WO2024114758A1 - 三维牙齿模型分割方法和装置 - Google Patents

三维牙齿模型分割方法和装置 Download PDF

Info

Publication number
WO2024114758A1
WO2024114758A1 PCT/CN2023/135600 CN2023135600W WO2024114758A1 WO 2024114758 A1 WO2024114758 A1 WO 2024114758A1 CN 2023135600 W CN2023135600 W CN 2023135600W WO 2024114758 A1 WO2024114758 A1 WO 2024114758A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
tooth
seed point
point
edge
Prior art date
Application number
PCT/CN2023/135600
Other languages
English (en)
French (fr)
Inventor
周金海
王勇
周达超
Original Assignee
广州黑格智造信息科技有限公司
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
Priority claimed from CN202211543154.6A external-priority patent/CN116168185B/zh
Application filed by 广州黑格智造信息科技有限公司 filed Critical 广州黑格智造信息科技有限公司
Publication of WO2024114758A1 publication Critical patent/WO2024114758A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating 3D models or images for computer graphics
    • G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61C—DENTISTRY; APPARATUS OR METHODS FOR ORAL OR DENTAL HYGIENE
    • A61C19/00—Dental auxiliary appliances
    • A61C19/04—Measuring instruments specially adapted for dentistry
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61C—DENTISTRY; APPARATUS OR METHODS FOR ORAL OR DENTAL HYGIENE
    • A61C7/00—Orthodontics, i.e. obtaining or maintaining the desired position of teeth, e.g. by straightening, evening, regulating, separating, or by correcting malocclusions
    • A61C7/002—Orthodontic computer assisted systems
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
    • B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
    • B29C64/00—Additive manufacturing, i.e. manufacturing of three-dimensional [3D] objects by additive deposition, additive agglomeration or additive layering, e.g. by 3D printing, stereolithography or selective laser sintering
    • B29C64/30—Auxiliary operations or equipment
    • B29C64/386—Data acquisition or data processing for additive manufacturing
    • B29C64/393—Data acquisition or data processing for additive manufacturing for controlling or regulating additive manufacturing processes
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B33—ADDITIVE MANUFACTURING TECHNOLOGY
    • B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
    • B33Y50/00—Data acquisition or data processing for additive manufacturing
    • B33Y50/02—Data acquisition or data processing for additive manufacturing for controlling or regulating additive manufacturing processes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/64—Three-dimensional objects
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61C—DENTISTRY; APPARATUS OR METHODS FOR ORAL OR DENTAL HYGIENE
    • A61C7/00—Orthodontics, i.e. obtaining or maintaining the desired position of teeth, e.g. by straightening, evening, regulating, separating, or by correcting malocclusions
    • A61C7/002—Orthodontic computer assisted systems
    • A61C2007/004—Automatic construction of a set of axes for a tooth or a plurality of teeth
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • the present disclosure relates to the technical field of dental orthodontics, and in particular to a three-dimensional tooth model segmentation method and device.
  • Tooth target position simulation refers to inputting the patient's scan data, and through simple operations and automatic operation of the system, the "corrected” effect can be obtained and displayed, and the continuous “animation” of the change from the original state to the target state can be demonstrated, so that the doctor and the patient can communicate better and promote the transaction.
  • the output result of the target position simulation is also one of the contents of the subsequent links to derive production, and tooth separation is an important link.
  • the present application provides a three-dimensional tooth model segmentation method and device to solve the above-mentioned technical problem of low accuracy in segmenting teeth in the three-dimensional tooth model.
  • the present application provides a method, comprising: acquiring a two-dimensional projection image of a three-dimensional tooth model, and identifying the above-mentioned two-dimensional projection image to obtain multiple tooth regions, and the tooth regions correspond one-to-one to the teeth; in the above-mentioned three-dimensional tooth model, confirming the original seed points corresponding to the above-mentioned multiple tooth regions; expanding the above-mentioned original seed points within a preset range to obtain target seed points of the teeth in the above-mentioned three-dimensional tooth model; and segmenting the above-mentioned three-dimensional tooth model based on the target seed points of each of the above-mentioned teeth to obtain segmented teeth.
  • the present application also provides a method for manufacturing a dental instrument, comprising:
  • the segmented teeth are obtained by any of the above methods.
  • the tooth is segmented and printed based on the target to obtain a formed tooth model; wherein the formed tooth model is used to obtain a dental appliance.
  • the present application also provides a three-dimensional tooth model segmentation device, comprising:
  • a memory and a processor wherein the memory stores a computer program, and when the computer program is executed by the processor, any one of the above methods is executed.
  • a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any one of the above methods is executed.
  • the present application determines the original seed points on the 3D tooth model by identifying the 2D projection image of the 3D tooth model, and expands the original seed points to obtain the target seed points, thereby marking the range of the teeth on the 3D tooth model through the target seed points, and further segmenting the 3D tooth model according to the target seed points to obtain the segmented teeth, thereby achieving the effect of accurately segmenting the teeth of the 3D tooth model, solving the problem of low accuracy in segmenting the teeth of the 3D tooth model in the prior art, and compared with the traditional tooth segmentation method, the present application can also avoid the segmentation of missing teeth. Set to split.
  • FIG1 is a schematic diagram of a hardware environment of an optional 3D tooth model segmentation method provided according to an embodiment of the present application
  • FIG2 is a flow chart of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG3 is a two-dimensional projection image of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG4 is a preliminary segmentation region diagram of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG5 is a preliminary segmentation region diagram of another three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG6 is a partial diagram of a three-dimensional tooth model corresponding to a preliminary segmented area provided according to an embodiment of the present application.
  • FIG7 is a schematic diagram of a flow network composed of seed points of a three-dimensional tooth model provided according to an embodiment of the present disclosure
  • FIG8 is a schematic diagram of a preliminary segmented area segmented by a minimum flow algorithm according to an embodiment of the present disclosure
  • FIG9 is a triangular facet diagram of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG10 is a smoothed boundary diagram of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG11 is a tooth sorting diagram of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG12 is a flow chart of a three-dimensional tooth model segmentation method provided according to an embodiment of the present application.
  • FIG13 is a schematic diagram of wave crests and wave troughs of a three-dimensional tooth model provided according to an embodiment of the present application.
  • FIG14 is a schematic diagram of a tooth separation of a three-dimensional tooth model provided according to an embodiment of the present application.
  • FIG15 is a flow chart of a three-dimensional tooth model provided according to an embodiment of the present application.
  • FIG16 is a schematic diagram of a combination of wave crest points of a three-dimensional tooth model provided according to an embodiment of the present application.
  • FIG17 is a segmentation curve diagram of a three-dimensional tooth model provided according to an embodiment of the present application.
  • FIG. 18 is a block diagram of a three-dimensional tooth model segmentation device provided according to an embodiment of the present application.
  • Flow Network refers to a special type of weighted directed complex network.
  • directed edges represent the direction of the flow of energy, matter, currency, information, attention, etc.
  • the weight of the edge represents the flow.
  • the flow network of this application is a graph composed of edges and points, in which there is a point with an in-degree of 0, called a source point, and a point with an out-degree of 0, called a sink point.
  • Source A point that emits a flow, with an infinite flow, usually represented by s.
  • the in-degree of the source point is 0, and the out-degree is greater than 0.
  • Terminal A point that receives a flow, usually represented by t.
  • the out-degree of a terminal is 0, and the in-degree is greater than 0.
  • Network flow The collection of all edge flows in a flow network.
  • Minimum Cut Algorithm It is one of the classic algorithms for image segmentation, also called Graph Cut, which is used to divide a graph into two parts so that the sum of the edge weights between the two parts is minimized.
  • Maximum flow minimum cut In a directed graph, the maximum flow that can reach the sink from the source is equal to the minimum volume sum of the set of edges that can cause network flow interruption if they are subtracted from the way.
  • the value of the maximum flow is equal to the capacity of the minimum cut.
  • an embodiment of a three-dimensional tooth model segmentation method is provided.
  • the above method can be applied to a hardware environment composed of a terminal 101 and a server 103 as shown in FIG1 .
  • the server 103 is connected to the terminal 101 via a network, and can be used to provide services for the terminal or a client installed on the terminal.
  • a database 105 can be set on the server or independently of the server to provide data storage services for the server 103.
  • the above network includes but is not limited to: a wide area network, a metropolitan area network or a local area network.
  • the terminal 101 can be a device for obtaining a three-dimensional tooth model, such as a mouth scan model scanner, or a terminal for receiving a three-dimensional tooth model, such as a mobile phone, a computer, etc.
  • the 3D tooth model segmentation method in the embodiment of the present application can be executed by the server 103, or can be executed by the server 103 and the terminal 101 together. As shown in FIG2 , the 3D tooth model segmentation method can include the following steps:
  • Step S202 obtaining a two-dimensional projection image of the three-dimensional tooth model, and identifying the two-dimensional projection image to obtain a plurality of tooth regions, where the tooth regions correspond to the teeth one by one;
  • Step S204 confirming original seed points corresponding to multiple tooth regions in the three-dimensional tooth model
  • Step S206 expanding the original seed point within a preset range to obtain the target seed point of the tooth in the three-dimensional tooth model
  • Step S208 segmenting the three-dimensional tooth model based on the target seed point of each tooth to obtain segmented teeth.
  • the purpose of the above three-dimensional tooth model segmentation method is to accurately segment each tooth on the three-dimensional tooth model into individual teeth.
  • the tooth region can correspond to each tooth one by one.
  • one tooth region can also correspond to multiple teeth.
  • the above-mentioned 3D tooth model can be an oral scan model, which refers to a 3D model including the user's teeth generated by scanning the inside of the user's mouth.
  • the data of the 3D tooth model is stored in a computer or server, and the shape or style of the 3D tooth model can be displayed on a display screen for easy viewing by the doctor.
  • the two-dimensional projection image can be an image obtained by projecting the three-dimensional tooth model onto a plane.
  • the two-dimensional projection image includes a tooth area and a non-tooth area.
  • the recognition method can use machine vision or artificial intelligence to recognize the two-dimensional projection image to obtain the tooth area.
  • the surface of the 3D tooth model is composed of polygonal patches, such as triangular patches.
  • a triangular patch contains 3 vertices, and all the vertices of the triangular patches can be regarded as seed points.
  • the above-mentioned original seed points are the seed points corresponding to the tooth area in the 2D projection image among all the seed points of the 3D tooth model. In other words, the tooth area can be preliminarily determined on the 3D tooth model through the original seed points.
  • the original seed point can be expanded to obtain the target seed point by expanding the original seed point.
  • the area covered by the target seed point can be regarded as the teeth on the 3D tooth model, and the 3D tooth model can be accurately divided by the target seed point.
  • the tooth area After the tooth area is identified, it can be understood that the position and area of the teeth have been determined on the two-dimensional projection image. Then, it is matched to the three-dimensional tooth model to determine the original seed point on the three-dimensional tooth model. Since the seed point is the vertex of the triangular facet on the surface of the three-dimensional tooth model, the original seed point can be understood as the area covering the teeth on the three-dimensional tooth model. In order to ensure accuracy, the original seed point is further expanded to obtain the target seed point, and the three-dimensional tooth model is segmented according to the target seed point to obtain the segmented teeth. At this point, the teeth in the three-dimensional tooth model can be divided into single teeth.
  • the above method determines the original seed points on the three-dimensional tooth model by identifying the two-dimensional projection image of the three-dimensional tooth model, and expands the original seed points to obtain the target seed points, thereby marking the range of the teeth on the three-dimensional tooth model through the target seed points.
  • the three-dimensional tooth model can be further segmented according to the target seed points to obtain segmented teeth, thereby achieving the effect of accurately segmenting the three-dimensional tooth model.
  • the original seed points are expanded within a preset range to obtain the target seed points of the teeth in the three-dimensional tooth model, including: determining the tooth edge based on the original seed points, the tooth edge is the edge of the surface of the three-dimensional tooth model covering all the original seed points of the tooth area; in the three-dimensional tooth model, confirming the non-target seed points corresponding to the areas outside the multiple tooth areas, and determining the gingival edge based on the non-target seed points, the gingival edge is the edge of the surface of the three-dimensional tooth model covering all the non-target seed points; on the surface of the three-dimensional tooth model, any point on the side of the tooth edge away from the gingival edge is determined as a source point, and any point on the side of the gingival edge away from the tooth edge is determined as a sink point; constructing multiple flow networks based on each source point and sink point, so that the seed points between the source point and the sink point are all nodes of the flow network, any two adjacent nodes are connected, and the
  • the tooth edge is determined according to the original seed point.
  • One tooth area corresponds to one tooth edge.
  • the gingival edge is determined according to the non-target seed point. There is only one gingival edge, and the part between the tooth edge and the gingival edge is unclear whether it is a tooth or gingiva.
  • the minimum cut algorithm is used to segment the flow network composed of seed points.
  • the nodes connected to the source nodes in the segmented flow network are the seed points of the teeth, and the nodes connected to the sink nodes in the segmented flow network are the seed points of the gingiva.
  • the teeth and gingiva can be accurately segmented.
  • non-target seed points corresponding to areas outside multiple tooth regions are confirmed, and the gingival margin is determined based on the non-target seed points, including: preliminarily dividing the three-dimensional tooth model according to multiple tooth regions to obtain multiple preliminary segmentation regions, and the preliminary segmentation regions cover the tooth regions one-to-one; in each preliminary segmentation region, groups of non-target seed points corresponding to areas outside each tooth region are confirmed; and corresponding gingival margins are determined based on each group of non-target seed points, the gingival margin being the edge of the surface of the three-dimensional tooth model covering a group of non-target seed points, and the gingival margin corresponds to the tooth one-to-one.
  • the preliminary segmentation area includes both the tooth area and the non-tooth area.
  • the gingival edge is the edge determined by the non-target seed point of the non-tooth area.
  • the above-mentioned gingival edge also corresponds to the tooth area one by one. This can reduce the size of the part between the tooth edge and the gingival edge, thereby achieving a better segmentation effect.
  • a minimum cut algorithm is used to segment the flow network, and the nodes connected to the source point in the segmented flow network are determined as target seed points, including: obtaining geometric features of the triangles corresponding to each two adjacent nodes, the geometric features being one or more of the length of the target edge, the angle between the normal corresponding to the two adjacent nodes and the target edge, the average curvature between the two adjacent nodes, and the distance between the target edge and the third vertex of the triangle, the target edge being the edge where two adjacent nodes in the triangle are located; calculating the weighted average of each geometric feature to obtain the capacity of each target edge; using a minimum cut algorithm to segment the flow network according to the capacity of each target edge, and determining the nodes connected to the source point in the segmented flow network as target seed points.
  • the geometric features corresponding to the edges are weighted averaged, and the weighted average is obtained as the capacity of each edge, so that the segmentation of teeth and gums can be achieved through the minimum cut algorithm.
  • the original seed point after the original seed point is obtained, the original seed point can be expanded to obtain the target seed point.
  • the process of expanding the original seed point can be divided into one or more stages.
  • the original seed point can be expanded according to a preset curvature threshold to obtain a target seed point.
  • the preset curvature threshold can be understood as a constraint used when expanding the original seed point to prevent the original seed point from expanding beyond the limit.
  • the preset curvature threshold may include one or more curvature values. If the preset curvature threshold includes a curvature value, the original seed point can be expanded according to the curvature value to obtain the target seed point.
  • the first curvature value can be used to expand the original seed point, and then the second curvature value can be used to expand the expansion result of the first curvature value, and the third curvature value can be used to expand the expansion result of the second curvature value, until all curvature values are used once.
  • the original seed point is expanded according to a preset curvature threshold to obtain the target seed point, which includes: The original seed point is expanded according to the initial curvature threshold to obtain a first seed point; the first seed point is expanded according to the target curvature threshold to obtain a target seed point, wherein the target curvature threshold is obtained according to the initial curvature threshold.
  • the original seed point is first expanded by the initial curvature threshold to obtain the first seed point, and then the first seed point is expanded by the target curvature threshold to obtain the target seed point.
  • the initial curvature threshold and the target curvature threshold can be the same or different, and the target seed point is obtained through two expansions.
  • the original seed point is expanded according to the initial curvature threshold to obtain the first seed point, which includes: taking the seed point adjacent to the original seed point as the current seed point; when the curvature of the current seed point is less than or equal to the initial curvature threshold, taking the current seed point and the original seed point as the first seed point.
  • the curvature of the seed point adjacent to the original seed point can be obtained.
  • Curvature is the rotation rate of the tangent direction angle to the arc length at a point on the curve, which is defined by differentiation and indicates the degree to which the curve deviates from a straight line. A numerical value indicating the degree of curvature of the curve at a certain point.
  • Each seed point corresponds to a curvature.
  • adjacent here and in the following text can be understood as two vertices that form the same side of the same triangle with the seed point.
  • adjacent to the original seed point means two vertices that form the same side of the same triangle with the original seed point.
  • the first seed point is expanded according to the target curvature threshold to obtain the target seed point, which includes: taking the seed point adjacent to the first seed point as the current seed point; and when the curvature of the current seed point is less than or equal to the target curvature threshold, taking the first seed point and the current seed point as the target seed points.
  • the first seed point can be expanded using the target curvature threshold to obtain the target seed point.
  • the curvature of the seed point adjacent to the first seed point is determined, and then the curvature is compared with the target curvature threshold. By comparing the size relationship, it is determined whether the seed point adjacent to the first seed point is the target seed point.
  • the original seed point is expanded using the initial curvature threshold and the target curvature threshold to obtain the target seed point.
  • the method before expanding the first seed point according to the target curvature threshold to obtain the target seed point, the method further includes: taking the sum of the initial curvature threshold and a preset value as the target curvature threshold, wherein the preset value is a positive number.
  • the initial curvature threshold and the target curvature threshold may be empirical values, or the initial curvature threshold may be an empirical value, and the target curvature threshold is determined according to the initial curvature threshold.
  • the sum of the initial curvature threshold and the preset value is used as the target curvature threshold, that is, the target curvature threshold is obtained according to the initial curvature threshold, and the target curvature threshold is greater than the initial curvature threshold.
  • the preset value is a pre-set value and can be modified according to different three-dimensional tooth models.
  • the above method also includes: when the corresponding point of the first seed point on the two-dimensional projection image does not fall into the corresponding tooth area, adjusting the first seed point to a non-first seed point; or when the corresponding point of the target seed point on the two-dimensional projection image does not fall into the corresponding tooth area, adjusting the target seed point to a non-target seed point.
  • the original seed point when the original seed point is expanded according to the initial curvature threshold or the first seed point is expanded according to the target curvature threshold, it is also necessary to check whether the expanded seed point meets the requirements, that is, whether the expansion exceeds the range, whether the seed point in the non-tooth area is used as the first seed point or the seed point in the non-tooth area is used as the target seed point.
  • the original seed point is expanded according to the initial curvature threshold, if the curvature of the seed point adjacent to the original seed point is less than or equal to the initial curvature threshold, it is also necessary to determine the area where the corresponding point of the seed point on the two-dimensional projection image is located.
  • the seed point is to be regarded as a non-first seed point.
  • the first seed point is expanded according to the target curvature threshold, if the curvature of the seed point adjacent to the first seed point is less than or equal to the target curvature threshold, it is also necessary to determine the area where the corresponding point of the seed point on the two-dimensional projection image is located. If it is not located in the tooth area, it means that the seed point has left the tooth area of the three-dimensional tooth model. The tooth area, therefore, this seed point is used as a non-target seed point.
  • the above method also includes: obtaining the target area by expanding the area obtained by expanding multiple tooth areas; in the target area, expanding the target seed points according to the initial curvature threshold and height.
  • the teeth of the three-dimensional tooth model can be segmented according to the target seed points.
  • the target seed points can be expanded again before segmenting the teeth.
  • the second stage expansion can also be performed.
  • the tooth area on the two-dimensional projection image can be adjusted first, and the tooth area can be expanded to obtain the target area. Then, with the target area as the limit, the target seed point is expanded in the second stage using the initial curvature threshold and height.
  • the purpose of expanding the tooth area to the target area is to ensure that all seed points on the teeth on the three-dimensional tooth model will be marked as target seed points to avoid omission.
  • the seed point adjacent to the target seed point can be used as the current seed point; when the height of the current seed point is greater than the preset standard height and the curvature of the current seed point is less than or equal to the target curvature threshold, the current seed point is used as the target seed point.
  • the height may be the value of the seed point in the preset direction of the three-dimensional tooth model.
  • the preset direction of the three-dimensional tooth model may be taken as the Z axis
  • the coordinate value of the seed point on the Z axis may be taken as the height of the seed point.
  • the preset direction may be any direction. After the three-dimensional tooth model is obtained, the orientation of the three-dimensional tooth model may be adjusted to the preset direction, so that the directions of all the obtained three-dimensional tooth models may be unified.
