System for detecting a marker point of a three-dimensional object and method thereof

Abstract

The application discloses a system and method for detecting a three-dimensional object mark point, and relates to the field of mark point detection, and comprises: a reading device for reading an image of a three-dimensional object and inputting the image into a statistical shape model; a registration module for registering the three-dimensional object with the statistical shape model; a parameter updating module for updating parameters of the statistical shape model; a mark point region identification module for searching for a near neighbor point of a mark point of the statistical shape model on the three-dimensional object as an identification result of a current mark point region; a judgment module for outputting a mark point region of the three-dimensional object when a parameter change value of the statistical shape model is less than a given threshold; and a mark point identification module for identifying a geometric mark point on the mark point region; the mark point device can quickly and stably identify a mark point region of a three-dimensional object of a patient, accurately and stably identify a position of a mark point in combination with geometric characteristics in the region, and identify a large number of anatomical and geometric mark points.

Description

Technical Field

[0001] This invention belongs to the field of marker detection, and more specifically, relates to a system and method for detecting markers on three-dimensional objects. Background Technology

[0002] Due to the complex structure of 3D objects, the geometric differences between different marker points are not significant, and the individual variations of 3D objects are substantial. Existing 3D feature point detection algorithms are not sensitive to specific marker points and are easily affected by individual differences, resulting in low accuracy and robustness. Therefore, current marker point detection mainly relies on manual annotation by restoration experts, a time-consuming and labor-intensive process that cannot provide sufficient accuracy. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a system and method for detecting landmarks on three-dimensional objects. This landmark device can quickly and stably identify the landmark regions of three-dimensional objects on a patient's body, accurately and stably identify the location of landmarks by combining the geometric characteristics of the region, without any type of user interaction, and identify a large number of anatomical and geometric landmarks through a strategy based on prior knowledge.

[0004] To achieve the above objectives, the present invention provides a system for detecting marker points of a three-dimensional object, comprising:

[0005] A reading device for reading images of three-dimensional objects and inputting those images into a statistical shape model;

[0006] The registration module registers the 3D object with the statistical shape template;

[0007] The parameter update module updates the parameters of the statistical shape model based on the differences between the statistical shape template and the 3D object.

[0008] The marker region recognition module searches and statistically analyzes the nearest neighbors of the markers in the shape template on the 3D object, and uses them as the recognition result of the current marker region.

[0009] The judgment module sets a given threshold. When the parameter change value of the statistical shape model is less than the given threshold, the judgment module outputs the marker point region that matches the 3D object.

[0010] The marker recognition module is able to identify geometric markers in the marker area.

[0011] Optionally, the statistical shape model includes multiple statistical shape templates.

[0012] Optionally, the registration module uses the ICP algorithm to register the three-dimensional object to obtain a minimized three-dimensional object.

[0013] Optionally, the parameter update module updates the parameters based on the difference between the minimized 3D object and the statistical shape template.

[0014] Optionally, the statistical shape model involves initially registering multiple 3D models at the same location to the same coordinate system, and then using a rigid iterative nearest-point algorithm to maximize the overlap of data between the multiple 3D models at the same location.

[0015] Optionally, the rigid iterative nearest point algorithm iteratively deforms the statistical shape template to approximate the shape of the 3D object, while simultaneously extracting the marker point regions of the 3D object that match the marker points of the statistical shape template.

[0016] The present invention also provides a method for detecting marker points of a three-dimensional object, utilizing the above-described system for detecting marker points of a three-dimensional object, the method comprising:

[0017] The image of the 3D object is read and input into the statistical shape model;

[0018] Register the 3D object with the statistical shape template;

[0019] The parameters of the statistical shape model are updated based on the differences between the statistical shape template and the 3D object;

[0020] The nearest neighbors of the marker points of the shape template on the 3D object are searched and statistically analyzed, and used as the identification result of the current marker point region.

[0021] Repeat the above steps until the parameter change value of the statistical shape model is less than the given threshold, then end the iterative optimization and output the marker point region of the matched 3D object.

[0022] Identify geometric markers within the marker area.

