techniques for spatially improved point feature based image matching
CN112200224A (en) Medical image feature processing method and device

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210803