  • the seed point adjacent to the target seed point can be used as the current seed point. If the curvature and height values of the current seed point meet the requirements of the initial curvature threshold and height, the current seed point can be used as the target seed point, thereby completing the expansion of the target seed point.
  • taking the current seed point as a target seed point includes: when the height of the current seed point is greater than the standard height, and the curvature of the current seed point is less than or equal to the target curvature threshold, and the corresponding point of the current seed point on the two-dimensional projection image is within the target area, taking the current seed point as the target seed point, wherein the target area is an area obtained by enlarging multiple tooth areas; when the corresponding point of the current seed point on the two-dimensional projection image is outside the target area, taking the current seed point as a non-target seed point.
  • the target area is the area obtained after the tooth area is expanded.
  • the purpose of expanding the tooth area is to include a part of the non-tooth area near the tooth area in the target area. In this way, when the target seed point is expanded using the initial curvature threshold and height, the target seed point can be allowed to expand to the non-tooth area near the tooth area. Doing so can make the expanded target seed point cover all tooth areas.
  • the method further includes: taking a seed point adjacent to the expanded target seed point as a current seed point; and taking the current seed point as a target seed point as well.
  • the expanded target seed point can be expanded again, and the seed points adjacent to the target seed point are also used as target seed points.
  • the purpose of this expansion is also to make the expanded target seed point cover all tooth areas, so that when the tooth is segmented according to the expanded target seed point, the tooth is complete.
  • the method before expanding the original seed point within a preset range, the method further includes: expanding multiple tooth areas to obtain a target area; marking points other than the target area on the three-dimensional tooth model as third seed points; and expanding the third seed point according to a curvature threshold.
  • the tooth area is first expanded to obtain the target area, the purpose of which is to make the target area include all tooth areas and also include non-tooth areas near the tooth areas. Therefore, when the points of the target area other than the points on the three-dimensional tooth model are marked as third seed points, the third seed points are not points on the teeth.
  • the points of the non-tooth part (gum part or the space between teeth) close to the tooth part can be marked as third seed points, so that the boundary between the part outside the tooth and the tooth part can be brought closer to the tooth part, reducing the scope of the target area, so that although the target area includes the non-tooth area, it includes less non-tooth area.
  • the dividing line between the tooth part and the part outside the tooth on the three-dimensional tooth model becomes "thinner".
  • expanding the third seed point according to the curvature threshold includes: using a seed point adjacent to the third seed point as a current seed point; and using the current seed point as the third seed point when the curvature of the current seed point is greater than the curvature threshold.
  • the third seed point when the third seed point is expanded, can be expanded according to the curvature and the curvature threshold.
  • the curvature of the third seed point can be calculated by calculating the rotation rate of the tangent direction angle of the point to the arc length through differentiation.
  • the curvature threshold can be a preset threshold, and the curvature threshold can be different for different three-dimensional tooth models.
  • the method further includes: when a point corresponding to the current seed point in the two-dimensional projection image is within the target area, taking the current seed point as a non-third seed point.
  • the third seed point After the third seed point is expanded, it is necessary to check whether the third seed point is expanded into the target area. Because the target area includes the tooth area, if the third seed point is expanded into the target area, it may be expanded into the tooth area. Therefore, the third seed point in the target area should be returned as a non-third seed point.
  • the above method after expanding the third seed point according to the curvature threshold, also includes: determining a first area composed of the third seed points; determining a sub-area from the first area; and when the sub-area is surrounded by the target seed point, determining the seed point in the sub-area as the target seed point.
  • the third seed point forms the first region.
  • the first region can be divided into multiple sub-regions.
  • Each of the multiple sub-regions can be understood as a region composed of multiple third seed points. If a sub-region is surrounded by the target seed point, it means that the sub-region is located within the tooth part on the three-dimensional tooth model, but the sub-region may not be a tooth, that is, a hole part on the tooth. Therefore, the seed point of the sub-region is used as the target seed point and divided into the tooth part.
  • obtaining a two-dimensional projection image of a three-dimensional tooth model includes: adjusting the orientation of the three-dimensional tooth model from an initial orientation to a target orientation; and projecting the three-dimensional tooth model with the target orientation onto a target surface to obtain a two-dimensional projection image.
  • the orientation of the three-dimensional tooth model when projecting the three-dimensional tooth model into a two-dimensional projection image, the orientation of the three-dimensional tooth model can be adjusted first, and the orientation can be adjusted to the target orientation.
  • the target orientation can be a pre-set orientation.
  • the purpose of adjusting the orientation of the three-dimensional tooth model is to make the tooth area on the two-dimensional projection image more complete and less blocked when the three-dimensional tooth model is projected into a two-dimensional projection image.
  • the horizontal plane can be used as the X-axis and Y-axis of the three-dimensional coordinate, and the upward direction of the horizontal plane can be used as the target orientation.
  • the tooth direction of the three-dimensional tooth model After acquiring the three-dimensional tooth model, the tooth direction of the three-dimensional tooth model can be oriented to the target orientation.
  • the three-dimensional tooth model may be projected onto a target plane, which may be a horizontal plane.
  • recognizing the two-dimensional projection image to obtain multiple tooth regions includes: inputting the two-dimensional projection image into a recognition model, and using the recognition model to mark multiple tooth regions on the two-dimensional projection image.
  • the two-dimensional projection image can be recognized by the recognition model, so that the tooth area can be recognized.
  • the recognition model extracts features and recognizes, and outputs multiple tooth areas.
  • confirming the original seed points corresponding to multiple tooth regions includes: taking the vertex of each triangular face in the three-dimensional tooth model as the current vertex; and when the corresponding point of the current vertex in the two-dimensional projection image is located in multiple tooth regions, taking the current vertex as an original seed point.
  • the surface of the three-dimensional tooth model is covered by polygonal patches, which can be triangular patches.
  • polygonal patches which can be triangular patches.
  • each triangular patch has three vertices. Two adjacent triangular patches share an edge.
  • the vertex can be used as the original seed point.
  • the method further includes: sorting the segmented teeth.
  • sorting the segmented teeth includes: taking the average of the midpoints of all teeth as the starting point and the midpoint of each tooth as the end point to form a vector for each tooth; taking the straight line where the two teeth with the largest distance between their midpoints are located among all teeth as the target straight line; and sorting all teeth according to the size of the angle between the vector and the target straight line.
  • the teeth in this embodiment After the teeth in this embodiment are segmented, they can be sorted.
  • the line connecting the midpoint of the tooth and the midpoint of each tooth is used as a vector, and the angle between the target straight line formed by the two teeth with the largest distance can be used to sort the teeth.
  • the teeth can be sorted starting from the first tooth on one side and ending at the last tooth on the other side.
  • the method further includes: smoothing the edges of the teeth to obtain smoothed edges.
  • smoothing the edge of a tooth to obtain a smoothed edge includes: sorting the vertices on the edge of the tooth; using the second vertex among the sorted vertices as the current vertex, and performing the following operations on the current vertex until the current vertex does not include the rear vertex: using the center point of the front vertex and the rear vertex of the current vertex as a smoothing point; each time a smooth point is obtained, using the rear vertex of the current vertex as a new current vertex, and using the obtained smooth point as the front vertex of the new current vertex; connecting the first vertex on the edge of the tooth with the obtained smooth point in order to obtain the smoothed edge of the tooth.
  • the purpose of smoothing the edges of the teeth is to make the edges of the dividing line between the teeth and the parts outside the teeth smaller, so as to avoid the tooth model cut out later being too sharp and causing discomfort to the user's gums after installation.
  • a series of seed points formed a curve among the seed points that the edge of the tooth passes through. Starting from the initial seed point, average it with the second seed point to get the first midpoint, average the first midpoint with the third seed point to get the second midpoint, average the second midpoint with the fourth seed point to get the third midpoint. Repeat the above steps until the last seed point of the edge of the tooth. Connect all the midpoints in order to get the smoothed edge of the tooth.
  • Figure 3 is a two-dimensional projection image of an exemplary three-dimensional tooth model.
  • the teeth and gums in Figure 3 are the three-dimensional tooth model obtained by oral scanning.
  • a two-dimensional projection image as shown in Figure 3 is obtained.
  • the direction of the 3D tooth model can be adjusted.
  • the purpose of adjusting the direction of the 3D tooth model is to project the 2D projection image so that the tooth part of the 3D tooth model can appear as much as possible on the 2D projection image.
  • the three-dimensional tooth model When adjusting the direction, the three-dimensional tooth model can be input and aligned in the positive direction of the Z axis.
  • the alignment method can adopt any alignment method in the art, and no excessive limitation is made here.
  • the table can be the X-axis and Y-axis of the three-dimensional rectangular coordinate system, and the normal of the table facing the teeth is the Z-axis.
  • Figure 3 can be a top view of the three-dimensional tooth model, that is, a two-dimensional projection image obtained by projecting the three-dimensional tooth model onto the table.
  • the recognition model can be used to recognize the two-dimensional projection image, so that the tooth area on the two-dimensional projection image can be recognized.
  • a box can be used for preliminary segmentation to obtain a preliminary segmentation area of a single tooth.
  • the box can be the minimum axial bounding box containing the mask area, or it can be an oblique box.
  • the preliminary segmentation area can be a regular shape or an irregular shape, for example, the box 402 in FIG. 4, That is, the preliminary segmentation area covers the identified tooth area.
  • the preliminary segmentation area includes the tooth area and part of the non-tooth area.
  • the box in Figure 4 is only an example. For example, as shown in Figure 5, Figure 5 is a preliminary segmentation area.
  • a single tooth 3D model can be obtained by performing preliminary segmentation on the 3D tooth model using preliminary segmentation areas.
  • 601 is a 3D model of a tooth
  • 602 is a partial 3D model of an adjacent tooth
  • 603 is a 3D model of the gums near the tooth.
  • the tooth edge is determined according to the original seed point 601, one tooth corresponds to one tooth edge, and the gingival edge is determined according to the non-target seed points 602 and 603, and the above-mentioned gingival edge also corresponds to the tooth one by one.
  • the minimum cut algorithm is used to segment the flow network composed of the seed points.
  • the nodes connected to the source point s in the segmented flow network are the seed points of the teeth, and the nodes connected to the sink point t in the segmented flow network are the seed points of the gingiva.
  • the teeth and gingiva can be accurately segmented.
  • the geometric features corresponding to the edges are weighted averaged, and the weighted average is obtained as the capacity of each edge, so that the segmentation of teeth and gums can be achieved through the minimum cut algorithm.
  • the actual tooth area in the three-dimensional model can be determined.
  • the tooth area on the two-dimensional projection image corresponds to the three-dimensional tooth model, which can correspond to the tooth part on the three-dimensional tooth model. All points on the tooth part are filtered, and the points whose normal direction intersects with the original model are removed, and the remaining points after filtering are used as the original seed points.
  • the surface (tooth part and non-tooth part, the surface of the entire model) of a three-dimensional tooth model is composed of triangular facets (it can also be composed of quadrilateral, pentagonal, etc. polygonal facets).
  • the triangular facets are not completely located on the same plane, and there are angles between each other.
  • the vertices of the triangular facets are seed points.
  • the remaining points are used as seed points. If the corresponding point of the seed point on the three-dimensional tooth model on the two-dimensional projection image is located in the tooth area as shown in FIG4 , the seed point is used as the original seed point.
  • the original seed point can be expanded to obtain the target seed point.
  • the purpose of expanding the original seed point is that the original seed point may not include all the tooth parts on the 3D tooth model, so the tooth parts of the 3D tooth model are covered by expansion.
  • the expansion of the original seed point is divided into multiple stages.
  • the original seed point is expanded by curvature. Expanding the original seed point actually means checking whether there are seed points adjacent to the original seed point that can be used as the target seed point together with the original seed point.
  • the target curvature threshold is the sum of the initial curvature threshold and a preset value. Therefore, for the seed point adjacent to the original seed point, first determine whether the curvature is less than the initial curvature threshold. If the curvature is less than the initial curvature threshold, it is used as the first seed point together with the original seed point. Then, if the curvature of the seed point adjacent to the first seed point is less than or equal to the target curvature threshold, it is used as the target seed point together with the first seed point. Thus, the two extensions of the original seed point are completed.
  • the first stage of extension is not over. Because the original seed point has been extended twice in the first stage of extension, there may be too many points extended, which exceed the tooth part. At this point, it is necessary to determine whether the corresponding position of the extended point on the two-dimensional projection image falls into the tooth area. If it falls into the tooth area, it means that the extension of the original seed point on the three-dimensional tooth model does not exceed the tooth part. If it does not fall into the tooth area, the corresponding seed point is no longer used as the target seed point. At this point, the first stage of extension is over.
  • the target seed points in the first stage can be expanded by the initial curvature threshold and height.
  • the seed points adjacent to the target seed points obtained after the expansion in the first stage can be used as the seed points to be expanded.
  • the curvature of these seed points is less than or equal to the initial curvature threshold, and the height is greater than the preset standard height. If both meet the requirements, then The seed points are also used as target seed points. If one condition is not met, they are not used as target seed points.
  • the preset standard height can be the height of the current point.
  • these seed points must also be within the target area on the two-dimensional projection image.
  • the target area is the area after the tooth area is enlarged. That is, on the two-dimensional projection image, the tooth area is slightly enlarged to obtain the target area. Then, during the second stage of expansion, if the expanded seed points correspond to the two-dimensional projection image and exceed the target area, the seed points that exceed the target area are not used as target seed points, that is, the expansion is rolled back. At this point, the second stage of expansion is completed.
  • each tooth seed point is first expanded according to the given curvature threshold, and it is ensured that it does not exceed the tooth area range on the two-dimensional projection image.
  • These points are used as the initial points of each tooth; based on the initial point, it is expanded again according to the given (curvature threshold + 0.2 to get the target curvature threshold), and it is ensured that it does not exceed the bounding box range of the tooth area;
  • the expansion is performed based on a given curvature threshold and height, and is ensured not to exceed the bounding box of the tooth area + the expanded range (i.e., the target area); each tooth boundary point is extended outward once (extended to the gaps between teeth and near the gum line), and the extension distance can be preset.
  • the initial seed point is not the final tooth area, and needs to be continuously diffused.
  • This solution forms the diffusion range based on the two-dimensional projection.
  • the bounding box of the tooth area + the expanded area refers to the range to which the seed point can diffuse.
  • the bounding box of the tooth area + the expanded area forms the diffusion range based on the two-dimensional projection image, and the diffusion range can be controlled by the algorithm parameters.
  • the reason for multiple extensions of the tooth area is not obvious.
  • the height is limited at the end of the tooth area extension to reduce the possibility of the tooth extending downward to the gums.
  • the tooth boundary is extended outward once because the teeth obtained by segmentation directly according to the curvature threshold will be much smaller than the original teeth due to the small curvature near the gum line, so extension is required.
  • the parts outside the teeth can also be expanded. That is, the points on the three-dimensional tooth model that correspond to the points outside the target area on the two-dimensional projection image are used as the third seed points.
  • the third seed points are points outside the tooth part, such as points on the gum part or the gaps between the teeth. This part of the seed points needs to be expanded toward the tooth part. When expanding, it can be expanded according to the curvature. If the curvature of the seed point adjacent to the third seed point is greater than the required curvature threshold, then this part of the seed points and the third seed point are used together as the third seed points, that is, the points outside the teeth.
  • the seed point may have expanded to the tooth part. Therefore, it is necessary to fall back and use the third seed point located in the target area as a non-third seed point.
  • the boundary between the target seed points obtained by the expansion of the first and second stages and the third seed points obtained after the expansion can be used as the boundary of the tooth.
  • the target seed point and the third seed point may be surrounded, such as the target seed point surrounds a part of the third seed point. This is because there are holes on the teeth, and the holes are identified as third seed points. In this case, the points in the area surrounded by the target seed point are also used as target seed points.
  • the tooth is segmented according to the target seed point and the third seed point to obtain a single tooth.
  • the tooth boundary can be smoothed before segmenting the tooth, and the smoothed boundary will be more gentle.
  • S1-S4 are the points of the tooth boundary, the midpoint A1 between S1 and S3, the midpoint A2 between A1 and S4, etc., and all the midpoints A1-An are connected to obtain the smoothed boundary.
  • the teeth can be sorted.
  • the sorting is shown in Figure 11.
  • the average value of the tooth midpoints of the three-dimensional tooth model is used as the starting point, and the midpoint of each tooth is used as the midpoint, which can be connected to multiple vectors.
  • the straight line between the two teeth with the largest tooth midpoint is used as the target line (the dotted line in Figure 11), and the angle between the vector and the target line is checked, and the vectors are sorted according to the angle.
  • the above-mentioned solution for automatically arranging misaligned teeth can be: establishing a tooth arrangement coordinate system; defining the characteristic points of a single tooth and establishing a local tooth coordinate system; on this basis, analyzing the position and posture of each tooth in the dentition from a low-dimensional perspective, using a weighted fitting optimization method to calculate the coordinate translation of the tooth and the rotation of the local coordinate axis, forming an associated constraint between the tooth posture and the spatial dentition curve, and combining the collision detection method of a rectangular bounding box to design a method based on the steepest descent.
  • the iterative algorithm of the method adjusts the tooth position within the constraints of the spatial dentition curve to complete the automatic arrangement of teeth.
  • a method for manufacturing a dental instrument is also provided, and the method for manufacturing a dental instrument may include the following steps:
  • Step 1 Get a digital tooth model.
  • the tooth model may be the target tooth model in the gum line extraction method embodiment. Accordingly, the tooth model may be obtained in the same manner as the target tooth model, that is, it may be obtained by mouth scanning or by traditional impression taking, which is not limited here.
  • Step 2 pre-process the digital tooth model.
  • the pre-processing action may include the steps of the three-dimensional tooth model segmentation method to obtain segmented teeth.
  • the pre-processing operation may also include the steps of the gum line extraction method to identify the gum line, and may further include converting the gum line into a cutting line for cutting the initial equipment in the subsequent step to obtain the dental instrument, or in some application scenarios, it can also be applied to automatic tooth segmentation, gum crown separation, etc.
  • the gum line extraction method can be the same as the gum line extraction method in the above-mentioned embodiment. Please refer to the above-mentioned content for relevant details, which will not be repeated here.
  • the three-dimensional tooth model segmentation method in the pre-processing may further include the following steps:
  • Step S222 obtaining the gum line of the three-dimensional tooth model and the tooth area of the three-dimensional tooth model
  • Step S224 extracting the peak points of the gum line, and pairing the peak points to obtain a peak point combination
  • Step S226, determining a segmentation path between teeth in the tooth region based on the combination of peak points
  • the purpose of the above three-dimensional tooth model segmentation method is to accurately segment each tooth on the three-dimensional tooth model into individual teeth.
  • the above-mentioned 3D tooth model can be an oral scan model, which refers to a 3D model including the user's teeth generated by scanning the inside of the user's mouth.
  • the data of the 3D tooth model is stored in a computer or server, and the shape or style of the 3D tooth model can be displayed on a display screen for easy viewing by the doctor.
  • the gum line can be a dividing line between the tooth area and the gum area on the three-dimensional tooth model.
  • the gum line can be determined in a variety of ways, for example, it can be determined by neural network model recognition, or by identifying a two-dimensional projection image of a three-dimensional tooth model.
  • the peak point of the above three-dimensional tooth model can be a point on the gum line of the three-dimensional tooth model that is higher than the adjacent points. For example, on the gum line of the three-dimensional tooth model, a straight line formed by multiple points continues to rise and starts to drop at one point, then this point is a peak point.
  • the peak point can be understood as the highest point of the gums located at the gap between two teeth.
  • the pairing of the crest points can determine every two crest points as a crest point combination, and the two crest points in a crest point combination are used to determine the boundary between two teeth.
  • Each crest point combination includes two crest points.
  • a segmentation path between two teeth in the tooth region can be determined according to the two peak points.
  • the tooth area can be segmented according to the segmentation path to obtain a single tooth.
  • there are peak points and trough points on the gum line between the teeth and gums of the three-dimensional tooth model (there are also peak points and trough points on the other side of the tooth not shown in Figure 13).
  • the peak point is the higher point on the gum part between the two teeth (it may be the highest point, or not the highest point, located on the gum line).
  • the trough point is the lower point on the gum line.
  • the present invention obtains the gum line of the three-dimensional tooth model and the tooth area of the three-dimensional tooth model, extracts the peak points of the gum line, and pairs the peak points to obtain the peak point combination, and then determines the segmentation path according to the peak point combination, and segments the tooth area according to the segmentation path to obtain the target segmented tooth method, so that the three-dimensional tooth model can be segmented.
  • Each tooth is segmented accurately, achieving the effect of accurately dividing the three-dimensional tooth model.
  • obtaining a tooth region of a three-dimensional tooth model includes: segmenting the three-dimensional tooth model based on a gum line to obtain the tooth region of the three-dimensional tooth model.