[0023] Optionally, a Gaussian process deformable model can be used when outputting the marker region of the matched 3D object.

[0024] Optionally, identifying geometric markers within the marked area includes both global and local geometric detection algorithms for statistical morphological models.

[0025] Optionally, geometric markers within the marked area can be identified using sphere fitting, surface fitting, curvature filtering, and normal filtering algorithms.

[0026] The present invention provides a system and method for detecting three-dimensional object landmarks. Its advantages are as follows: the system for detecting three-dimensional object landmarks can combine global landmark region recognition and local geometric feature detection, and the device identifies geometric landmarks based on the landmark region of a statistical morphological model. The device also has a local landmark detection algorithm based on prior geometric information.

[0027] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0028] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the invention.

[0029] Figure 1 A method for detecting marker points of a three-dimensional object according to an embodiment of the present invention is shown.

[0030] Figure 2 A flowchart illustrating a system for detecting landmarks of a three-dimensional object according to an embodiment of the present invention is shown for landmark recognition on a skeletal specimen. Detailed Implementation

[0031] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0032] This invention provides a system for detecting marker points on a three-dimensional object, comprising:

[0033] A reading device for reading images of three-dimensional objects and inputting those images into a statistical shape model;

[0034] The registration module registers the 3D object with the statistical shape template;

[0035] The parameter update module updates the parameters of the statistical shape model based on the differences between the statistical shape template and the 3D object.

[0036] The marker region recognition module searches and statistically analyzes the nearest neighbors of the markers in the shape template on the 3D object, and uses them as the recognition result of the current marker region.

[0037] The judgment module sets a given threshold. When the parameter change value of the statistical shape model is less than the given threshold, the judgment module outputs the marker point region that matches the 3D object.

[0038] The marker recognition module is able to identify geometric markers in the marker area.

[0039] Specifically, the system for detecting marker points of a 3D object inputs the target 3D object into a statistical shape model, finds the corresponding statistical shape template, registers the target 3D object with the statistical shape template, uses an alternating minimization scheme to obtain a minimized 3D object, and then updates the parameters of the statistical shape template based on the difference between the minimized 3D object and the statistical shape template to obtain an updated statistical shape template that approximates the target 3D object. In this way, the marker point region that matches the target 3D object can be obtained, thereby obtaining the geometric marker points within the marker point region.

[0040] Optionally, the statistical shape model includes multiple statistical shape templates.

[0041] Optionally, the registration module uses the ICP algorithm to register the 3D object to obtain a minimized 3D object.

[0042] Optionally, the parameter update module updates the parameters based on the difference between the minimized 3D object and the statistical shape template.

[0043] Optionally, the statistical shape model involves initially registering multiple 3D models at the same location to the same coordinate system, and then using a rigid iterative nearest-point algorithm to maximize the overlap of data between the multiple 3D models at the same location.

[0044] Specifically, the process of establishing the statistical shape model of a 3D model is as follows: First, identical 3D models are initially registered to a common coordinate system, and ICP (rigid iterative nearest point algorithm) is used to ensure maximum overlap of data among all sample 3D models. Then, principal component analysis is used to represent the points of each sample as vectors, which can represent the common components and individual differences of any personalized 3D model, resulting in a statistical shape model. This statistical shape model can be geometrically modified by changing the parameter vectors to adapt to the shape of individual 3D models. Finally, the geometric markers of key 3D models are marked in the average template of the statistical shape model. As the statistical shape model adaptively deforms, the geometric markers will change accordingly within the deformation field defined by the statistical shape model. In this way, the statistical shape model of the 3D model is established, and when a specific statistical shape template is needed, it can be extracted from the established model.

[0045] Optionally, the rigid iterative nearest point algorithm iteratively deforms the statistical shape template to approximate the shape of the 3D object, while simultaneously extracting the marker point regions of the 3D object that match the marker points of the statistical shape template.

[0046] Specifically, the statistical shape template is iteratively deformed to approximate the shape of the target 3D object. This involves taking the 3D object shape of the transformed statistical shape template as the target, matching it with the target 3D object, finding the corresponding marker point region, and finally confirming the geometric marker point through the marker point region.