  • the three-dimensional tooth model before dividing the three-dimensional tooth model, can be segmented using the gum line, and after segmentation, the tooth region of the three-dimensional tooth model is obtained. Segmenting the three-dimensional tooth model can be regarded as hiding or not processing the gum region of the three-dimensional tooth model, and when dividing the three-dimensional tooth model later, the gum region can be prevented from affecting the accuracy of dividing the teeth.
  • the method before obtaining the gum line of the three-dimensional tooth model, the method further includes: determining an initial orientation of the three-dimensional tooth model; and adjusting the three-dimensional tooth model from the initial orientation to a target orientation.
  • the direction of the three-dimensional dental model can be adjusted.
  • the purpose of adjusting the direction of the three-dimensional dental model is to "straighten" the three-dimensional dental model, so that all three-dimensional dental models can be adjusted to a fixed orientation, making curvature recognition more accurate.
  • the initial orientation may be different for all three-dimensional dental models. If the desktop is taken as a horizontal plane, the direction of the upward normal of the desktop can be taken as the target orientation. Then, after the three-dimensional dental model is adjusted to the target orientation, it can be in a state where the teeth are facing upward. After obtaining the three-dimensional dental model, the orientation of the three-dimensional dental model can be adjusted to the target orientation, so that the directions of all the acquired three-dimensional dental models can be unified.
  • determining the initial orientation of the three-dimensional tooth model includes: if the three-dimensional tooth model is a non-closed model, determining the minimum directed bounding box of the three-dimensional tooth model; and determining the initial orientation of the three-dimensional tooth model according to the minimum directed bounding box.
  • the three-dimensional tooth model can be divided into two types, the first type is a closed model, and the second type is a non-closed model.
  • the closed model is formed by adding a plane as the bottom surface after the edge is stretched.
  • a minimum directed bounding box can be used to enclose the 3D tooth model.
  • the two faces with the largest area of the minimum directed bounding box are the side of the 3D tooth model facing the teeth and the side of the gum part away from the teeth.
  • the direction of one normal vector can be selected from the two largest normal vectors of the two faces of the minimum directed bounding box as the target orientation.
  • determining the initial orientation of the three-dimensional tooth model based on the minimum directed bounding box includes: confirming a preset axis direction based on the two faces with the largest area of the minimum directed bounding box; based on the preset axis direction, obtaining a first average coordinate value of the boundary edge of the three-dimensional tooth model and a second average coordinate value of the vertices of the polygonal facets of the three-dimensional tooth model; and determining the initial orientation based on the first average coordinate value and the second average coordinate value.
  • the preset axis may be a preset direction, such as a coordinate axis of a three-dimensional space or other directions. Any direction is taken as the preset direction, and the direction of the preset axis is the same as the preset direction.
  • the first average coordinate value of the boundary edge of the three-dimensional tooth model and the second average coordinate value of all vertices can be obtained.
  • the boundary edge is the boundary between the teeth and gums.
  • the first average coordinate value of the boundary edge can be understood as the average value of the dividing points of the teeth and gums
  • the second average coordinate value can be understood as the center of gravity of the three-dimensional tooth model.
  • the center of gravity of the three-dimensional tooth model is biased toward one side of the tooth area. Therefore, of the two largest faces of the minimum directed bounding box that encloses the three-dimensional tooth model, the face that is closer to the second average coordinate value is the side where the teeth are located. Therefore, the initial orientation of the three-dimensional tooth model can be determined.
  • determining the initial orientation of the three-dimensional tooth model includes: when the three-dimensional tooth model is a closed model, determining the polygonal patch group to which the polygonal patches on the surface of the three-dimensional tooth model belong, wherein the plane angle between any two polygonal patches in the same polygonal patch group is less than a first threshold; and determining the initial orientation of the three-dimensional tooth model based on the sum of the areas of the polygonal patches in the polygonal patch group.
  • the closed model can be obtained by filling a non-closed model, or by directly scanning the inside of the oral cavity to obtain a closed three-dimensional tooth model.
  • the closed three-dimensional tooth model can determine the initial orientation by calculating the area of a polygonal patch group composed of polygonal patches.
  • the polygonal The polygonal patch group to which the shape patch belongs includes: selecting any first patch from all patches that have not been divided into the polygonal patch group as a patch in a polygonal patch group, and determining the polygonal patch whose plane normal angle with the first patch is less than a first threshold as a patch in the same polygonal patch group as the first patch, wherein, in an initial case, all polygonal patches on the three-dimensional tooth model are not divided into the polygonal patch group; continuing to select any target patch from all patches that have not been divided into the polygonal patch group as a patch in another polygonal patch group, and determining the polygonal patch whose plane normal angle with the target patch is less than the first threshold as a patch in the same polygonal patch group as the target patch, until all polygonal patches are divided into the polygonal patch group.
  • determining the polygonal patch group to which the polygonal patches on the surface of the three-dimensional tooth model belong includes: selecting any first patch from all patches that have not been divided into the polygonal patch group as a patch in a polygonal patch group, and determining the polygonal patch whose normal has an angle with the normal of the first patch less than a first threshold as a patch located in the same polygonal patch group as the first patch, wherein, in an initial case, all polygonal patches on the three-dimensional tooth model are not divided into the polygonal patch group; selecting any target patch from all faces that have not been divided into the polygonal patch group as a patch in another polygonal patch group, and determining the polygonal patch whose normal has an angle with the normal of the target patch less than the first threshold as a patch located in the same polygonal patch group as the target patch, until all polygonal patches are divided into the polygonal patch group.
  • the polygonal patches may be grouped to obtain different polygonal patch groups, and the angles between the polygonal patches in the polygonal patch groups are smaller than a first threshold.
  • any polygonal patch among all polygonal patches can be used as a polygonal patch in a polygonal patch group. Then, with the polygonal patch as the base patch, all polygonal patches whose face angle with the polygonal patch is less than the first threshold or whose normal angle with the normal of the polygonal patch is less than the first threshold are added to the polygonal patch group. Then, among the remaining polygonal patches, the face angle of each polygonal patch with the face angle of the base patch is greater than or equal to the first threshold. Repeat this operation until all polygonal patches belong to a polygonal patch group.
  • the polygonal patches in each polygonal patch group can be regarded as being on the same plane.
  • the polygonal patch in the polygonal patch group with the largest sum of areas is regarded as the gum side of the closed model, and the other side is the tooth side, thereby determining the initial orientation of the three-dimensional tooth model.
  • obtaining the gum line of the three-dimensional tooth model includes: extracting features of the three-dimensional tooth model; and identifying the features to obtain the gum line.
  • a neural network model can be used to extract the gum line of the three-dimensional tooth model. After the data of the three-dimensional tooth model is input into the neural network model, the neural network model extracts the features of the three-dimensional tooth model, performs convolution and pooling on the features, and outputs the recognized gum line.
  • extracting the crest point of the gum line includes: determining the coordinate value of each point on the gum line based on the target orientation; determining the crest point from the gum line according to the coordinate value, wherein the coordinate value of the crest point is greater than the coordinate value of the adjacent point of the crest point on the gum line.
  • the orientation of the three-dimensional tooth model is adjusted to the target orientation, for a point on the gum line, if the coordinate value in the target direction is greater than that of an adjacent point, then the point is regarded as a peak point.
  • the two peak points cannot be too close to each other. If they are too close, it means that the peak point may be determined incorrectly.
  • pairing the peak points to obtain the peak point combination includes: dividing the peak points into a first peak point group and a second peak point group according to the positional relationship between the peak points and the three-dimensional tooth model; taking each peak point in the first peak point group as a first peak point, and taking the first peak point and the second peak point in the second peak point group as a group of peak point combinations, wherein the second peak point is the peak point in the second peak point group that is closest to the first peak point, and the angle between the line connecting the first peak point and the second peak point and the tooth midline of the three-dimensional tooth model is greater than a third threshold.
  • the tooth center line is the connecting line of the midpoints of the teeth on the three-dimensional tooth model. If the crest points pass through the connecting line, it means that the crest points on both sides of the tooth are paired, and the crest points on one side of the tooth are not paired.
  • the method further includes: deleting the successfully paired peak points from the first peak point group and the second peak point group; taking each peak point in the second peak point group as a third peak point and combining the third peak point and the fourth peak point in the first peak point group as a group of peak points, wherein the fourth peak point is the peak point in the first peak point group that is closest to the third peak point, and the angle between the line connecting the third peak point and the fourth peak point and the tooth midline of the three-dimensional tooth model is greater than a third threshold.
  • pairing can be performed one by one starting from the peak points on one side of the tooth in the three-dimensional tooth model. For each peak point on one side of the tooth in the three-dimensional tooth model, a peak point with the smallest distance is selected from the other side of the tooth and the angle between the line connecting the two peak points and the tooth midline meets the preset threshold condition. The two peak points are combined into a peak point combination. After the pairing is completed, all successfully paired peak points are deleted, and then pairing is performed again starting from the other side of the tooth, thereby achieving pairing of all peak points.
  • the peak point with the smallest distance is found only based on the distance, it is easy to make matching errors in the case of bevel teeth or incisors.
  • the above matching method can avoid matching errors in this scenario by using whether the angle meets the preset threshold condition.
  • pairing fails such as the distance between the two peak points is too large, or one peak point on one side can be paired with both peak points on the other side.
  • determining the segmentation path between teeth in the tooth region based on the peak point combination includes: determining each peak point combination as a current combination; and determining the shortest distance between two peak points in the current combination on the three-dimensional tooth model as a segmentation path.
  • a segmentation path may be determined according to the peak point combination.
  • the segmentation path may be the shortest path between two peak points on the tooth.
  • determining the segmentation path between teeth in the tooth area includes: determining each peak point combination as the current combination; determining the shortest distance between two peak points in the current combination on the three-dimensional tooth model as the first path; correcting the first path by superimposing curvature to obtain a target path; and determining the target path as the segmentation path.
  • the segmentation path can be used as the path for segmenting the teeth, or the segmentation path can be adjusted. After the shortest path between two peak points on the teeth is obtained, the path can be corrected by superimposing curvature, and the corrected path is used as the segmentation path.
  • the first path is corrected by superimposing curvature to obtain a target path, including: taking each point on the first path as a current point, and determining a replacement point for the current point within a preset range on the three-dimensional tooth model; wherein the replacement point is a point that meets the curvature threshold requirement; and taking the line connecting the replacement points as the target path.
  • the shortest path is corrected by superimposing curvature, that is, the points on the path are locally adjusted according to the curvature, and each point on the shortest path is corrected. Each point is used as the current point. If there is a point that meets the curvature threshold requirement within a preset range near the current point on the three-dimensional tooth model, the current point is replaced with the point that meets the curvature threshold requirement. The replaced points are connected as the segmentation path.
  • the 3D tooth model segmentation method of the present application can be mainly used in the scene of correction display. After the teeth are segmented, the tooth position is adjusted according to the correction plan, and the corrected teeth are generated and displayed. Furthermore, the tooth segmentation technology can also be applied to the design of orthodontic guides, the design of substitute dental models, and the design of temporary teeth.
  • the user's oral cavity can be scanned first to obtain data of a three-dimensional tooth model of the user's teeth.
  • the data can be displayed to the user and the doctor through a display screen, so that the doctor can explain the condition of the teeth to the user.
  • the 3D dental model of the user obtained by mouth scanning can be a closed or non-closed model.
  • the 3D dental model includes the tooth area and the gum area, and the boundary between the tooth area and the gum area is not marked. Therefore, if you want to separate the teeth, you need to determine the dividing line between the teeth and the gums (gum line) and the segmentation path between the teeth.
  • the scanning data output by the 3D scanner is obtained; based on the identified model features, the model is automatically aligning; based on the automatically aligning model, the characteristic values of the three-dimensional dental mold are extracted by the geometric algorithm of curvature, and the extracted characteristic values are filtered and denoised to obtain the gum line and its peak points, and the tooth part and the gum part are separated to obtain the pure tooth part; the peak points of the gum line are used for pairing, and the segmentation path between teeth is determined by the curvature and shortest path method to obtain all segmented teeth.
  • the above-mentioned three-dimensional tooth model segmentation method can be applied to the fully automatic orthodontic production process. First, a digital three-dimensional model of the patient is obtained, and the above-mentioned segmentation method is performed on the digital three-dimensional model to obtain a single tooth model, and then the misaligned teeth are automatically arranged, and multiple sets of corrective tooth models are generated. At the same time, positioning parts and placement identification information are set for multiple sets of corrective tooth models. The multiple sets of corrective tooth models are printed, and after the printing is completed, the preheated polymer material film is pressed on the dental mold to form a shell-shaped film.
  • the bottom of the dental mold (with a film attached to the surface of the dental mold) is photographed by an image acquisition terminal (CCD), and then the identification information is identified by OCR recognition technology, and the identification information is sent to the server, and the server calls out the corresponding marking instruction or cutting instruction in the database according to the identification information.
  • the marking instruction When executing the marking instruction, the 3D label to be printed is determined; the boundary constraint of the 3D label is determined, and the label area is determined from the bottom plate according to the boundary constraint, and the label area is used to print the 3D label; the 3D label is perforated to the label area.
  • the image acquisition terminal When executing the cutting instruction, acquires an image of the bottom of the product to be cut; identifies the identification code in the image acquired by the image acquisition terminal, and sends the identification code or the identification information contained in the identification code to the receiving terminal; the receiving terminal matches the corresponding cutting instruction, and the cutting execution terminal performs a cutting operation on the product to be cut according to the cutting instruction.
  • a corresponding identification code is set on the bottom surface of each product to be cut, and different products correspond to different identification codes.
  • the product to be cut is placed above the image acquisition terminal. By acquiring an image of the bottom of the product to be cut, the identification code in the image can be identified.
  • the identification code corresponds to the corresponding cutting instruction. According to the cutting instruction, the cutting execution terminal performs the corresponding cutting operation, removes the excess part and finally obtains the product.
  • the above-mentioned solution for automatic arrangement of misaligned teeth can be: establishing a tooth arrangement coordinate system; defining the characteristic points of a single tooth and establishing a local tooth coordinate system; on this basis, analyzing the position and posture of each tooth in the dentition from a low-dimensional perspective, and using the weighted fitting optimization method to calculate the coordinate translation of the tooth and the rotation of the local coordinate axis, respectively, to form the associated constraints between the tooth posture and the spatial dentition curve, and combined with the collision detection method of the rectangular bounding box, designing an iterative algorithm based on the steepest descent method to adjust the tooth posture within the constraint range of the spatial dentition curve to complete the automatic arrangement of teeth.
  • FIG. 15 is a flow chart of this embodiment.
  • segmenting the three-dimensional tooth model is mainly divided into the following steps:
  • Importing a model means importing a 3D tooth model.
  • a 3D tooth model is a model obtained by oral scanning. After the 3D tooth model is imported into the system, the system can display the structure of the 3D tooth model on the display screen.
  • the three-dimensional tooth model can be a scanned model (non-closed model) output by a 3D scanner, and the input three-dimensional tooth model can be in any direction.
  • any other type of tooth model with a plane does not affect the implementation of the present invention.
  • the 3D tooth model is a digital 3D body composed of a series of polygonal patches.
  • the orientation of the 3D tooth model can be adjusted to the target orientation first, so as to straighten the 3D tooth model.
  • the imported 3D tooth model its features are identified.
  • the purpose of identifying features is to find the alignment angle of the 3D tooth model. Due to the different application types of different dental models, their features are inconsistent. At the same time, according to different dental model types, they can be classified by feature extraction. When aligning, it is necessary to determine the initial orientation of the 3D tooth model and then adjust it to the target orientation.
  • the model is a mouth scan model, which is a non-closed model
  • the oriented bounding box (OBB) technology is first used to determine the bounding box of the model, and the direction of a normal line of the two largest surfaces of the six surfaces of the bounding box is determined as the final tooth orientation.
  • This method determines the size and direction of the box based on the geometric shape of the object itself, and the box does not need to be perpendicular to the coordinate axis. In this way, the most suitable and compact bounding box can be selected.
  • the first average coordinate value of the boundary edge of the three-dimensional tooth model and the second average coordinate value of the vertex of the polygonal face of the three-dimensional tooth model are obtained based on the preset axis direction. Then, the side where the second average coordinate value is closer to the face is determined as the side of the tooth orientation, and the initial orientation of the three-dimensional tooth model is determined.
  • the enclosing box is rotated using the rotation matrix so that the model is rotated to the target direction.
  • the method for detecting the largest plane of the tooth is: set a certain polygonal patch, superimpose the set polygonal patch and the 3D tooth model composed of polygonal patches, set the error threshold e, when e is greater than a certain value, it is considered that the set polygonal patch and the patch on the 3D tooth model are not flat; otherwise, they are considered to be in the same plane.
  • the set polygonal patch and a patch of the tooth model are in the same plane, superimpose them together, and continue to look for the next polygonal patch and judge the error threshold. The above steps are repeated until the largest plane of the tooth model is obtained. Determine the current normal vector of the largest plane of the tooth model, and obtain the target normal vector of the largest plane of the tooth model.
  • the rotation angle and rotation axis are solved according to the vector values before and after the rotation (using the cross multiplication method, the current normal vector and the target normal vector are solved to obtain the rotation angle and rotation axis).
  • the cross multiplication method is a binary operation of vectors in vector space, and its operation result is a vector rather than a scalar; using the above rotation angle and its rotation axis, any model can be rotated to the desired spatial position.
  • the purpose of identifying and extracting the gum line of a three-dimensional tooth model is to divide the three-dimensional tooth model according to the gum line.
  • the gum line can be determined in a variety of ways. For example, it can be determined by a neural network model, and the three-dimensional tooth model can be input into the neural network model, and the gum line of the three-dimensional tooth model can be marked by the neural network model.
  • the three-dimensional tooth model can be projected onto a two-dimensional plane, and then the tooth area on the two-dimensional plane can be identified and then corresponded to the three-dimensional tooth model, thereby marking the boundary between the teeth and the gums, that is, the gum line.
  • the geometric calculation method of curvature is used to extract the characteristic values of the tooth model, and then filter and remove noise to obtain the three-dimensional tooth model contour;
  • the curvature plan calculation method is the rotation rate of the tangent direction angle of a certain surface on the dental model to the arc length, indicating the degree of concavity and convexity of the surface, which is also called a feature in this scheme.
  • the characteristic value of the concave and convex area of the dental model can be obtained through the curvature method (the real gum line on the dental model is also reflected by concavity and convexity);
  • the eigenvalue is simplified into a contour line, that is, an initial gum line. Since the initial gum line is not a smooth curve, there will be local folding and deviation.
  • the principal component analysis method is used to process the initial gum line to obtain the points of the main direction. The data is determined by whether the main direction is consistent. For example, the peaks and troughs are the locations where the curve bends and changes direction. These locations can determine the points of the main direction. In view of this situation, this solution uses the principal component analysis method to obtain its main contour shape, determine the points of the main direction of the contour, avoid the non-smooth area of the line segment, and obtain the final smooth gum line required.
  • the tooth part and the gum part of the three-dimensional tooth model can be segmented according to the position of the gum line.
  • the purpose of segmenting the tooth part and the gum part is to make the tooth division more accurate.
  • the points that determine the main direction of the contour are also the peak and trough points. Determine the insertion control point (the peak and trough points are both curved areas), and because the trough point is always lower than the peak point, the peak point can be determined based on the up and down distance (the height difference in the target direction) and the trough point can be filtered to obtain a complete peak point.
  • the two peak points with the shortest distance and whose connecting line passes through the tooth midline are paired as the starting and ending points of the tooth segmentation. For example, as shown in FIG16 , the peak points on both sides of the tooth are paired. (In FIG16 , a dotted line connects a peak point combination, and not all dotted lines are shown).
  • This solution first uses the shortest distance method to determine the initial cutting curve.
  • the shortest distance shown is from the paired starting peak point, along the surface of the triangular face of the dental mold, to reach the terminal peak point with the shortest distance. Since the shortest distance is not the best cutting line, this solution corrects the shortest distance by superimposing the curvature. Since the contact between teeth is uneven, the shortest distance segmentation line must be within the range of large curvature. The final segmentation line is established by superimposing the two methods.
  • the determined segmentation path is shown in Figure 17.
  • the gum line extraction method may include the following steps:
  • Step S202 obtaining a target tooth model, wherein the target tooth model includes a plurality of polygonal patches;
  • Step S204 determining edge categories of the edges of the plurality of polygonal facets, wherein the edge categories include tooth edges and gum edges;
  • Step S206 determining the gum line of the target tooth model according to the edge category of the edge.
  • the target tooth model can be a mouth scan model, which refers to a digital three-dimensional model of the user's teeth and gums generated by scanning the inside of the user's mouth; of course, it can also be a digital three-dimensional model obtained by taking an impression, first taking an impression of the teeth and gums, and then scanning the impression.
  • the data of the three-dimensional tooth model is stored on a computer or server, and the shape or style of the three-dimensional tooth model can be displayed on a display screen for the doctor to view.