[0047] The present invention also provides a method for detecting marker points of a three-dimensional object, utilizing the above-described system for detecting marker points of a three-dimensional object, the method comprising:

[0048] The image of the 3D object is read and input into the statistical shape model;

[0049] Register the 3D object with the statistical shape template;

[0050] The parameters of the statistical shape model are updated based on the differences between the statistical shape template and the 3D object;

[0051] The nearest neighbors of the marker points of the shape template on the 3D object are searched and statistically analyzed, and used as the identification result of the current marker point region.

[0052] Repeat the above steps until the parameter change value of the statistical shape model is less than the given threshold, then end the iterative optimization and output the marker point region of the matched 3D object.

[0053] Identify geometric markers within the marker area.

[0054] Optionally, a Gaussian process deformable model can be used when outputting the marker region of the matched 3D object.

[0055] Optionally, identifying geometric markers within the marked area includes both global and local geometric detection algorithms for statistical morphological models.

[0056] Specifically, under certain circumstances, statistical shape model registration can also be used as a standalone algorithm for marker detection; other feature detection algorithms, such as region weighting and corner detection, can be used within the marker region.

[0057] Optionally, geometric markers within the marked area can be identified using sphere fitting, surface fitting, curvature filtering, and normal filtering algorithms.

[0058] Specifically, the sphere fitting algorithm can detect the center point of a spherical anatomical structure, the surface fitting algorithm can detect the joint surface region and center point, the curvature filtering algorithm can detect the extreme points of concave and convex anatomy, and the normal filtering algorithm can detect the significant orientation region and center.

[0059] Example

[0060] This invention provides a system for detecting marker points on a three-dimensional object, comprising:

[0061] A reading device for reading images of three-dimensional objects and inputting those images into a statistical shape model;

[0062] The registration module registers the 3D object with the statistical shape template;

[0063] The parameter update module updates the parameters of the statistical shape model based on the differences between the statistical shape template and the 3D object.

[0064] The marker region recognition module searches and statistically analyzes the nearest neighbors of the markers in the shape template on the 3D object, and uses them as the recognition result of the current marker region.

[0065] The judgment module sets a given threshold. When the parameter change value of the statistical shape model is less than the given threshold, the judgment module outputs the marker point region that matches the 3D object.

[0066] The marker recognition module is able to identify geometric markers in the marker area.

[0067] In this embodiment, the statistical shape model includes multiple statistical shape templates.

[0068] In this embodiment, the registration module uses the ICP algorithm to register the three-dimensional object to obtain a minimized three-dimensional object.

[0069] In this embodiment, the parameter update module updates the parameters based on the difference between the minimized 3D object and the statistical shape template.

[0070] In this embodiment, the statistical shape model involves initially registering multiple 3D models at the same location to the same coordinate system, and then using a rigid iterative nearest-point algorithm to maximize the overlap of data between the multiple 3D models at the same location.

[0071] In this embodiment, the rigid iterative nearest point algorithm iteratively deforms the statistical shape template to approximate the shape of the three-dimensional object, while simultaneously extracting the marker point region of the three-dimensional object that matches the marker point of the statistical shape template.

[0072] like Figure 1 As shown, the present invention also provides a method for detecting three-dimensional object markers. Utilizing the aforementioned system for detecting three-dimensional object markers, the method includes:

[0073] The image of the 3D object is read and input into the statistical shape model;

[0074] Register the 3D object with the statistical shape template;

[0075] The parameters of the statistical shape model are updated based on the differences between the statistical shape template and the 3D object;

[0076] The nearest neighbors of the marker points of the shape template on the 3D object are searched and statistically analyzed, and used as the identification result of the current marker point region.

[0077] Repeat the above steps until the parameter change value of the statistical shape model is less than the given threshold, then end the iterative optimization and output the marker point region of the matched 3D object.

[0078] Identify geometric markers within the marker area.

[0079] In this embodiment, a Gaussian process deformable model is used when outputting the marker point region of the matched 3D object.