  • the above-mentioned polygonal patches are patches that constitute the surface of the target tooth model (the tooth part and the non-tooth part, the surface of the entire model), which can be specifically triangles, quadrilaterals, pentagons, etc.
  • Each polygonal patch is located on a plane, and different polygonal patches can be located on the same or different planes.
  • the surface of the target tooth model is composed of multiple polygonal patches, each of which includes multiple edges, and two adjacent polygonal patches share an edge.
  • the endpoints of the edges are the vertices of the corresponding polygonal patches, and a vertex can be shared by multiple polygonal patches.
  • the target tooth model includes a tooth part and a gum part, and the gum line is the boundary between the tooth part and the gum part.
  • the edges constituting the polygonal patch can be located in the tooth part or the gum part, and thus can be classified into tooth edges and gum edges.
  • the gum line of the target tooth model can be determined according to the edge category of the edge of the polygonal patch, and the determined gum line is used to segment the tooth part and the gum part of the target tooth model.
  • the edge category can be determined by determining the edge category of the polygonal patch on the target tooth model, and then the gum line of the target tooth model can be determined.
  • the edge as a component of the polygonal patch can further refine the target tooth model, thereby improving the accuracy of the gum line extraction of the tooth model.
  • determining edge categories of the edges of the plurality of polygonal patches includes: acquiring target features of the edges of the plurality of polygonal patches; and determining edge categories of the edges of the plurality of polygonal patches according to the target features.
  • the target feature includes a geometric feature.
  • the target feature may also include a non-geometric feature, which is not specifically limited here.
  • the edge category can be determined by extracting the target feature of the edge of the polygonal face and identifying the target feature. It should be noted that when the target feature includes a geometric feature, the topological structure relationship of the space will be considered when classifying the edge, thereby improving the accuracy of gum line recognition.
  • the target features of all edges on the target tooth model may be extracted, or the target features of a portion of the edges may be selected, and the edge category of the edge may be determined by identifying the target features of the extracted edge.
  • the target features of an edge include at least one of the following: a dihedral angle between a first polygonal patch and a second polygonal patch that share the edge, a first curvature value of a first vertex of the edge, a second curvature value of a second vertex of the edge, a first distance from a vertex of the first polygonal patch away from the edge to the edge, a second distance from a vertex of the second polygonal patch away from the edge to the edge, a first opposite angle of the edge in the first polygonal patch, and a second opposite angle of the edge in the second polygonal patch.
  • the target features of an edge include the above seven items.
  • the target features of an edge may include at least one of the following features: the dihedral angle between a first polygonal patch and a second polygonal patch sharing an edge, a first curvature value of a first vertex of the edge, a second curvature value of a second vertex of the edge, a first spatial coordinate of the first vertex, a second spatial coordinate of the second vertex, the length of the edge, an angle between the edge and an adjacent edge, a first distance from a vertex of the first polygonal patch farthest from the edge to the edge, a second distance from a vertex of the second polygonal patch farthest from the edge to the edge, a first opposite angle of the edge in the first polygonal patch, a second opposite angle of the edge in the second polygonal patch, a first normal of the first vertex, a second normal of the second vertex, a normal of the first polygonal patch, and a normal of the second polygonal patch.
  • the dihedral angle between the first polygonal patch and the second polygonal patch of the above-mentioned shared edge is the angle between the two polygonal patches that share an edge.
  • the above-mentioned first vertex and the second vertex are the two endpoints of the edge, and the above-mentioned curvature value is the value of the curvature of the patch corresponding to the vertex at the vertex.
  • Curvature is the rotation rate of the tangent direction angle to the arc length at a point on the curve, which is defined by differentiation and indicates the degree to which the curve deviates from the straight line. A numerical value indicating the degree of curvature of the curve at a certain point.
  • the above-mentioned spatial coordinates are the coordinates in the three-dimensional rectangular coordinate system where the vertex is located.
  • the three-dimensional rectangular coordinate system can be the coordinate system where the target tooth model is located.
  • the plane where the base plate of the target tooth model is located can be defined as the xOy plane, and the normal direction of the plane can be defined as the z-axis.
  • the direction of the coordinate axis can be predetermined.
  • the gingival surface of the target tooth model is used as the plane where the X-axis and the Y-axis are located, and the tooth direction is used as the Z-axis direction.
  • the length of an edge is the distance between the two endpoints of the edge; because there may be multiple adjacent edges to an edge, there may be multiple angles between the edges, ranging from 0 to 180 degrees; when the polygon is a quadrilateral or more polygonal, there may be more than one distance from the vertex away from the edge to the edge, and the distance from each vertex to the edge may be different, and when the polygonal patch is a triangular patch, there is only one vertex away from the edge, so the distance from the vertex to the edge is also only one; similarly, the edge in the polygonal patch may have one or more diagonals, and when there are multiple diagonals, the angle of each diagonal may be different.
  • the polygonal patches are triangular patch 302, quadrilateral patch 304, and pentagonal patch 306, and each edge of the three polygonal patches corresponds to at least one or at least two of the above features.
  • the angle between the normals of the two polygonal patches can be obtained by cross product based on the normals of the polygonal patches, and the dihedral angle between the two polygonal patches is complementary to the angle between the normals, and the dihedral angle of the two polygonal patches can be obtained by subtracting the angle between the normals from 180°;
  • the curvature value of the two vertices can be obtained according to the curvature geometry calculation method or other methods;
  • the distance from the vertex to the edge of the polygonal patch away from the edge can be obtained according to the distance from the point to the straight line;
  • two vectors are obtained according to the orientation of the two edges, and the angle between the two edges at the vertex of the polygonal patch can be obtained by dot product based on the two vectors, that is, the diagonal of the edge in each polygonal patch.
  • determining the edge category of the edge includes: upgrading the target feature of the initial dimension of the edge to obtain a first feature of the first dimension; reducing the first feature to obtain a second feature of the target dimension; and determining whether the edge is a tooth edge or a gum edge according to the value of the second feature.
  • the target dimension can be two-dimensional, corresponding to the tooth edge and the gum edge, respectively.
  • it can also be multi-dimensional, as long as the edge category of the corresponding edge can be determined, which is not limited here.
  • the initial dimension of the target features can be determined according to the type and quantity of the extracted target features.
  • the target features are first dimensionally upgraded to the first dimension to obtain the first feature, and then the first feature is dimensionally reduced to the target dimension.
  • the dimension is first upgraded and then reduced, the first feature after dimension reduction is no longer the target feature, but the second feature of the target dimension, based on which the edge category of the corresponding edge can be obtained.
  • the seven-dimensional target features of the corresponding edge are input into the edge classification network to classify the corresponding edge.
  • the input dimension is first seven dimensions, corresponding to seven geometric features, and after one convolution operation, the target edge is The dimension of the target feature is increased to N dimensions, and then k identical convolution operations are performed to keep the dimension of the target feature unchanged, and then the pooling operation is performed to reduce the number of edges. Then the above process is repeated, that is, one dimensionality increase convolution + k same-dimensional convolutions + pooling, so that the dimension of the target feature becomes 2N, 4N, and 8N in turn.
  • the initial dimension is not limited to 7 dimensions, and the number of cycles in the dimension increase and dimension decrease process is not limited.
  • N is an integer not less than 7, such as 7, 8, 9, 10, etc., which can be adjusted according to actual conditions when implementing the specific implementation plan.
  • the convolution and pooling in this step are different from the convolution and pooling operations of the image neighborhood, and the operations are performed based on the edges in the target tooth model.
  • the original edge features are processed according to the following formula (1), where f(a) and f′(a) represent the original features and processed features of edge a respectively.
  • the convolution operation of edge e after processing is defined in formula (2), where ⁇ w k
  • k 0, 1, 2, 3, 4 ⁇ represents the weight parameters to be trained.
  • the pooling operation transforms the five edges (e, a, b, c, d) into two edges (h, i), which is defined in formula (3).
  • the upper pooling operation is the opposite of the pooling operation, transforming the two edges (h, i) into (e′, a′, b′, c′, d′), which is defined in formula (4).
  • C(e) w 0 f′(e)+w 1 f′(a)+w 2 f′(b)+w 3 f′(c)+w 4 f′(d) (2)
  • the classification network needs to be trained to obtain network weight values and some hyperparameter values: such as k and N, etc.
  • multiple groups of geometric feature data and corresponding edge classification category values are input into the edge classification network for training.
  • 500 groups, 1000 groups or 2000 groups of classified tooth model data are input, and each group of tooth model data contains 5000, 10000 or 20000 geometric feature data and classification data corresponding to the edges, so that the classification network can classify the edges according to the geometric feature data of the edges.
  • the method based on edge classification can automatically identify the gum line data by combining deep learning technology, which can greatly improve the recognition accuracy and efficiency, improve the efficiency of the entire invisible orthodontic and other dental diagnosis and treatment processes, and enhance the user experience in diagnosis and treatment.
  • determining the gum line of the target tooth model according to the edge category includes: determining the patch category of the polygonal patch according to the edge category, wherein the patch category includes tooth patch and gum patch; and extracting the gum line based on the common edges of adjacent polygonal patches with different patch categories.
  • the patch category of the polygonal patch can be determined according to the edge category.
  • the tooth patch is a patch located in the tooth part of the tooth model
  • the gum patch is a patch located in the gum part.
  • the patch category is determined according to the edge category of the edge, and the patch category of the polygonal patch can be determined according to all the edges of the polygonal patch, or the patch category of the polygonal patch can be determined according to at least one edge of the polygonal patch.
  • determining the patch category of a polygonal patch based on the edge category includes: when the number of gingival edges in the polygonal patch is greater than the number of tooth edges, determining the polygonal patch to be a gingival patch; when the number of gingival edges in the polygonal patch is less than the number of tooth edges, determining the polygonal patch to be a tooth patch.
  • the patch category of the polygonal patch is determined according to the edge categories of all the edges of the polygonal patch. According to the number of tooth edges and gum edges in the polygonal patch, it is determined whether the polygonal patch is a tooth edge or a gum edge.
  • extracting the gum line based on the common edges of adjacent polygonal patches of different patch categories includes: determining a first feature point based on the vertices of the common edge; determining a target feature point from the first feature point based on a curvature relationship; and determining the gum line based on the target feature point.
  • the patch category of the polygonal patch can be determined according to the edge category of the edge, after the patch category is determined, the common edges of the polygonal patches of different categories are the boundaries between the tooth patch and the gum patch. Therefore, the gum line can be determined according to all the common edges of the tooth patch and the gum patch.
  • a first feature point can be determined from the vertices of the shared edge, and then a target feature point can be determined from the first feature point, and the gum line can be determined based on the target feature point.
  • the target feature point can be obtained by calculating the curvature.
  • a vertex of a common edge can be used as a starting point to search for connected common edges, and the vertices of each connected common edge can be recorded in a counterclockwise or clockwise order as feature points.
  • the first feature point can be all or part of the recorded feature point. Specifically, the first feature point is located on the initial tooth model, so that the extracted gum line has a higher accuracy.
  • the target feature point is a feature point with specific features in part of the first feature point.
  • determining the first feature point based on the vertices of the common edge includes: when the target tooth model is a model obtained by performing edge shrinkage processing on the initial tooth model, determining the nearest point of the vertex on the common edge on the initial tooth model as the first feature point; when the target tooth model is the initial tooth model, determining the vertex on the common edge as the first feature point.
  • the above-mentioned initial tooth model is the tooth model initially obtained.
  • the initial tooth model can be subjected to edge contraction processing to obtain the target tooth model, or the initial tooth model can be directly used as the target tooth model. Whether the initial tooth model is subjected to edge contraction processing affects whether the vertex of the common edge is directly used as the first feature point.
  • the first feature point is the vertex of the common edge, or the closest point of the common edge on the initial tooth model.
  • the recorded feature points can be projected back to the original model, and the point on the original model closest to the recorded feature point can be found as a new feature point.
  • KDTree can be established with the vertices of the initial tooth model.
  • KDTree is a tree data structure that stores instance points in a k-dimensional space for quick retrieval.
  • all recorded feature points can be traversed, and the nearest point can be found in KDTree as a new feature point.
  • the new feature point is the first feature point.
  • determining the target feature point from the first feature points according to the curvature relationship includes: determining a first feature point whose curvature is greater than that of an adjacent first feature point as the target feature point.
  • the target feature point can be determined from the first feature point based on the curvature of the first feature point.
  • Each first feature point corresponds to a curvature.
  • the first feature point whose curvature is greater than the curvature of the adjacent first feature point is used as the target feature point, that is, the point in the curve where the position is more curved is determined as the target feature point, which can better reflect the actual trend of the curve.
  • the curvature of the discrete points can be estimated using the curvature geometry algorithm, and the feature points with large curvature are retained as target feature points.
  • the reason for retaining feature points with large curvature is that points with large curvature are often turning points. Turning points are highly characteristic and representative.
  • determining the gum line based on the target feature points includes: performing an interpolation operation on the target feature points to obtain interpolated target feature points; and connecting the interpolated target feature points in sequence to form the gum line.
  • interpolation operation is performed on the target feature points to compensate for the loss of feature points caused by determining the target feature points from the first feature points, and the target feature points can be combined into a closed curve through interpolation operation.
  • the closed curve is the gum line.
  • B-spline curve interpolation can be performed on adjacent target feature points; all points are connected in sequence, and the resulting curve is the gum line. It should be noted that when interpolating the target feature points, it is not limited to B-spline curves, and other spline curves can also be used, which is not specifically limited here.
  • obtaining a target tooth model includes: obtaining an initial tooth model, wherein the initial tooth model includes multiple polygonal patches, and the number of edges in the initial tooth model is greater than the number of edges in the target tooth model; reducing the number of edges in the initial tooth model to a preset number in a preset manner to obtain the target tooth model.
  • the preset mode may be a shrinking operation or a collapsing operation.
  • the shrinking operation is performed on the initial tooth model to reduce the number of edges in the initial tooth model, thereby obtaining a target tooth model, simplifying the tooth model, and improving the recognition rate of the gum line.
  • the number of edges of the initial tooth model can be reduced one by one or a batch of edges, such as reducing one edge at a time until it is reduced to a preset number or a preset number of times.
  • the number of edges can be reduced by a batch at a time until it is reduced to a preset number or a preset number of times.
  • the imported initial tooth model can be a tooth model in any direction.
  • the initial tooth model is a digital three-dimensional body composed of a series of vertices and polygonal patches.
  • the tooth model is downsampled in the way of edge collapse.
  • the polygonal patches of the tooth model are sparser, the number of polygonal patches is reduced, and the number of edges is also reduced. Downsampling stops until the specified number of edges is reached.
  • the imported initial tooth model often has many edges and the number of edges is inconsistent.
  • the classification network is used for classification in the subsequent steps, the number of edges required is often specific. Therefore, the imported initial tooth model needs to be downsampled, some edges are deleted, and the specified number of edges are retained.
  • the number of edges can be determined by considering the classification speed (the fewer the number of edges, the faster the speed) and the accuracy of the final gum line (the more the number of edges, the higher the accuracy, and the more the edges, the less the information loss of the original model).
  • the main process is as follows:
  • the sum of the distances from the vertex to the adjacent facets is recorded as the error value, with the initial value being 0; wherein the adjacent facet refers to the facet where the vertex is located; the initial error value of 0 means that, before edge contraction is performed, the sum of the distances from the vertex to its adjacent facets is 0.
  • the number of edges in the initial tooth model is reduced to a preset number in a preset manner to obtain a target tooth model, including: for each edge of the tooth model to be processed, calculating the sum of distances from the vertex corresponding to each edge to the adjacent polygonal facets corresponding to the vertex of each edge in the tooth model to be processed after edge shrinkage processing is performed on each edge in each of multiple shrinkage methods; wherein, in the first processing, the tooth model to be processed is the initial tooth model; determining the edge and shrinkage method corresponding to the minimum value among the calculated sums of multiple distances; performing edge shrinkage processing on the edge corresponding to the minimum value according to the shrinkage method corresponding to the minimum value to obtain the target tooth model.
  • the edge corresponding to the minimum value is subjected to edge shrinkage processing according to the shrinkage method corresponding to the minimum value, so as to obtain
  • the target tooth model includes: after edge shrinkage processing, if the number of edges in the obtained tooth model is not greater than the preset number, the obtained tooth model is the target tooth model; if the number of edges is greater than the preset number, the obtained tooth model is determined as the tooth model to be processed to continue the edge shrinkage processing.
  • edge shrinkage processing when the initial tooth model is subjected to edge shrinkage processing, it is also necessary to determine which edge to shrink.
  • This process can be a cyclic process. That is, shrink one edge each time and cycle, or shrink a batch of edges each time and cycle, until the tooth model to be processed after shrinkage meets the conditions. Therefore, in each shrinkage, it is necessary to determine which edge or which batch of edges to shrink.
  • the initial model can be used as the model to be processed, and the edges to be shrunk for the model to be processed can be determined by simulation or calculation first. If a certain edge is shrunk, the sum of the distances from the vertex corresponding to each edge in the shrunk model to the adjacent polygonal facets corresponding to the vertex of the edge in the tooth model to be processed is calculated. If the sum is the smallest, the edge is shrunk according to the strategy. Each time the edge is shrunk, the calculation is repeated, and the edge shrinkage is repeated until the number of edges of the tooth model to be processed meets the preset number requirement. At this point, the edge shrinkage is completed, and the target tooth model is obtained.
  • Step three performing molding processing based on the pre-processed digitized tooth model to obtain a molded tooth model, that is, a physical tooth mold.
  • a shaped tooth model is obtained based on target segmented teeth printing, wherein the shaped tooth model is used to obtain a dental instrument. That is, the target segmented teeth are arranged together to obtain a digital tooth model, and a physical tooth mold is obtained based on a molding process (also called printing).
  • the 3D printing can adopt light-curing 3D printing, such as SLA, DLP, LCD, or other 3D printing methods, such as 3DP, MJF, FDM, Polyjet, etc.
  • the obtained physical dental model may have a hollow base plate, fixed accessories, identification information, etc.
  • the 3D printing process also includes a post-processing step, which can be selected according to the 3D printing method used.
  • the post-processing step can include post-curing, cleaning, and the like.
  • the molding method is not limited to 3D printing, and other molding methods such as injection molding may also be used, which are not limited here.
  • Step 4 Perform film pressing on the solid dental model.
  • the preheated polymer material film can be laminated on the solid dental model to obtain a shell-shaped film covering the solid dental model, that is, the initial device that has not been cut.
  • the shell-shaped film at least covers the tooth part of the solid dental model.
  • an image of a physical dental mold can be collected through an image acquisition terminal, such as a CCD image sensor, and then the identification information can be identified through recognition technologies such as OCR. Then, based on the identification information, the corresponding lamination instructions are called out in the database, and lamination processing is performed based on the dental mold instructions, wherein the lamination instructions may include pressure parameters, diaphragm preheating time, lamination temperature, etc.
  • Step 5 Marking the shell-shaped diaphragm.
  • the shell-shaped membrane is mainly marked based on the identification information added in the pre-processing operation to form a mark on the shell-shaped membrane to obtain an initial device with a mark.
  • an image of the physical dental model can be captured through an image acquisition terminal, such as a CCD image sensor, and then the identification information can be identified through OCR and other recognition technologies. Then, the corresponding marking instructions are called out based on the identification information to mark the shell-shaped membrane based on the parameter information contained in the marking instructions.
  • an image acquisition terminal such as a CCD image sensor
  • the execution of this step is not limited, that is, whether the marking operation is required can be selected according to actual needs.
  • Step 6 Cut the initial equipment to obtain dental instruments.
  • This step mainly involves cutting the initial device based on the cutting line converted from the extracted gum line, thereby removing the useless part of the initial device and retaining only the useful part, thereby obtaining a dental device.
  • the dental device can be Invisible braces.
  • step five and step six is not limited in this embodiment.
  • an image of a physical dental model can be captured through an image acquisition terminal, such as a CCD image sensor, and then the identification information can be identified through OCR or other recognition technologies, and then the corresponding cutting instructions can be called out based on the identification information to cut the initial equipment based on the parameter information contained in the cutting instructions.
  • an image acquisition terminal such as a CCD image sensor
  • a three-dimensional tooth model segmentation device including: a memory 1801 and a processor 1803 , wherein the memory stores a computer program, and when the computer program is executed by the processor, the above-mentioned three-dimensional tooth model segmentation method is executed.
  • a computer program product or a computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium.
  • the processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any of the above embodiments.
  • the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof.
  • the processing unit can be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application, or a combination thereof.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPs digital signal processing devices
  • PLDs programmable logic devices
  • FPGAs field programmable gate arrays
  • general purpose processors controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application, or a combination thereof.
  • the technology described herein can be implemented by a unit that performs the functions described herein.
  • the software code can be stored in a memory and executed by a processor.
  • the memory can be implemented in the processor or outside the processor.
  • the disclosed devices and methods can be implemented in other ways.
  • the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed.
  • the coupling or direct coupling or communication connection between each other shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units.
  • each functional unit in each embodiment of the present application may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit.
  • the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) perform all or part of the steps of the above-mentioned methods of each embodiment of the present application.
  • the aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk.
  • program codes such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Materials Engineering (AREA)
  • Chemical & Material Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Theoretical Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • General Engineering & Computer Science (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Dentistry (AREA)
  • Epidemiology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Computer Graphics (AREA)
  • Architecture (AREA)
  • Multimedia (AREA)
  • Mechanical Engineering (AREA)
  • Optics & Photonics (AREA)
  • Dental Tools And Instruments Or Auxiliary Dental Instruments (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本申请涉及一种三维牙齿模型分割方法和装置。该方法包括：获取三维牙齿模型的二维投影图像，并对二维投影图像进行识别，得到多个牙齿区域，牙齿区域与牙齿一一对应；在三维牙齿模型中，确认与多个牙齿区域对应的原始种子点；在预设范围内对原始种子点进行扩展，得到三维牙齿模型中牙齿的目标种子点；基于各牙齿的目标种子点对三维牙齿模型进行分割，得到分割牙齿。本申请解决了三维牙齿模型分牙准确度低的技术问题。

Description

三维牙齿模型分割方法和装置
本公开要求于2022年12月02日提交中国专利局、申请号为202211543154.6、申请名称“三维牙齿模型分割方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及牙齿正畸技术领域，尤其涉及一种三维牙齿模型分割方法和装置。
背景技术
随着科技信息的发展，制造技术、数字建模技术、材料科学、数控技术等技术迅速发展壮大以及这些学科的相互融合，计算机技术已经越来越多地渗透于医学各领域的教学、科研和临床应用的各个方面且能够较好的互相合作。而随着测量技术地发展及普及，人们能够很方便地获取数字化牙齿模型，其在口腔临床诊断和治过程中起着重要作用。3D打印技术的出现和发展已成为现在的热口之一，3D打印技术应用于医疗领域也越来越屡见不鲜。3D打印在医学领域的应用已有二十余年，广泛应用于口腔种植、骨科、神经外科等手术。
发明人发现：在目前的正畸诊疗的应用场景中，需要将结果快速呈现到用户，增强诊疗信心，即进行牙齿目标位模拟，牙齿目标位模拟指输入患者的扫描数据，通过简单的操作和系统的自动运行能够得到“矫正”后的效果并展示，且能够演示从原始状态向目标状态变化的连续“动画”，以使得医生与患者得到更好的沟通，促使成交。其作为正畸的前置环节，承载辅助医患沟通的主要职责，且目标位模拟的输出结果也是后续环节导出生产的内容之一，而分牙是其中重要的环节。
目前，相关技术中，三维牙齿模型分割技术尚属起步阶段，难以满足高效处理需求。
发明内容
本申请提供了一种三维牙齿模型分割方法和装置，以解决上述三维牙齿模型分牙准确度低的技术问题。
根据本申请实施例的一个方面，本申请提供了一种方法，包括：获取三维牙齿模型的二维投影图像，并对上述二维投影图像进行识别，得到多个牙齿区域，牙齿区域与牙齿一一对应；在上述三维牙齿模型中，确认与上述多个牙齿区域对应的原始种子点；在预设范围内对上述原始种子点进行扩展，得到上述三维牙齿模型中牙齿的目标种子点；基于各上述牙齿的目标种子点对上述三维牙齿模型进行分割，得到分割牙齿。
根据本申请实施例的另一个方面，本申请还提供了一种牙科器械的制作方法，包括：
如上述任一项方法得到分割牙齿；
基于目标分割牙齿打印得到成型牙齿模型；其中，成型牙齿模型用于得到牙科器械。
根据本申请实施例的另一个方面，本申请还提供了一种三维牙齿模型分割装置，包括：
存储器和处理器，存储器存储有计算机程序，计算机程序被处理器运行时执行上述任一项方法。
根据本申请实施例的另一方面，还提供了一种计算机可读的存储介质，计算机可读的存储介质存储有计算机程序，计算机程序被处理器运行时执行任一项上述方法。
本申请在获取到三维牙齿模型之后，通过识别三维牙齿模型的二维投影图像来确定三维牙齿模型上的原始种子点，并对原始种子点进行扩展，得到目标种子点，从而通过目标种子点标记出三维牙齿模型上的牙齿的范围，进一步可以根据目标种子点对三维牙齿模型进行分割，得到分割牙齿，从而实现了准确对三维牙齿模型进行分牙的效果，解决了现有技术中对三维牙齿模型进行分牙准确度低的问题，相较于传统分牙手段，本申请还可以避免对缺牙位 置进行分割。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请的原理。
为了更清楚地说明本申请实施例或相关技术中的技术方案，下面将对实施例或相关技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为根据本申请实施例提供的一种可选的三维牙齿模型分割方法硬件环境示意图；
图2为根据本申请实施例提供的一种三维牙齿模型分割方法流程图；
图3为根据本申请实施例提供的一种三维牙齿模型分割方法的二维投影图像；
图4为根据本申请实施例提供的一种三维牙齿模型分割方法的初步分割区域图；
图5为根据本申请实施例提供的另一种三维牙齿模型分割方法的初步分割区域图；
图6为根据本申请实施例提供的一个初步分割区域对应的三维牙齿模型的局部图；
图7为根据本公开实施例提供的一种三维牙齿模型的种子点构成的流网络的示意图；
图8为根据本公开实施例提供的一种初步分割区域被最小流算法分割后的示意图；
图9为根据本申请实施例提供的一种三维牙齿模型分割方法的三角面片图；
图10为根据本申请实施例提供的一种三维牙齿模型分割方法的平滑边界图；
图11为根据本申请实施例提供的一种三维牙齿模型分割方法的牙齿排序图；
图12为根据本申请实施例提供的一种三维牙齿模型分割方法的流程图；
图13为根据本申请实施例提供的一种三维牙齿模型的波峰点和波谷点示意图；
图14为根据本申请实施例提供的一种三维牙齿模型的分牙示意图；
图15为根据本申请实施例提供的一种三维牙齿模型的流程图；
图16为根据本申请实施例提供的一种三维牙齿模型的波峰点组合的示意图；
图17为根据本申请实施例提供的一种三维牙齿模型的分割曲线图；
图18为根据本申请实施例提供的一种三维牙齿模型分割装置框图。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请的一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本申请保护的范围。
在后续的描述中，使用用于表示元件的诸如“模块”、“部件”或“单元”的后缀仅为了有利于本申请的说明，其本身并没有特定的意义。因此，“模块”与“部件”可以混合地使用。
为了便于描述，以下对本申请实施例涉及的部分名词或术语进行说明：
流网络(Flow Network)：是指一类特殊的加权有向的复杂网络。其中，有向连边表示能量、物质、货币、信息、注意力等流动的方向，连边的权重则表示流量。本申请的流网络是一个由边和点组成的图，其中，存在一个点入度为0，称作源点，一个点出度为0，称作汇点。
源点(source)：发出流的一个点，拥有无限的流，一般用s表示。源点入度为0，出度大于0。
汇点(terminal)：接收流的一个点，一般用t表示。汇点出度为0，入度大于0。
网络流(network-flow)：流网络中の所有边上流量的集合。
最小割算法(Minimum Cut)：是图像分割的经典算法之一，也叫图割(Graph Cut)，用于将图分成两个部分使得两个部分之间的边权和最小。
最大流最小割：是指在一个有向的图中，能够从源点到达汇点的最大流量等于如果从途中减除就能导致网络流中断的边的集合的最小容积和。
任何网络中，最大流的值等于最小割的容量。
为了解决背景技术中提及的问题，根据本申请实施例的一方面，提供了一种三维牙齿模型分割方法的实施例。
可选地，在本申请实施例中，上述方法可以应用于如图1所示的由终端101和服务器103所构成的硬件环境中。如图1所示，服务器103通过网络与终端101进行连接，可用于为终端或终端上安装的客户端提供服务，可在服务器上或独立于服务器设置数据库105，用于为服务器103提供数据存储服务，上述网络包括但不限于：广域网、城域网或局域网，终端101可以为用于获取三维牙齿模型的装置，如口扫模型扫描仪，或者用于接收三维牙齿模型的终端，如手机、电脑等。
本申请实施例中的三维牙齿模型分割方法可以由服务器103来执行，还可以是由服务器103和终端101共同执行。如图2所示，三维牙齿模型分割方法可以包括以下步骤：
步骤S202，获取三维牙齿模型的二维投影图像，并对二维投影图像进行识别，得到多个牙齿区域，牙齿区域与牙齿一一对应；
步骤S204，在三维牙齿模型中，确认与多个牙齿区域对应的原始种子点；
步骤S206，在预设范围内对原始种子点进行扩展，得到三维牙齿模型中牙齿的目标种子点；
步骤S208，基于各牙齿的目标种子点对三维牙齿模型进行分割，得到分割牙齿。
上述三维牙齿模型分割方法的目的在于，将三维牙齿模型上的每一颗牙齿进行准确的分牙，分为单颗的牙齿，牙齿区域可以与牙齿一一对应，当然，一个牙齿区域也可以对应多个牙齿。
上述三维牙齿模型可以为口扫模型，口扫模型指通过扫描用户的口腔内部而生成的包括用户的牙齿部分的三维模型。三维牙齿模型的数据在计算机或服务器上，通过显示屏可以显示三维牙齿模型的形状或样式，以方便医生观看。
上述二维投影图像可以为将三维牙齿模型投影到一个平面上得到的图像。二维投影图像上包含了牙齿区域和非牙齿区域。通过识别二维投影图像，可以识别出多个牙齿区域，从而对三维牙齿模型上的牙齿进行分牙。其中，识别的方法可以采用机器视觉或人工智能的方式对二维投影图像进行识别，得到牙齿区域。
三维牙齿模型的表面由多边形面片组成，如三角面片，一个三角面片包含了3个顶点，所有三角面片的顶点可以视为种子点。上述的原始种子点为三维牙齿模型的所有种子点中，与二维投影图像中的牙齿区域对应的种子点。也就是说，通过原始种子点，可以在三维牙齿模型上初步确定出牙齿的区域。
由于原始种子点确定出的牙齿的区域可能并不准确，因此，可以通过对原始种子点进行扩展的方法，来扩充原始种子点，得到目标种子点。目标种子点所覆盖的区域可以视为三维牙齿模型上的牙齿，通过目标种子点，可以对三维牙齿模型进行准确的分牙。
在识别到牙齿区域后，可以理解为在二维投影图像上已经确定了牙齿所在的位置与区域。然后，将其对应到三维牙齿模型上，确定出三维牙齿模型上的原始种子点。由于种子点是三维牙齿模型的表面的三角面片的顶点，因此，原始种子点可以理解为覆盖了三维牙齿模型上的牙齿的区域。为了保证准确性，进一步对原始种子点进行扩充，得到目标种子点，根据目标种子点对三维牙齿模型进行分割，得到分割牙齿，至此，可以将三维牙齿模型中的牙齿分为单颗的牙齿。
上述方法通过识别三维牙齿模型的二维投影图像来确定三维牙齿模型上的原始种子点，并对原始种子点进行扩展，得到目标种子点，从而通过目标种子点标记出三维牙齿模型上的牙齿的范围，进一步可以根据目标种子点对三维牙齿模型进行分割，得到分割牙齿，从而实现了准确对三维牙齿模型进行分牙的效果。
在一些实施例中，在预设范围内对原始种子点进行扩展，得到三维牙齿模型中牙齿的目标种子点包括：根据原始种子点确定牙齿边缘，牙齿边缘为覆盖牙齿区域的所有的原始种子点的三维牙齿模型的表面的边缘；在三维牙齿模型中，确认与多个牙齿区域以外的区域对应的非目标种子点，并根据非目标种子点确定牙龈边缘，牙龈边缘为覆盖所有的非目标种子点的三维牙齿模型的表面的边缘；在三维牙齿模型的表面上，将远离牙龈边缘的牙齿边缘的一侧的任意一点确定为源点，将远离牙齿边缘的牙龈边缘的一侧的任意一点确定为汇点；根据各源点和汇点构建多个流网络，使得源点和汇点之间的种子点均为流网络的节点，任意两个相邻的节点连接，种子点为三角面片的顶点；采用最小割算法对各流网络进行分割，并将分割后的各流网络中与源点连接的节点确定为对应的各牙齿的目标种子点。
根据原始种子点确定牙齿边缘，一个牙齿区域对应一个牙齿边缘，根据非目标种子点确定牙龈边缘，牙龈边缘只有一个，牙齿边缘和牙龈边缘中间的部分不清楚是牙齿还是牙龈，采用最小割算法对种子点组成的流网络进行分割，分割后的流网络中与源结点连接的结点是牙齿的种子点，分割后的流网络中与汇点连接的结点是牙龈的种子点，可以准确分割出牙齿与牙龈。
在一些实施例中，在三维牙齿模型中，确认与多个牙齿区域以外的区域对应的非目标种子点，并根据非目标种子点确定牙龈边缘，包括：根据多个牙齿区域对三维牙齿模型进行初步分牙，得到多个初步分割区域，初步分割区域一一对应地覆盖牙齿区域；在各初步分割区域中，确认与各牙齿区域以外的区域对应的各组非目标种子点；根据各组非目标种子点确定对应的各牙龈边缘，牙龈边缘为覆盖一组非目标种子点的三维牙齿模型的表面的边缘，牙龈边缘与牙齿一一对应。
当然，牙龈边缘也可以有多个，牙齿区域进行初步分牙，即可得到多个初步分割区域，初步分割区域既包括牙齿区域和非牙齿区域，牙龈边缘即为非牙齿区域的非目标种子点确定的边缘，上述牙龈边缘也与牙齿区域一一对应，这样可以减少牙齿边缘和牙龈边缘中间的部分的大小，使得分割效果更好。
在一些实施例中，采用最小割算法对流网络进行分割，并将分割后的流网络中与源点连接的节点确定为目标种子点，包括：获取各相邻两个节点对应的三角面片的几何特征，几何特征为目标边的边长、相邻两个节点对应的法线与目标边的夹角、相邻两个节点之间的平均曲率以及目标边与三角面片的第三个顶点的距离中的一个或者多个，目标边为三角面片中相邻两个节点所在的边；计算各几何特征的加权平均值，得到各目标边的容量；采用最小割算法，根据各目标边的容量对流网络进行分割，并将分割后的流网络中与源点连接的节点确定为目标种子点。
由于流网络是根据各边的容量进行分割的，最小割算法应用于牙齿与牙龈的分割时，将边对应的几何特征进行加权平均，得到加权平均值作为各边的容量，即可实现通过最小割算法实现牙齿与牙龈的分割。
在一些实施例中，在得到原始种子点后，可以对原始种子点进行扩展得到目标种子点。对原始种子点进行扩展的过程可以分为一个或者多个阶段。
例如，在一个阶段中，可以按照预设曲率阈值对原始种子点进行扩展，得到目标种子点。预设曲率阈值可以理解为对原始种子点进行扩展时所使用的约束条件，避免原始种子点扩展超出限制。预设曲率阈值可以包括一个或者多个曲率值，如果预设曲率阈值包括一个曲率值，则可以按照该一个曲率值对原始种子点进行扩展，得到目标种子点，如果预设曲率阈值包括多个曲率值，则可以使用第一个曲率值对原始种子点进行扩展，然后使用第二个曲率值对第一个曲率值的扩展结果进行扩展，使用第三个曲率值对第二个曲率值的扩展结果进行扩展，直到所有的曲率值均使用一次。
在一些实施例中，按照预设曲率阈值对原始种子点进行扩展，得到目标种子点包括：按 照初始曲率阈值对原始种子点进行扩展，得到第一种子点；按照目标曲率阈值对第一种子点进行扩展，得到目标种子点，其中，目标曲率阈值根据初始曲率阈值得到。
以预设曲率阈值由初始曲率阈值和目标曲率阈值组成为例，先由初始曲率阈值对原始种子点进行扩展，得到第一种子点，然后使用目标曲率阈值对第一种子点进行扩展，得到目标种子点。初始曲率阈值和目标曲率阈值可以相同也可以不同，通过先后的两次扩展，得到目标种子点。
在一些实施例中，按照初始曲率阈值对原始种子点进行扩展，得到第一种子点包括：将与原始种子点相邻的种子点作为当前种子点；在当前种子点的曲率小于或等于初始曲率阈值的情况下，将当前种子点和原始种子点作为第一种子点。
本实施例中，在使用初始曲率阈值对原始种子点进行扩展时，可以获取与原始种子点相邻的种子点的曲率。曲率就是针对曲线上某个点的切线方向角对弧长的转动率，通过微分来定义，表明曲线偏离直线的程度。表明曲线在某一点的弯曲程度的数值。每一个种子点都对应一个曲率。通过比对曲率与初始曲率阈值的大小关系，从而确定与原始种子点相邻的种子点是否可以做为第一种子点。原始种子点无需比对曲率，即可以作为第一种子点。
此处与下文中提及的相邻可以理解为与种子点组成同一个三角面片的同一条边的两个顶点。如与原始种子点相邻，即与原始种子点组成同一个三角面片的同一条边的两个顶点。
在一些实施例中，按照目标曲率阈值对第一种子点进行扩展，得到目标种子点包括：将与第一种子点相邻的种子点作为当前种子点；在当前种子点的曲率小于或等于目标曲率阈值的情况下，将第一种子点和当前种子点作为目标种子点。
在使用初始曲率阈值对原始种子点进行扩展得到第一种子点之后，可以使用目标曲率阈值对第一种子点进行扩展，得到目标种子点。确定与第一种子点相邻的种子点的曲率，然后将该曲率与目标曲率阈值进行比对，通过比对大小关系，从而确定与第一种子点相邻的种子点是否为目标种子点。至此，则完成使用初始曲率阈值和目标曲率阈值对原始种子点进行扩展得到目标种子点。
在一些实施例中，在按照目标曲率阈值对第一种子点进行扩展，得到目标种子点之前，上述方法还包括：将初始曲率阈值与预设值的和作为目标曲率阈值，其中，预设值为正数。