[0080] In this embodiment, identifying geometric markers within the marked area includes a global geometric detection algorithm and a local geometric detection algorithm based on the statistical morphological model.

[0081] In this embodiment, geometric markers within the marked area are identified using a sphere fitting algorithm, a surface fitting algorithm, a curvature filtering algorithm, and a normal filtering algorithm.

[0082] like Figure 2 As shown, the algorithm steps for detecting landmarks on a skeletal specimen in this system for detecting landmarks on a 3D object are as follows:

[0083] Step 1: Read the image of the target skeletal specimen;

[0084] Step 2: Register the target skeletal specimen with the statistical shape template;

[0085] Step 3: Calculate the difference between the target bone specimen and the statistical shape template by using the point correspondence between the target bone specimen and the statistical shape template;

[0086] Step 4: Calculate the iteration parameters of the statistical shape model and update the statistical shape model;

[0087] Step 5: Search the nearest neighbor region of the marker points in the statistical shape template to determine the landmark region of the target bone specimen;

[0088] Step 6: Repeat steps 2 to 5. When the parameter change value of the statistical shape model is less than the given threshold, end the iterative optimization and output the matched anatomical marker region.

[0089] Step 7: Identify geometric landmarks within the anatomical landmark area.

[0090] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A system for detecting marker points on a three-dimensional object, characterized in that, The system includes: A reading device for reading images of three-dimensional objects and inputting those images into a statistical shape model; The registration module registers the 3D object with the statistical shape template; The parameter update module updates the parameters of the statistical shape model based on the differences between the statistical shape template and the 3D object. The marker region recognition module searches and statistically analyzes the nearest neighbors of the markers in the shape template on the 3D object, and uses them as the recognition result of the current marker region. The judgment module sets a given threshold. When the parameter change value of the statistical shape model is less than the given threshold, the judgment module outputs the marker point region that matches the 3D object. The marker recognition module is capable of identifying geometric markers within a marker area; The registration module uses the ICP algorithm to register the three-dimensional objects to obtain the minimized three-dimensional object; The parameter update module updates parameters based on the difference between the minimized 3D object and the statistical shape template; The process of establishing a statistical shape model is as follows: First, identical 3D models are initially registered to a common coordinate system, and the rigid iterative nearest point algorithm is used to ensure that the data of all sample 3D models overlap to the maximum extent. Then, principal component analysis is used to represent the points of each sample as vectors, which can represent the common components and individual differences of any personalized 3D model, thus obtaining a statistical shape model. The rigid iterative nearest point algorithm approximates the shape of a 3D object by iteratively deforming a statistical shape template, while simultaneously extracting the landmark region of the 3D object that matches the landmark points of the statistical shape template.

2. The system for detecting marker points of a three-dimensional object according to claim 1, characterized in that, The statistical shape model includes multiple statistical shape templates.

3. A method for detecting marker points of a three-dimensional object, utilizing the system for detecting marker points of a three-dimensional object according to any one of claims 1-2, characterized in that, The method includes: The image of the 3D object is read and input into the statistical shape model; Register the 3D object with the statistical shape template; The parameters of the statistical shape model are updated based on the differences between the statistical shape template and the 3D object; The nearest neighbors of the marker points of the shape template on the 3D object are searched and statistically analyzed, and used as the identification result of the current marker point region. Repeat the above steps until the parameter change value of the statistical shape model is less than the given threshold, then end the iterative optimization and output the marker point region of the matched 3D object. Identify geometric markers within the marker area.

4. The method for detecting marker points of a three-dimensional object according to claim 3, characterized in that, A Gaussian process deformable model is used when outputting the marker point regions of the matched 3D objects.

5. The method for detecting marker points of a three-dimensional object according to claim 3, characterized in that, Identifying geometric landmarks within a marked area includes both global and local geometric detection algorithms for statistical morphological models.

6. The method for detecting marker points of a three-dimensional object according to claim 3, characterized in that, Geometric markers within the marked area are identified using sphere fitting, surface fitting, curvature filtering, and normal filtering algorithms.

Images