本实施例中，初始曲率阈值和目标曲率阈值可以为经验值，或者初始曲率阈值为经验值，根据初始曲率阈值来确定目标曲率阈值。例如，将初始曲率阈值与预设值的和作为目标曲率阈值，即目标曲率阈值根据初始曲率阈值得到，且目标曲率阈值大于初始曲率阈值。预设值为预先设置的值，可以根据三维牙齿模型的不同进行修改。
在一些实施例中，在按照初始曲率阈值对原始种子点进行扩展得到第一种子点，或者按照目标曲率阈值对第一种子点进行扩展得到目标种子点之后，上述方法还包括：在第一种子点在二维投影图像上的对应点未落入对应的牙齿区域的情况下，将第一种子点调整为非第一种子点；或者在目标种子点在二维投影图像上的对应点未落入对应的牙齿区域的情况下，将目标种子点调整为非目标种子点。
本实施例中，在按照初始曲率阈值对原始种子点进行扩展或者按照目标曲率阈值对第一种子点进行扩展时，还需要查看扩展后的种子点是否符合要求，即扩展是否超出了范围，是否将非牙齿区域的种子点作为了第一种子点或者将非牙齿区域的种子点作为了目标种子点。如在按照初始曲率阈值对原始种子点进行扩展时，如果与原始种子点相邻的种子点的曲率小于或等于初始曲率阈值，则还要判断该种子点在二维投影图像上的对应点所在的区域。如果没有位于牙齿区域，则说明该种子点已经脱离了三维牙齿模型的牙齿的区域，因此，要将该种子点作为非第一种子点。在按照目标曲率阈值对第一种子点进行扩展时，如果与第一种子点相邻的种子点的曲率小于或等于目标曲率阈值，则还要判断该种子点在二维投影图像上的对应点所在的区域。如果没有位于牙齿区域，则说明该种子点已经脱离了三维牙齿模型的牙 齿的区域，因此，要将该种子点作为非目标种子点。
在一些实施例中，在按照曲率阈值对原始种子点进行扩展，得到目标种子点之后，在基于各牙齿的目标种子点对三维牙齿模型进行分割，得到分割牙齿之前，上述方法还包括：对多个牙齿区域进行扩大后得到的区域，得到目标区域；在目标区域中，按照初始曲率阈值和高度对目标种子点进行扩展。
在对原始种子点进行扩展得到目标种子点之后，可以按照目标种子点对三维牙齿模型的牙齿进行分割。此外，还可以在分割牙齿前，对目标种子点再次进行扩展。也就是说，除了使用初始曲率阈值和目标曲率阈值对原始种子点进行第一阶段的扩展得到目标种子点后，还可以进行第二阶段的扩展。
在第二阶段的扩展时，可以先调整二维投影图像上的牙齿区域，对牙齿区域进行扩大，得到目标区域。然后，以目标区域为限制，使用初始曲率阈值和高度对目标种子点进行第二阶段的扩展。将牙齿区域扩大为目标区域的目的在于，保证三维牙齿模型上位于牙齿上的种子点均会被标记为目标种子点，避免遗漏。
在一些实施例中，在按照初始曲率阈值和高度对目标种子点进行扩展时，可以将与目标种子点相邻的种子点作为当前种子点；在当前种子点的高度大于预设标准高度，且当前种子点的曲率小于或等于目标曲率阈值的情况下，将当前种子点作为目标种子点。
上述的高度可以为种子点在三维牙齿模型的预设方向上的数值。例如，可以将三维牙齿模型的预设方向作为Z轴，将种子点在Z轴上的坐标值作为种子点的高度。
上述预设方向可以为任意方向。在得到三维牙齿模型之后，可以将三维牙齿模型的朝向调整为预设方向，从而对获取的所有的三维牙齿模型，可以进行方向的统一。
在使用初始曲率阈值与高度对目标种子点进行扩展时，可以将与目标种子点相邻的种子点作为当前种子点，如果当前种子点的曲率和高度值都符合初始曲率阈值与高度的要求，那么可以将当前种子点作为目标种子点，从而完成对目标种子点的扩展。
在一些实施例中，在当前种子点的高度大于预设标准高度，且当前种子点的曲率小于或等于目标曲率阈值的情况下，将当前种子点作为目标种子点包括：在当前种子点的高度大于标准高度，且当前种子点的曲率小于或等于目标曲率阈值，且当前种子点在二维投影图像上的对应点位于目标区域内的情况下，将当前种子点作为目标种子点，其中，目标区域为对多个牙齿区域进行扩大后得到的区域；在当前种子点在二维投影图像上的对应点位于目标区域外的情况下，将当前种子点作为非目标种子点。
在使用初始曲率阈值与高度对目标种子点进行扩展时，要保证扩展后的目标种子点没有超出目标区域。目标区域是对牙齿区域扩大后得到的区域。扩大牙齿区域的目的在于，将牙齿区域附近的一部分非牙齿区域也囊括在目标区域内，这样在使用初始曲率阈值与高度对目标种子点进行扩展时，可以允许目标种子点扩展到牙齿区域附近的非牙齿区域。这么做可以使扩展后的目标种子点覆盖所有的牙齿区域。
在一些实施例中，在按照初始曲率阈值和高度对目标种子点进行扩展之后，上述方法还包括：将与扩展后的目标种子点相邻的种子点作为当前种子点；将当前种子点同样作为目标种子点。
在使用初始曲率阈值和高度对目标种子点进行扩展之后，可以对扩展后的目标种子点再进行一次扩展，将与目标种子点相邻的种子点同样作为目标种子点。该次扩展的目的同样在于使扩展后的目标种子点覆盖所有的牙齿区域，从而可以按照扩展后的目标种子点分割牙齿的时候，牙齿是完整的。
在一些实施例中，在预设范围内对原始种子点进行扩展之前，上述方法还包括：将多个牙齿区域进行扩大，得到目标区域；将目标区域在三维牙齿模型上的点之外的点标记为第三种子点；按照曲率阈值对第三种子点进行扩展。
本实施例中，先对牙齿区域进行扩大得到目标区域，目的在于，使目标区域包含了所有的牙齿区域，而且还包含了牙齿区域附近的非牙齿区域。因此，当将目标区域在三维牙齿模型上的点之外的点标记为第三种子点的时候，第三种子点都不是牙齿上的点。对第三种子点进行扩展，可以将非牙齿部分(牙龈部分或牙齿之间的空隙)的接近牙齿部分的点标记为第三种子点，从而可以将牙齿外的部分和牙齿部分的分界向牙齿部分靠拢，减少目标区域的范围，使目标区域虽然包含了非牙齿区域，但是包含的非牙齿区域更少。
对第三种子点进行扩展之后，三维牙齿模型上牙齿部分和牙齿外的部分之间的分割线会变“细”。
在一些实施例中，按照曲率阈值对第三种子点进行扩展包括：将与第三种子点相邻的种子点作为当前种子点；在当前种子点的曲率大于曲率阈值的情况下，将当前种子点作为第三种子点。
本实施例中，在对第三种子点进行扩展的时候，可以按照曲率和曲率阈值对第三种子点进行扩展。第三种子点的曲率可以通过计算该点的切线方向角对弧长的转动率，通过微分来计算。曲率阈值可以为预先设置的阈值，对于不同的三维牙齿模型，曲率阈值可以不同。
在一些实施例中，在将与第三种子点相邻的种子点作为当前种子点之后，上述方法还包括：在当前种子点在二维投影图像中对应的点位于目标区域内的情况下，将当前种子点作为非第三种子点。
当对第三种子点进行扩展之后，还要查看第三种子点是否扩展到了目标区域内。因为目标区域内包含了牙齿区域，如果第三种子点扩展到目标区域内，则可能扩展到了牙齿区域，因此，要将位于目标区域内的第三种子点退回，作为非第三种子点。
在一些实施例中，在按照曲率阈值对第三种子点进行扩展之后，上述方法还包括：确定第三种子点组成的第一区域；从第一区域中确定出子区域；在子区域被目标种子点包围的情况下，将子区域中的种子点确定为目标种子点。
本实施例中，在对第三种子点进行扩展之后，第三种子点组成了第一区域。第一区域可以划分为多个子区域。多个子区域的每一个子区域可以理解为多个第三种子点组成的区域。如果某一个子区域被目标种子点包围，则说明该子区域在三维牙齿模型上位于牙齿部分之内，但该子区域可能并不是牙齿，也就是牙齿上的孔洞部分。因此，要将该子区域的种子点作为目标种子点，划分到牙齿部分。
在一些实施例中，获取三维牙齿模型的二维投影图像包括：将三维牙齿模型的朝向由初始朝向调整为目标朝向；将目标朝向的三维牙齿模型投影到目标面上，得到二维投影图像。
本实施例中，在将三维牙齿模型投影为二维投影图像的时候，可以先调整三维牙齿模型的朝向，将朝向调整为目标朝向。目标朝向可以为预先设定的朝向。调整三维牙齿模型的朝向目的在于将三维牙齿模型投影为二维投影图像的时候，可以使二维投影图像上的牙齿区域更完整，遮挡更少。例如，可以将水平面作为三维坐标的X轴与Y轴，将水平面向上的方向作为目标朝向，在获取到三维牙齿模型之后，可以将三维牙齿模型的牙齿方向朝向目标朝向。
调整三维牙齿模型的朝向之后，可以将三维牙齿模型投影到目标面，目标面可以为水平面。
在一些实施例中，对二维投影图像进行识别，得到多个牙齿区域包括：将二维投影图像输入到识别模型中，由识别模型在二维投影图像上标记出多个牙齿区域。
本实施例中，可以由识别模型识别二维投影图像，从而让识别出牙齿区域。二维投影图像输入到识别模型中之后，由识别模型提取特征并识别，输出多个牙齿区域。
在一些实施例中，在三维牙齿模型中，确认与多个牙齿区域对应的原始种子点包括：将三维牙齿模型中的每一个三角面的顶点作为当前顶点；在当前顶点在二维投影图像中的对应点位于多个牙齿区域内的情况下，将当前顶点作为一个原始种子点。
本实施例中，三维牙齿模型的表面由多边形面片覆盖，多边形面片可以为三角形面片，以三角形面片为例，每一个三角形面片有3个顶点。两个相邻的三角形面片共用一条边。每一个三角形面片的顶点中，对应到二维投影图像中的点位于牙齿区域内，则可以将该顶点作为原始种子点。
在一些实施例中，在基于各牙齿的目标种子点对三维牙齿模型进行分割，得到分割牙齿之后，上述方法还包括：对分割后的牙齿进行排序。
在一些实施例中，对分割后的牙齿进行排序包括：将所有牙齿的牙中点的平均值作为起始点，将每一颗牙齿的牙中点作为终点，形成每一颗牙齿的向量；将所有牙齿中，牙中点之间的距离最大的两颗牙齿所在的直线作为目标直线；按照向量与目标直线之间的夹角的大小，对所有牙齿进行排序。
本实施例中的牙齿进行分割之后，可以进行排序。上述的牙中点和每一颗牙的中点的连线作为向量，可以与两颗距离最大的牙齿所构成的目标直线的夹角来对牙齿进行排序。也就是说，可以从牙齿的某一边的第一颗牙齿开始，向另一边的最后一颗牙齿进行排序。
在一些实施例中，在基于牙齿的目标种子点对三维牙齿模型进行分割，得到分割牙齿之后，上述方法还包括：对牙齿的边缘进行平滑，得到平滑后的边缘。
在一些实施例中，对牙齿的边缘进行平滑，得到平滑后的边缘包括：将牙齿的边缘上的顶点进行排序；将排序后的顶点中，第二个顶点作为当前顶点，对当前顶点执行如下操作，直到当前顶点不包括后顶点：将当前顶点的前顶点和后顶点的中心点作为平滑点；每得到一个平滑点，将当前顶点的后顶点作为新的当前顶点，将得到的平滑点作为新的当前顶点的前顶点；将牙齿的边缘上的第一个顶点与得到的平滑点按照先后顺序相连，得到牙齿平滑后的边缘。
本实施例中，平滑牙齿的边缘的目的在于，使牙齿和牙齿外的部分的分割线部分的棱角小一些，避免后续切割出的牙齿模型太尖锐导致安装后引起用户的牙龈不适。
在平滑时，牙齿的边缘所经过的种子点中，一系列种子点构成一条曲线。从起始的种子点开始，与第二个种子点做平均值，得到第一个中点，将第一个中点与第三个种子点求平均值，得到第二个中点，将第二个中点与第四个种子点球平均值，得到第三个中点。重复上述步骤一直到牙齿的边缘的最后一个种子点。将所有的中点按照顺序先后相连，即得到平滑后的牙齿的边缘。
以下结合一个实例对上述三维牙齿模型分割方法进行说明。
例如，如图3所示，图3为一个实例性的三维牙齿模型的二维投影图像。图3中的牙齿与牙龈部分即为口扫得到的三维牙齿模型，将三维牙齿模型投影到平面上，则到如图3所示的二维投影图像。
在获取用户的口扫数据得到用户的三维牙齿模型之后，可以对三维牙齿模型的方向进行调整。调整三维牙齿模型的方向目的在于，投影得到二维投影图像时，三维牙齿模型的牙齿部分可以尽可能多的出现在二维投影图像上。
调整方向时，可以输入三维牙齿模型，并将三维牙齿模型朝Z轴正方向摆正。摆正的方法可以采用本领域任意一种摆正方法，在此不做过多的限定。以将三维牙齿模型的牙齿向上放到桌面上为例，则桌面可以为三维直角坐标系的X轴和Y轴，桌面朝向牙齿一侧的法线即为Z轴。按照该方向摆正三维牙齿模型后，三维牙齿模型的牙齿朝上，图3可以为对三维牙齿模型的顶视图，即将三维牙齿模型投影到桌面上所得到的二维投影图像。
在对三维牙齿模型投影到平面上得到二维投影图像之后，可以使用识别模型识别二维投影图像，从而可以识别出二维投影图像上的牙齿区域。识别出牙齿区域后可以使用方框进行初步分割，得到单颗牙齿的初步分割区域。该方框可以是包含掩膜区域的最小轴向包围框，也可以是斜框。初步分割区域可以为规则的形状或者不规则的形状，例如，图4中的方框402， 即为初步分割区域，使其覆盖识别出的牙齿区域。初步分割区域包含了牙齿区域和部分非牙齿的区域。图4中的方框仅为示例。例如，如图5所示，图5为初步分割区域。
如图6所示，采用初步分割区域对三维牙齿模型进行初步分割即可得到单个牙齿三维模型，图中，601为牙齿的三维模型，602为邻牙的部分三维模型，603为牙齿附近的牙龈的三维模型。
在单个牙齿三维模型中，根据601的原始种子点确定牙齿边缘，一个牙齿对应一个牙齿边缘，根据602和603的非目标种子点确定牙龈边缘，上述牙龈边缘也与牙齿一一对应，如图7所示，牙齿边缘和牙龈边缘中间的部分不清楚是牙齿还是牙龈，采用最小割算法对种子点组成的流网络进行分割，分割后的流网络中与源点s连接的结点是牙齿的种子点，分割后的流网络中与汇点t连接的结点是牙龈的种子点，可以准确分割出牙齿与牙龈，单个牙齿三维模型分割完成后如图8所示，图中，801是牙齿，802是该颗牙齿以外的部分。
由于流网络是根据各边的容量进行分割的，最小割算法应用于牙齿与牙龈的分割时，将边对应的几何特征进行加权平均，得到加权平均值作为各边的容量，即可实现通过最小割算法实现牙齿与牙龈的分割。
二维投影图像是识别出牙齿区域之后，可以确定出三维模型中的牙齿实际区域。二维投影图像上的牙齿区域对应到三维牙齿模型上，可以对应有三维牙齿模型上的牙齿部分。对该牙齿部分上的所有点进行过滤，去除法线方向和原模型有交集的点，将过滤后剩余的点作为原始种子点。
例如，如图9所示，一个三维牙齿模型(图9中仅示出了两颗门牙)的表面(牙齿部分和非牙齿部分，整个模型的表面)是由三角面片组合而成的(也可以由四边形、五边形......等多边形面片)。三角面片并不是完全位于同一个平面上，彼此之间有面夹角。三角面片的顶点为种子点。或者，三角面片的顶点去除噪声后，剩下的点作为种子点。如果三维牙齿模型上的种子点在二维投影图像上的对应点位于如图4所示的牙齿区域，则种子点作为原始种子点。
本实施例中，可以对原始种子点进行扩展，得到目标种子点。对原始种子点进行扩展目的在于，原始种子点可能并未包含三维牙齿模型上的所有的牙齿部分，因此，通过扩展来覆盖三维牙齿模型的牙齿部分。
对原始种子点的扩展分为多个阶段。
第一个阶段，通过曲率对原始种子点进行扩展。对原始种子点进行扩展，实际上就是查看与原始种子点相邻的种子点中，是否有种子点可以与原始种子点一同作为目标种子点。
判断种子点是否相邻，可以判断种子点是否位于同一条直线上。
在通过曲率对原始种子点进行扩展时，可以进行先后两次扩展。可以先通过初始曲率阈值进行扩展，再通过目标曲率阈值进行扩展。目标曲率阈值为初始曲率阈值和一个预设值的和。因此，对于与原始种子点相邻的种子点，先判断曲率是否小于初始曲率阈值。如果曲率小于初始曲率阈值，则与原始种子点一同作为第一种子点。然后，与第一种子点相邻的种子点，如果曲率小于或等于目标曲率阈值，则与第一种子点一同作为目标种子点。从而完程对原始种子点的先后两次扩展。此时，第一阶段的扩展还未结束。因为第一阶段的扩展，原始种子点经过了先后两次扩展，因此，可能扩展的点太多，超出了牙齿部分。此时，需要判断扩展后的点在二维投影图像上的对应位置是否落入了牙齿区域。如果落入了牙齿区域，则说明在三维牙齿模型上对原始种子点的扩展并没有超出牙齿部分。如果没有落入牙齿区域，则对应的种子点不再作为目标种子点。至此，第一阶段扩展结束。
在第二阶段，可以通过初始曲率阈值和高度对第一阶段的目标种子点进行扩展。在第二阶段，可以将与第一阶段扩展后得到的目标种子点相邻的种子点作为要扩展的种子点，该部分种子点的曲率小于或等于初始曲率阈值，而且高度要大于预设标准高度。两者均符合，则 该部分种子点同样作为目标种子点。如果有一个条件不符合，则不作为目标种子点。在一个具体示例中，预设标准高度可以为当前点的高度。
此外，该部分种子点除了曲率和高度符合条件之外，该部分种子点在二维投影图像上还要处于目标区域内。目标区域是牙齿区域扩大后的区域。也就是在二维投影图像上，对牙齿区域稍作扩大，得到目标区域。然后，在第二阶段的扩展时，如果扩展后的种子点对应到二维投影图像上超出了目标区域，则超出目标区域的种子点不作为目标种子点，即回退扩展。至此，第二阶段的扩展完成。
也就是说，先按照给定曲率阈值先扩展每颗牙齿种子点，且保证不超过二维投影图像上的牙齿区域范围，这些点作为每颗牙的初始点；依据初始点按照给定(曲率阈值+0.2得到目标曲率阈值)再次进行扩展，且保证不超过牙齿区域的包围框范围；
依据给定曲率阈值和高度进行扩展，且保证不超过牙齿区域的包围框+扩大的范围(即目标区域)；每颗牙边界点向外延伸一次(延伸至牙缝和牙龈线附近)，延伸的距离可以预先设置。
初始的种子点并不是最终的牙齿所有区域，需要不断扩散，本方案即根据二维投影形成扩散范围。牙齿区域的包围框+扩大区域是指种子点能够扩散至的范围，牙齿区域的包围框+扩大区域根据二维投影图像形成扩散范围，并且通过算法参数可控制扩散范围。
牙齿区域扩展时分多次原因：部分牙模相邻牙齿之间分隔不明显。牙齿区域扩展时最后限制高度，是为了减少出现牙齿向下扩展到牙龈的可能性。牙齿边界向外延伸一次，是由于直接按照曲率阈值进行分割后，分割得到的牙齿会比原先小一大圈，是由于牙龈线附近曲率小造成，所以需要延伸。
除了第一阶段和第二阶段的扩展之外，牙齿外的部分也可以进行扩展。即，将三维牙齿模型上，对应到二维投影图像上的目标区域之外的点作为第三种子点，第三种子点即为牙齿部分之外的点，如牙龈部分的点或者牙齿之间的缝隙。该部分种子点要向牙齿部分进行扩展。扩展时，可以按照曲率进行扩展。如与第三种子点相邻的种子点的曲率大于要求的曲率阈值，则该部分种子点与第三种子点一同作为第三种子点，即牙齿外的部分的点。但需要注意的是，如果扩展后的第三种子点在二维投影图像上位于目标区域内，则该种子点可能扩展到了牙齿部分，因此，要回退，将位于目标区域内的第三种子点作为非第三种子点。
对于第一阶段和第二阶段的扩展得到目标种子点，以及扩展之后得到的第三种子点，两者的边界可以作为牙齿的边界。
如果牙齿上包含了孔洞，则目标种子点与第三种子点之间，可能包含了包围的情况，如目标种子点包围了一部分的第三种子点，该种情况则是因为牙齿上有孔洞，孔洞被识别为第三种子点。在此情况下，将被目标种子点包围的区域中的点同样作为目标种子点。
在扩展完成后，按照目标种子点和第三种子点对牙齿进行分割，从而得到单颗的牙齿。或者，还可以在分割牙齿之前，先对牙齿的边界进行平滑，平滑后的边界会更缓和。例如，如图10所示，S1-S4为牙齿边界的点，S1和S3的中点A1，A1和S4的中点A2......，将所有中点A1-An相连，则得到了平滑后的边界。
在分割牙齿得到单颗牙齿之后，可以对牙齿进行排序。三维牙齿模型为例，排序如图11所示。三维牙齿模型的牙中点的平均值作为起始点，每一刻牙的牙中点作为中点，可以连接到多个向量。以牙中点最大的两颗牙的直线为目标直线(图11中虚线)，查看向量与目标直线构成的夹角，按照夹角大小来排序。
需要说明的是，上述对错位牙齿进行自动排列的方案可以为：建立排牙坐标系；定义单颗牙齿特征点并建立牙齿局部坐标系；在此基础上，从低维角度分析牙列中各颗牙齿的位置和姿态，采用加权拟合优化的方法分别计算牙齿的坐标平移量与局部坐标轴旋转量，形成牙齿位姿与空间牙列曲线的关联约束，并结合矩形包围盒的碰撞检测方法，设计基于最速下降 法的迭代算法在空间牙列曲线约束范围内调整牙齿位姿，完成牙齿的自动排列。
根据本申请实施例的另一方面，还提供了一种牙科器械的制作方法，该牙科器械的制作方法可包括如下步骤：
步骤一，获取数字化的牙齿模型。
其中，该牙齿模型可以为前述牙龈线提取方法实施例中的目标牙齿模型。相应地，该牙齿模型的获取方式可与目标牙齿模型的获取方式一致，即，可通过口扫的方式获取，也可通过传统取印模的方式获取，此处不做限定。
步骤二，对该数字化的牙齿模型进行前处理。
在一实施例中，该前处理动作可以包括三维牙齿模型分割方法的步骤，以得到分割牙齿。该前处理操作还可包括牙龈线提取方法中的步骤，以识别出牙龈线，并可进一步包括将该牙龈线转化为切割线，以供后续步骤中切割初始器材以得到牙科器械使用，或者在一些应用场景中，还可以应用于自动牙齿分割、牙龈牙冠分离等。其中，牙龈线提取方法可与前述的牙龈线提取方法实施例中的相同，相关详细内容请参阅前述内容，此处不再赘述。
在其中一个实施例中，如图12所示，该前处理中的三维牙齿模型分割方法还可以采用以下步骤：
步骤S222，获取三维牙齿模型的牙龈线以及三维牙齿模型的牙齿区域；
步骤S224，提取牙龈线的波峰点，并对波峰点进行配对，得到波峰点组合；
步骤S226，基于波峰点组合，确定牙齿区域中牙齿间的分割路径；
步骤S228，根据分割路径对牙齿区域进行分割处理，得到目标分割牙齿。
上述三维牙齿模型分割方法的目的在于，将三维牙齿模型上的每一颗牙齿进行准确的分牙，分为单颗的牙齿。
上述三维牙齿模型可以为口扫模型，口扫模型指通过扫描用户的口腔内部而生成的包括用户的牙齿部分的三维模型。三维牙齿模型的数据在计算机或服务器上，通过显示屏可以显示三维牙齿模型的形状或样式，以方便医生观看。
上述牙龈线可以为三维牙齿模型上牙齿区域和牙龈区域的分割线，牙龈线可以通过多种方式确定，例如，可以通过神经网络模型识别的方式来确定，或者，通过识别三维牙齿模型的二维投影图像的方式确定。
上述三维牙齿模型的波峰点可以为三维牙齿模型的牙龈线上，高于相邻的点的点。例如，三维牙齿模型的牙龈线上，多个点连成的直线持续走高，在其中一个点开始下降，那么，该点就是一个波峰点。波峰点可以理解为位于两颗牙之间牙缝处的牙龈的最高点。
本实施例中的波峰点的配对，可以将每两个波峰点确定为成一个波峰点组合，一个波峰点组合中的两个波峰点用于确定两颗牙的分界。每一个波峰点组合包含两个波峰点。
本实施例中，确定出波峰点组合之后，可以根据两个波峰点，确定牙齿区域中两颗牙齿之间的分割路径。
按照分割路径可以对牙齿区域进行分割，得到单颗的牙齿。如图13所示，三维牙齿模型的牙齿与牙龈之间的牙龈线上，有波峰点和波谷点(图13中牙齿另一侧未显示的部分也有)，波峰点就是牙龈上位于两颗牙齿之间的牙龈部分的较高的点(可能为最高的点，或者不为最高的点，位于牙龈线上)。波谷点就是牙龈线上，较低的点。对波峰点进行配对，可以将每两颗牙齿之间的牙缝位置的波峰点配为一个波峰点组合，通过该波峰点组合，可以确定两颗牙齿之间的分割路径，最后通过分割路径分割牙齿，得到单颗的牙齿。分割结果可以如图14所示。
本申请通过获取到三维牙齿模型的牙龈线以及三维牙齿模型的牙齿区域，提取牙龈线的波峰点，并对波峰点进行配对，得到波峰点组合，然后根据波峰点组合确定分割路径，根据分割路径对牙齿区域进行分割处理，得到目标分割牙齿的方法，从而可以对三维牙齿模型的 每一颗牙齿进行准确的分割，实现了准确对三维牙齿模型进行分牙的效果。
在一些实施例中，获取三维牙齿模型的牙齿区域，包括：基于牙龈线对三维牙齿模型进行分割处理，得到三维牙齿模型的牙齿区域。
本实施例中，在对三维牙齿模型进行分牙之前，可以先使用牙龈线对三维牙齿模型进行分割处理，分割后，得到三维牙齿模型的牙齿区域。对三维牙齿模型进行分割，可以视为将三维牙齿模型的牙龈区域隐藏或者不做处理，后续对三维牙齿模型进行分牙时，可以避免牙龈区域影响分牙精度。
在一些实施例中，在获取三维牙齿模型的牙龈线之前，该方法还包括：确定三维牙齿模型的初始朝向；将三维牙齿模型由初始朝向调整为目标朝向。
本实施例中，在获取用户的口扫数据得到用户的三维牙齿模型之后，可以对三维牙齿模型的方向进行调整。调整三维牙齿模型的方向目的在于，可以将三维牙齿模型“摆正”，这样对于所有的三维牙齿模型，均可以调整到一个固定的朝向，使得曲率识别更加准确。例如，对于所有的三维牙齿模型，初始朝向可能是不同的。如果将桌面作为水平面，则可以将桌面的向上的法线的方向作为目标朝向。那么，三维牙齿模型调整到目标朝向后，可以为牙齿向上的状态。在得到三维牙齿模型之后，可以将三维牙齿模型的朝向调整为目标朝向，从而对获取的所有的三维牙齿模型，可以进行方向的统一。
在一些实施例中，确定三维牙齿模型的初始朝向包括：在三维牙齿模型为非闭合模型的情况下，确定三维牙齿模型的最小有向包围盒；根据最小有向包围盒确定三维牙齿模型的初始朝向。
本实施例中，三维牙齿模型可以分为两种，第一种为闭合模型，第二种为非闭合模型。闭合模型为边缘拉伸后加入一个平面作为底面，从而构成封闭模型。
对于非闭合的三维牙齿模型，可以使用最小有向包围盒包围三维牙齿模型。该最小有向包围盒的面积最大的两个面即为三维牙齿模型的牙齿朝向一侧和牙龈部分远离牙齿的一侧。可以将最小有向包围盒两个最大的面的两个法向量中，选择一个法向量的方向作为目标朝向。
在一些实施例中，根据最小有向包围盒确定三维牙齿模型的初始朝向包括：根据最小有向包围盒的面积最大的两个面，确认预设轴方向；基于预设轴方向，获取三维牙齿模型的边界边的第一平均坐标值，以及三维牙齿模型的多边形面片的顶点的第二平均坐标值；根据第一平均坐标值和第二平均坐标值，确定初始朝向。
本实施例中，上述预设轴可以为预设的方向，如三维空间的一条坐标轴或者其他方向。以任意一个方向作为预设方向，预设轴的方向与预设方向相同。
在确定预设轴之后，可以获取三维牙齿模型的边界边的第一平均坐标值，以及所有顶点的第二平均坐标值。边界边为牙齿和牙龈的分界，边界边的第一平均坐标值可以理解为牙齿和牙龈的分割点的平均值，第二平均坐标值可以理解为三维牙齿模型的重心。三维牙齿模型的重心是偏向牙齿区域一侧的。因此，包围三维牙齿模型的最小有向包围盒的两个面积最大的面中，更靠近第二平均坐标值的面是牙齿所在的一侧。因此，可以确定出三维牙齿模型的初始朝向。
在一些实施例中，确定三维牙齿模型的初始朝向包括：在三维牙齿模型为闭合模型的情况下，确定三维牙齿模型表面的多边形面片所属于的多边形面片组，其中，同一个多边形面片组中的任意两个多边形面片的平面夹角小于第一阈值；根据多边形面片组中的多边形面片的面积之和确定三维牙齿模型的初始朝向。
本实施例中，闭合模型可以为对非闭合模型进行填充得到，也可以为对口腔内部进行口扫操作时，直接口扫得到闭合三维牙齿模型。闭合的三维牙齿模型可以通过计算多边形面片所组成的多边形面片组的面积来确定初始朝向。
在一些实施例中，在三维牙齿模型为闭合模型的情况下，确定三维牙齿模型表面的多边 形面片所属于的多边形面片组包括：从未划分到多边形面片组中的所有面片中选择任意一个第一面片作为一个多边形面片组中的面片，将与第一面片的平面法向夹角小于第一阈值的多边形面片确定为与第一面片位于同一个多边形面片组中的面片，其中，在初始情况下，三维牙齿模型上的所有多边形面片均未划分到多边形面片组中；继续从未划分到多边形面片组中的所有面片中选择任意一个目标面片作为另一个多边形面片组中的面片，并将与目标面片的平面夹角小于第一阈值的多边形面片确定为与目标面片位于同一个多边形面片组中的面片，直到所有多边形面片均划分到多边形面片组中。
在一些实施例中，在三维牙齿模型为闭合模型的情况下，确定三维牙齿模型表面的多边形面片所属于的多边形面片组包括：从未划分到多边形面片组中的所有面片中选择任意一个第一面片作为一个多边形面片组中的面片，将法线与第一面片的法线的夹角小于第一阈值的多边形面片确定为与第一面片位于同一个多边形面片组中的面片，其中，在初始情况下，三维牙齿模型上的所有多边形面片均未划分到多边形面片组中；在未划分到多边形面片组中的所有面中选择任意一个目标面片作为另一个多边形面片组中的面片，并将法线与目标面片的法线的夹角小于第一阈值的多边形面片确定为与目标面片位于同一个多边形面片组中的面片，直到所有多边形面片均划分到多边形面片组中。
本实施例中，可以对多边形面片进行分组，得到不同的多边形面片组，多边形面片组中的多边形面片相互之间的面夹角小于第一阈值。
在划分多边形面片组时，可以将所有的多边形面片中的任意一个多边形面片作为一个多边形面片组中的一个多边形面片，然后，以该多边形面片为基础面片，将所有的多边形面片中，与该多边形面片的面夹角小于第一阈值或者法线与该多边形面片的法线的夹角小于第一阈值的多边形面片全部加入到该多边形面片组中。则剩余的多边形面片中，每一个多边形面片的面夹角与该基础面片的面夹角是大于或等于第一阈值的。重复该操作，直到所有的多边形面片均属于一个多边形面片组。每一个多边形面片组中的多边形面片可以视为位于同一个平面上。计算多边形面片组中的多边形面片的面积之和，面积之和最大的多边形面片组中的多边形面片视为封闭模型的牙龈一侧，则另一侧是牙齿一侧，从而确定出三维牙齿模型的初始朝向。
在一些实施例中，获取三维牙齿模型的牙龈线包括：提取三维牙齿模型的特征；对特征进行识别，得到牙龈线。
本实施例中，可以使用神经网络模型来提取三维牙齿模型的牙龈线。三维牙齿模型的数据输入神经网络模型之后，神经网络模型提取三维牙齿模型的特征，对特征进行卷积与池化，并输出识别得到的牙龈线。
在一些实施例中，提取牙龈线的波峰点包括：基于目标朝向，确定牙龈线上的每一个点的坐标值；根据坐标值，从牙龈线上确定出波峰点，其中，波峰点的坐标值大于牙龈线上波峰点的相邻点的坐标值。
本实施例中，由于三维牙齿模型的朝向调整为了目标朝向，因此，对于牙龈线上的点，如果在目标方向上的坐标值大于相邻点，那么该点即作为波峰点。
相邻的两个波峰点的距离如果太近，则可以将其中一个坐标值更大的点作为波峰点，另一个点不作为波峰点，因为波峰点是两颗牙齿中间的牙缝区域，因此，两个波峰点不能相距太近，太近则说明可能波峰点确定错了。
在一些实施例中，对波峰点进行配对，得到波峰点组合包括：按照波峰点与三维牙齿模型的位置关系，将波峰点分为第一波峰点组和第二波峰点组；将第一波峰点组中的每一个波峰点作为第一波峰点，将第一波峰点与第二波峰点组中的第二波峰点作为一组波峰点组合，其中，第二波峰点为第二波峰点组中，距离第一波峰点最近的波峰点，第一波峰点与第二波峰点的连线与三维牙齿模型的牙中线之间的角度大于第三阈值。
本实施例中，在对波峰点进行配对时，将两个距离最近而且连线经过了牙中线的波峰点配对为一对波峰点组合。牙中线即为三维牙齿模型上的牙齿的中点的连线。波峰点经过连线则说明是将牙齿两侧的波峰点配对，而不会对牙齿一侧的波峰点配对。
在一些实施例中，在将第一波峰点组中的每一个波峰点作为第一波峰点，将第一波峰点与第二波峰点组中的第二波峰点作为一组波峰点组合之后，方法还包括：将配对成功的波峰点从第一波峰点组与第二波峰点组中删除；将第二波峰点组中的每一个波峰点作为第三波峰点，将第三波峰点与第一波峰点组中的第四波峰点作为一组波峰点组合，其中，第四波峰点为第一波峰点组中，距离第三波峰点最近的波峰点，第三波峰点与第四波峰点的连线与三维牙齿模型的牙中线之间的角度大于第三阈值。
本实施例中，可以从三维牙齿模型的牙齿的一侧的波峰点开始，逐一进行配对，对于三维牙齿模型的牙齿的一侧的每一个波峰点，都从牙齿的另一侧选择距离最小的波峰点且两个波峰点连线与牙中线的夹角满足预设阈值条件，将该两根波峰点组合成一个波峰点组合，在配对完成之后，将所有配对成功的波峰点删除，然后从牙齿的另一侧开始，再次进行配对，从而实现所有的波峰点的配对。
进一步的，若仅根据距离找距离最小的波峰点，在斜牙或切牙的情况下容易配对错误，而上述配对方法通过采用夹角是否满足预设阈值条件的方式可以避免该种场景下的配对错误。
配对失败的情况可能有多种，比如两个波峰点的距离太大，或者一侧的一个波峰点和另一侧的两个波峰点均能配对等。
在一些实施例中，基于波峰点组合，确定牙齿区域中牙齿间的分割路径包括：将每一个波峰点组合确定为当前组合；将当前组合中的两个波峰点在三维牙齿模型上的最短距离确定为一个分割路径。
当确定出波峰点组合之后，可以按照波峰点组合确定分割路径。分割路径可以为牙齿上两个波峰点之间的最短路径。
在一些实施例中，基于波峰点组合，确定牙齿区域中牙齿间的分割路径包括：将每一个波峰点组合确定为当前组合；将当前组合中的两个波峰点在三维牙齿模型上的最短距离确定为第一路径；通过叠加曲率的方式对第一路径进行修正，得到目标路径；将目标路径确定为分割路径。
本实施例中，可以将分割路径作为分割牙齿的路径，或者，还可以对分割路径进行调整，当得到牙齿上两个波峰点之间的最短路径之后，还可以对该路径使用叠加曲率的方式进行修正，修正之后的路径作为分割路径。
在一些实施例中，通过叠加曲率的方式对第一路径进行修正，得到目标路径包括：将第一路径上的每一个点作为当前点，为当前点在三维牙齿模型上的预设范围内确定出当前点的替换点；其中，替换点为符合曲率阈值要求的点；将替换点的连线作为目标路径。
本实施例中，使用叠加曲率的方式对最短路径进行修正，也即依据曲率对路径上的点进行局部调整，对最短路径上的每一个点进行修正。将每一个点作为当前点，如果三维牙齿模型上，位于当前点附近的预设范围内存在符合曲率阈值要求的点，则使用符合曲率阈值要求的点替换当前点。替换后的点连接后作为分割路径。
以下结合一个实例对上述三维牙齿模型分割方法进行说明。
本申请的三维牙齿模型分割方法可以主要应用于矫正展示的场景下，在牙齿分割后，根据矫正方案调整牙位，生成矫正后的牙齿并显示。进一步的，牙齿分割技术还可以应用于正畸导板的设计、代型牙模的设计以及临时牙的设计中。
本实施例中，如果要对用户的牙齿进行检查或者进行正畸，可以先扫描用户的口腔，得到用户的牙齿的三维牙齿模型的数据。数据可以通过显示屏显示给用户和医生，从而可以由医生对用户说明牙齿的情况。
口扫得到的用户的三维牙齿模型可以为闭合的或者非闭合的模型。三维牙齿模型上包括了牙齿区域和牙龈区域，牙齿区域和牙龈区域之间的分界并未被标注。因此，如果要分牙，就要确定出牙齿和牙龈之间的分割线(牙龈线)以及牙齿之间的分割路径。
也就是说，本方案中，通过获取3D扫描仪输出的扫描数据；基于识别的模型特征，对模型自动摆正；基于自动摆好的模型，通过曲率的几何算法提取三维牙模的特征值，并对提取的特征值进行过滤、去噪音，得到牙龈线及其波峰点，并将牙齿部分和牙龈部分分割开，得到纯牙齿部分；运用牙龈线的波峰点进行配对，并通过曲率及最短路径方法确定牙齿与牙齿间的分割路径，得到分割的所有牙齿。
在一个实施例中，在一个实施例中，上述三维牙齿模型分割方法可以应用于正畸全自动生产流程中。首先获取患者的数字三维模型，对该数字三维模型执行上述的分割方法，得到单颗牙模型，然后对错位牙齿进行自动排列，并生成多套矫正牙齿模型。同时对多套矫正牙齿模型设置定位部以及摆放标识信息。对该多套矫正牙齿模型进行打印，并在打印结束后把经预热的高分子材料膜片在牙模上压膜制成壳状膜片。通过图像采集终端(CCD)对牙模(牙模表面附有膜片)底部进行拍摄，然后通过OCR识别技术，识别出标识信息，将标识信息发送给服务器，服务器根据标识信息在数据库中调出对应的打标指令或切割指令。在执行打标指令时，确定待打印的3D标签；确定所述3D标签的边界约束，并根据所述边界约束从所述底板中确定标签区域，所述标签区域用于打印所述3D标签；将所述3D标签镂穿至所述标签区域。在执行切割指令时，图像采集终端采集待切割产品底部的图像；识别出图像采集终端采集的图像中的标识码，将标识码或者标识码包含的标识信息发送至接收终端；接收终端匹配对应的切割指令，切割执行终端根据切割指令对待切割产品进行切割操作。每个待切割产品的底面设置相应的标识码，不同的产品对应不同的标识码，待切割产品置于图像采集终端的上方，通过采集待切割产品底部的图像，进而可识别图像中的标识码，标识码对应相应的切割指令，根据切割指令，切割执行终端执行相应的切割操作，去除多余的部分最终获得产品。
需要说明的是，上述对错位牙齿进行自动排列的方案可以为：建立排牙坐标系；定义单颗牙齿特征点并建立牙齿局部坐标系；在此基础上，从低维角度分析牙列中各颗牙齿的位置和姿态，采用加权拟合优化的方法分别计算牙齿的坐标平移量与局部坐标轴旋转量，形成牙齿位姿与空间牙列曲线的关联约束，并结合矩形包围盒的碰撞检测方法，设计基于最速下降法的迭代算法在空间牙列曲线约束范围内调整牙齿位姿，完成牙齿的自动排列。
图15是本实施例的一种流程图，本实施例对三维牙齿模型进行分割主要分为以下步骤：
1导入模型
导入模型即为导入三维牙齿模型。三维牙齿模型即为口扫得到的模型。将三维牙齿模型导入到系统中，系统可以通过显示屏显示三维牙齿模型的构造。
三维牙齿模型可以为通过3D扫描仪输出的扫描模型(非闭合模型)，输入的三维牙齿模型可以是任意方向。在本方案中，其他任意带有平面的牙齿模型类型并不影响本发明的实现。
三维牙齿模型为由一系列多边形面片组成的数字化三维体。
2模型自动摆正
三维牙齿模型在导入到模型中之后，可能有不同的朝向。因此，可以先将三维牙齿模型的方向调整为目标朝向，从而将三维牙齿模型摆正。
针对导入的三维牙齿模型，识别其特征，识别特征的目的是为了找到三维牙齿模型的摆正角度，由于不同的牙模应用类型不同，则其特征不一致，同时根据不同的牙模类型，通过特征的提取可对其进行归类。摆正时，需要确定三维牙齿模型的初始朝向，然后将其调整为目标朝向。
确定初始朝向的方法有多种。例如，
1)若模型为口扫模型，此类模型为非闭合模型，则首先应用定向包容盒子(Oriented Bounding Box，OBB)技术确定包容模型的包容盒子，并将包容盒子的6个面中，面积最大的两面的一条法线的方向确定为最终的牙齿朝向。这种方法是根据物体本身的几何形状来决定盒子的大小和方向，盒子无须和坐标轴垂直。这样就可以选择最合适的最紧凑的包容盒子。
得到包容盒子之后，基于预设轴方向，获取三维牙齿模型的边界边的第一平均坐标值，以及三维牙齿模型的多边形面片的顶点的第二平均坐标值。然后，将第二平均坐标值距离更近的面所在的一侧确定为牙齿朝向的一侧，二确定出三维牙齿模型的初始朝向。
根据目标的角度，运用旋转矩阵将包容盒子进行旋转，从而使得模型旋转至目标朝向。
2)若模型为带平底面的封闭模型，则依据模型的最大平底面贴底的方式进行摆正，检测牙齿最大平面的方法为：设定某个多边形面片，将设定的多边形面片和由多边形面片组成的3D牙齿模型进行叠加，设定误差阈值e，当e大于某个值时，认为设定的多边形面片与3D牙齿模型上的面片不平；反之则认为处于同一平面。当设定的多边形面片与牙齿模型的某个面片在同一平面时，将其两者叠加一起，并继续寻找下一个多边形面片并判断误差阈值。循环上述步骤直至得到牙齿模型最大的平面。确定牙齿模型最大的平面的当前法向量，并获取牙齿模型最大的平面的目标法向量。
根据叉乘运算方法，根据旋转前后的向量值求解旋转角度及旋转轴(利用叉乘运算方法，对当前法向量和目标法向量进行求解得到旋转角度及旋转轴)，叉乘运算方法是一种在向量空间中向量的二元运算，其的运算结果是一个向量而不是一个标量；由上述旋转角度及其旋转轴，可以将任意模型旋转至想要的空间位置上。
3基于特征识别提取牙龈线
识别提取三维牙齿模型的牙龈线的目的在于，根据牙龈线对三维牙齿模型进行分牙。牙龈线可以有多种方式确定。例如，可以通过神经网络模型确定，可以将三维牙齿模型输入到神经网络模型中，由神经网络模型标注出三维牙齿模型的牙龈线。或者，可以将三维牙齿模型投影到二维平面上，然后识别二维平面上的牙齿区域，在对应到三维牙齿模型上，从而标注出牙齿与牙龈的分界线，即牙龈线。
以神经网络模型识别牙龈线为例，将模型摆放至指定位置之后，按照曲率的几何计算方法，对牙齿模型提取的特征值，并进行过滤、去噪音，可得到三维牙齿模型轮廓；的曲率计划计算方法为针对牙模上某个面的切线方向角对弧长的转动率，表明面的凹凸程度，在本方案中亦称为特征。则通过曲率方法可获取牙模凹凸区域的特征值(牙模上真实牙龈线也是由凹凸体现的)；
得到初始的牙模特征值，需要将其去除噪音，的去噪音为除了牙龈线区域之外，牙齿的其他地方也有凹凸，也被认为是特征，这部分会影响后续与轮廓线的拟合，因此需要将其过滤或者去除，最终得到最优的特征值。
将特征值简化变成一条轮廓线，即变成一条初始的牙龈线，由于初始牙龈线不是平滑的曲线，会有局部的折叠、偏离的情况，采用主成分分析方法处理初始的牙龈线，得到主方向的点。通过利用数据是主方向是否一致性来确定，比如在波峰和波谷的地方是曲线弯曲的改向的位置，这些位置就可以确定主方向的点。针对此情况，本方案运用主成分分析方法获取其主要轮廓形状，确定轮廓主方向的点，规避线段不平滑区域，得到最终所需的平滑牙龈线。
4基于牙龈线分割牙龈与牙齿
在识别到牙龈线之后，可以按照牙龈线的位置分割三维牙齿模型的牙齿部分和牙龈部分。分割牙齿部分和牙龈部分的目的在于，可以使分牙更准确。
5基于牙龈线获得波峰点并进行配对
上述确定轮廓主方向的点，也即确定波峰波谷点。本方案根据牙龈线的曲率弯曲程度来 确定插入控制点(波峰波谷点都是会弯曲的区域)，并由于波谷点总是低于波峰点，因此根据上下距离(目标方向上高度的差)即可判断波峰点并将波谷点过滤，从而得到完整的波峰点。
得到波峰点之后，按距离最短且连线经过牙中线的两个波峰点进行配对，作为牙齿分割的起始和终止点。例如，如图16所示，图16中，为对牙齿两侧的波峰点进行配对。(图16中一条虚线连接的一个波峰点组合，并未示出所有虚线)。
6根据点配对按照曲率阈值与最小路径得到牙齿的分割曲线
由于牙齿与牙齿之前是接触相连的，要通过曲线去切割牙齿，形成独立的单牙，本方案首先采用最短距离的方式确定初始的切割曲线，所示最短距离为从已经配对的起始波峰点，沿着牙模三角面片的表面行走，用最短距离到达终止波峰点。由于最短距离并非最好的切割线，因此本方案叠加曲率的方式对最短距离进行修正，由于牙齿与牙齿之间的接触都是凹凸不平的，因此最短距离分割线要在曲率大的范围内，通过两种方法叠加确立最终的分割线。确定出的分割路径如图17所示。
7按照分割曲线分割牙齿
按分割曲线将所有牙齿都分割开，得到分割牙齿。
在其中一个实施例中，该牙龈线提取方法可以包括以下步骤：
步骤S202，获取目标牙齿模型，其中，目标牙齿模型包括多个多边形面片；
步骤S204，确定多个多边形面片的边的边类别，其中边类别包括牙齿边、牙龈边；
步骤S206，根据边的边类别，确定目标牙齿模型的牙龈线。
上述的目标牙齿模型可以为口扫模型，口扫模型指通过扫描用户的口腔内部而生成的包括用户的牙齿、牙龈部分的数字化三维模型；当然也可以为通过取印模的方式，先获取牙齿、牙龈部分的印模，然后扫描印模从而得到数字化的三维模型。三维牙齿模型的数据在计算机或服务器上，通过显示屏可以显示三维牙齿模型的形状或样式，以方便医生观看。
上述的多边形面片为组成目标牙齿模型的表面(牙齿部分和非牙齿部分，整个模型的表面)的面片，具体可以为三角形、四边形、五边形......等面片。每一个多边形面片位于一个平面上，不同的多边形面片可位于相同或不同的平面上。目标牙齿模型的表面由多个多边形面片组成，每个多边形面片包括多条边，相邻的两个多边形面片共用一条边。边的端点即为对应的多边形面片的顶点，一个顶点可以被多个多边形面片共用。
需要说明的是，目标牙齿模型包括牙齿部分和牙龈部分，而牙龈线则是牙齿部分与牙龈部分的边界。组成多边形面片的边则可位于牙齿部分或牙龈部分，从而可分类为牙齿边和牙龈边。基于此，本实施例中，可以根据多边形面片的边的边类别来确定目标牙齿模型的牙龈线，确定出的牙龈线用于分割目标牙齿模型的牙齿部分和牙龈部分。
本实施例中，在获取到目标牙齿模型之后，可以通过确定目标牙齿模型上的多边形面片的边的边类别的方法来确定边的类别，进而确定目标牙齿模型的牙龈线，而边作为多边形面片的组成部分能够对目标牙齿模型进行进一步的细化，从而能够提高牙齿模型的牙龈线提取的准确性。
在一些实施例中，确定多个多边形面片的边的边类别包括：获取多个多边形面片的边的目标特征；根据目标特征，确定多个多边形面片的边的边类别。
本实施例中，目标特征包括几何特征，当然，在它实施例中，目标特征也可以包括非几何特征，此处不做具体限定。本实施例中，可以通过提取多边形面片的边的目标特征，并且识别目标特征的方法，确定边类别。需要说明的是，在目标特征包括几何特征时，进行边分类时会考虑空间的拓扑结构关系，进而提高牙龈线识别的精确度。
本实施例中，可以提取目标牙齿模型上所有边的目标特征，也可以选择其中的一部分边的目标特征。通过识别提取的边的目标特征从而确定出边的边类别。
在一些实施例中，边的目标特征包括以下至少一种：共用边的第一多边形面片和第二多边形面片之间的二面角、边的第一顶点的第一曲率值、边的第二顶点的第二曲率值、第一多边形面片的远离边的顶点到边的第一距离、第二多边形面片的远离边的顶点到边的第二距离、边在第一多边形面片中的第一对角、边在第二多边形面片中的第二对角。在一个应用场景中，边的目标特征包括上述的七项。
作为另一种示例，边的目标特征可以包括以下特征的至少一种：共用边的第一多边形面片和第二多边形面片之间的二面角、边的第一顶点的第一曲率值、边的第二顶点的第二曲率值、第一顶点的第一空间坐标、第二顶点的第二空间坐标、边的长度、边与相临边的夹角、第一多边形面片的远离边的顶点到边的第一距离、第二多边形面片的远离边的顶点到边的第二距离、边在第一多边形面片中的第一对角、边在第二多边形面片中的第二对角、第一顶点的第一法线、第二顶点的第二法线、第一多边形面片的法线、第二多边形面片的法线。
上述的共用边的第一多边形面片和第二多边形面片之间的二面角为共用一条边的两个多边形面片的夹角。上述的第一顶点和第二顶点是边的两个端点，上述曲率值为顶点对应的面片在该顶点处的曲率的值。曲率就是针对曲线上某个点的切线方向角对弧长的转动率，通过微分来定义，表明曲线偏离直线的程度。表明曲线在某一点的弯曲程度的数值。曲率越大，对应点所在的曲线在该点处越弯曲。上述的空间坐标为顶点所在的三维直角坐标系中的坐标。三维直角坐标系可以为目标牙齿模型所在的坐标系，例如可以将目标牙齿模型的底板所在的平面定义为xOy面，该面的法线方向定义为z轴。坐标轴的方向可以预先确定。例如，将目标牙齿模型的牙龈面作为X轴和Y轴所在面，将牙齿方向作为Z轴方向。边的长度为边的两个端点间的距离；因为边的相邻边可能有多条，因此边与相邻边的夹角可以有多个，位于0-180度之间；在多边形为四边形或更多边形时，远离边的顶点到边的距离可以有不只一个，每一个顶点到边的距离可以不同，而在多边形面片为三角形面片时，远离边的顶点只有一个，因此该顶点到该边的距离也只有一个；相似地，边在多边形面片中的对角也可以有一个或多个，在有多个时，每一个对角的角度可以不同。例如，多边形面片分别为三角面片302、四边形面片304和五边形面片306，三个多边形面片的每一条边均对应上述至少一项或者至少两项特征。
具体地，在提取特征时，找出目标牙齿模型的边对应的多边形面片；依据多边形面片的法线通过叉乘可以得到两个多边形面片的法线之间的夹角，而两个多边形面片之间的二面角与该法线之间的夹角互补，通过180°减去该法线之间的夹角可得到该两个多边形面片的二面角；按照曲率几何计算方法或者其它方法可以得到两个顶点的曲率值；按照点到直线距离可以得到多边形面片的远离边的顶点到边的距离；依据两条边的朝向得到两个向量，依据两个向量通过点乘可以得到多边形面片的顶点处两条边的夹角，即得到边在各多边形面片中的对角。需要说明的是，各顶点对应的空间坐标以及点、边、面的邻近关系均是已知的，可基于此得出各目标特征。
在一些实施例中，根据目标特征，确定边的边类别包括：对边的初始维度的目标特征进行升维，得到第一维度的第一特征；对第一特征进行降维，得到目标维度的第二特征；根据第二特征的值的大小确定边为牙齿边或牙龈边。也就是说，本实施例中，目标维度可以是二维，分别与牙齿边和牙龈边对应。当然，在其它实施例中也可以是多维，只要能够确定对应边的边类别即可，此处不做限定。
本实施例中，在提取边的目标特征后，目标特征的初始维度可根据所提取的目标特征的种类、数量等确定。先对目标特征进行升维，升到第一维度，得到第一特征，再对第一特征进行降维，降低到目标维度，但是，由于先升维再降维，因此，降维后的第一特征已经不是目标特征，而是目标维度的第二特征，据此可得出对应边的边类别。
以初始维度为七维为例，将对应边的七维目标特征输入边分类网络中以对该对应边进行分类。具体地，首先输入维度为7个维度，分别对应7个几何特征，经过一次卷积操作，目 标特征的维度上升到N维，其次经过k个相同卷积操作，保持目标特征的维度不变，再经过池化操作减少边的数量。然后重复以上过程，即一个升维卷积+k个同维卷积+池化，使得目标特征的维度依次变为2N、4N、8N，需要说明的是，在最后一个升维的循环中，经过k个同维卷积后，进行上池化操作，以逐渐恢复边的数量，然后进行降维卷积，使得特征维度降至4N维，再经过k个相同卷积操作，保持特征维度不变，然后重复以上过程，即上池化+一个降维卷积+k个升维卷积，使得特征维度依次变为4N、2N、N，需要说明的是，在最后一个降维循环后，进一步直接将特征维度降为2维即可。最终的两个维度的值即为边是牙齿边和牙龈边类别的概率值，可选取概率值大的类别作为边的类别，至此，模型的每条边都被分为牙齿边或者牙龈边。
需要说明的是，以上仅为举例不作为限定。例如，初始维度不限定为7维，升维和降维过程中的循环次数也不做限定，N为不小于7的整数，如7、8、9、10等，在具体实施方案时，可根据实际情况进行调整。
需要进一步说明的是，该步骤中卷积和池化不同于图像邻域的卷积池化操作，其操作是基于目标牙齿模型中的边进行的。
卷积操作中为排除边顺序不同(如：(a，b，c，d)和(c，d，a，b))带来的影响，对原始边特征按如下(1)式进行处理，其中f(a)和f′(a)分别表示边a的原始特征和处理后的特征，处理后边e的卷积操作定义见(2)式，其中{wk|k＝0，1，2，3，4}表示需要训练的权重参数。池化操作将5条边(e，a，b，c，d)转变为2条边(h，i)，其定义见(3)式，上池化操作和池化操作相反，将2条边(h，i)转变成(e′，a′，b′，c′，d′)，其定义见(4)式。

C(e)＝w0f′(e)+w1f′(a)+w2f′(b)+w3f′(c)+w4f′(d)   (2)

另外，在进行边分类之前，需先对分类网络进行训练，以得到网络权重值以及一些超参数的值：如k和N等，具体是将多组几何特征数据及对应的边的分类的类别值(如牙龈对应0，牙齿对应1)输入到边分类网络中进行训练，例如输入500组、1000组或2000组等已分类好的牙齿模型数据，每组牙齿模型数据包含5000、10000或20000条边对应的几何特征数据和分类数据，使得该分类网络能够依据边的几何特征数据对该边进行分类。需要说明的是，本实施例中结合深度学习技术，基于边分类的方法能自动识别出牙龈线数据，能够极大提升识别准确度和效率，提高整个隐形正畸等牙齿诊疗流程的效率，增强用户在诊疗中的体验。
在一些实施例中，根据边的边类别，确定目标牙齿模型的牙龈线包括：根据边类别，确定多边形面片的面片类别，其中，面片类别包括牙齿面片、牙龈面片；基于面片类别不同且相邻的多边形面片的共用边提取牙龈线。
本实施例中，在确定了边的类别之后，可以根据边的类别确定多边形面片的面片类别。可以理解地，牙齿面片是位于牙齿模型的牙齿部分的面片，牙龈面片是位于牙龈部分的面片。根据边的边类别确定面片类别，可以根据多边形面片的所有边确定多边型面片的面片类别，或者根据多边形面片的至少一条边确定多边形面片的面片类别。
在一些实施例中，根据边类别，确定多边形面片的面片类别包括：在多边形面片中，牙龈边的数量大于牙齿边的数量时，确定多边形面片为牙龈面片；在多边形面片中，牙龈边的数量小于牙齿边的数量时，确定多边形面片为牙齿面片。
本实施例中，根据多边形面片的所有的边的边类别确定多边形面片的面片类别。根据多边形面片中，牙齿边和牙龈边的数量的多少，从而确定出多边形面片是牙齿边或者是牙龈边。
遍历目标牙齿模型所有边，如果边对应的两个面片类别不同，则边标记为共用边。
在一些实施例中，基于面片类别不同且相邻的多边形面片的共用边提取牙龈线包括：基于共用边的顶点确定第一特征点；根据曲率关系，从第一特征点中确定出目标特征点；根据目标特征点确定牙龈线。
本实施例中，由于可以根据边的边类别确定多边形面片的面片类别，在确定面片类别之后，类别不同的多边形面片的共用边则为牙齿面片和牙龈面片的分界。因此，可以根据牙齿面片和牙龈面片的所有共用边来确定牙龈线。
本实施例中，可以从共用边的顶点中确定出第一特征点，然后从第一特征点中确定出目标特征点，根据目标特征点确定牙龈线。目标特征点可以根据曲率计算得到。
本实施例中，可以一条共用边的顶点为起点，查找相连的共用边，将各相连的共用边的顶点按逆时针或者顺时针顺序记录下来作为特征点。其中，第一特征点可以是该记录下来的特征点的全部或者部分，具体地，第一特征点位于初始牙齿模型上，如此，所提取的牙龈线的准确度更高。目标特征点则为第一特征点中的部分具有特定特征的特征点。
在一些实施例中，基于共用边的顶点确定第一特征点包括：在目标牙齿模型为对初始牙齿模型进行边收缩处理得到的模型的情况下，将共用边上的顶点在初始牙齿模型上的最近的点确定为第一特征点；在目标牙齿模型为初始牙齿模型的情况下，将共用边上的顶点确定为第一特征点。
上述初始牙齿模型为最初获取到的牙齿模型。可以对初始牙齿模型进行边收缩处理得到目标牙齿模型，或者直接将初始牙齿模型作为目标牙齿模型。是否对初始牙齿模型进行边收缩处理，影响着是否直接将共用边的顶点作为第一特征点。本实施例中，第一特征点是共用边的顶点，或者是共用边在初始牙齿模型上最近的点。
也就是说，本实施例中，在以一条共用边的顶点为起点，查找相连的共用边，将顶点按逆时针或者顺时针顺序记录下来作为特征点之后，可以将记录下的特征点投影回原模型，找到原模型上离该记录下的特征点最近的点作为新的特征点。
具体可以初始牙齿模型的顶点建立KDTree，KDTree是一种对k维空间中的实例点进行存储以便对其进行快速检索的树形数据结构。本实施例中，可遍历所有记录的特征点，在KDTree中查找出与其最近的点作为新的特征点。新的特征点即为第一特征点。
在一些实施例中，根据曲率关系从第一特征点中确定出目标特征点包括：将第一特征点中，曲率大于邻近的第一特征点的曲率的第一特征点确定为目标特征点。
当根据初始牙齿模型是否经过边收缩处理从而确定出第一特征点之后，可以根据第一特征点的曲率从第一特征点中确定出目标特征点。每一个第一特征点都对应一个曲率。本实施例中，将第一特征点中，曲率大于邻近的第一特征点的曲率的第一特征点作为目标特征点，也就是将曲线中，所在的位置更弯曲的点确定为目标特征点，如此更能体现曲线的实际走势。
记录的特征点投影回原模型后，可以利用曲率的几何算法估计出离散点的曲率，保留曲率大的特征点作为目标特征点；保留曲率大的特征点的原因是，曲率大的点往往是转折点， 转折点的特征性较强，具有代表性。
在一些实施例中，根据目标特征点确定牙龈线包括：对目标特征点执行插值操作，得到插值后的目标特征点；将插值后的目标特征点按顺序连接为牙龈线。
本实施例中，对目标特征点执行插值操作，可以补充由于从第一特征点中确定目标特征点所造成的特征点的损失，以及通过插值操作可以将目标特征点组合为闭合的曲线。闭合的曲线即为牙龈线。
对目标特征点进行插值时，可以对相邻目标特征点进行B样条曲线插值；将所有点按顺序连接，形成的曲线即为牙龈线。需要说明的是，在对目标特征点进行插值时，不限定于B样条曲线，也可采用其它样条曲线，此处不做具体限定。
在一些实施例中，获取目标牙齿模型包括：获取初始牙齿模型，其中，初始牙齿模型包括多个多边形面片，初始牙齿模型中的边的数量大于目标牙齿模型中的边的数量；按照预设方式将初始牙齿模型中的边的数量减少到预设数量，得到目标牙齿模型。
本实施例中，预设方式可以为边收缩操作或边塌陷操作。对初始牙齿模型进行边收缩操作，目的在于减少初始牙齿模型中的边的数量，从而得到目标牙齿模型，以对牙齿模型进行简化，提高牙龈线的识别速率。
减少初始牙齿模型的边的数量，可以有多种方法。可以一条边或者一批边的减少，如每次减少一条，直到减少到预设数量或者减少预设次数。或者，每次减少一批次，直到减少到预设数量或者减少预设次数。
以边塌陷操作为例，导入的初始牙齿模型，可为任意方向的牙齿模型，初始牙齿模型为由一系列顶点及多边形面片组成的数字化三维体。在导入初始牙齿模型后，将该牙齿模型按照边塌陷的方式进行下采样，下采样后牙齿模型的多边形面片更加稀疏，多边形面片数量减少，边的数量也随之减少，直到具有指定边数量才停止下采样。
导入的初始牙齿模型往往具有很多条边且边的数量不一致，而在后续步骤中通过分类网络进行分类时，需要的边的数量往往是特定的，因此需对导入的初始牙齿模型进行下采样，删除一些边，保留指定数量的边。边数量的确定可考虑分类速度(边数越少速度越快)和最终得出的牙龈线的准确率(边数越多准确率越高，边越多原模型信息损失越小)。其主要的流程如下：
(1)对初始牙齿模型的每条边的顶点，记录其到相邻面片的距离之和作为误差值，初始值为0；其中，相邻面片是指该顶点所在的面片；初始误差值为0是指，在尚未进行边收缩时，该顶点到其相邻面片的距离和为0。
(2)计算删掉边后顶点对应的误差值(这里指顶点到原始未删除边时的相邻面片的距离之和)，按照误差值从小到大排列，首先删除误差小的边；其中，一条边收缩成为一个顶点时，删掉相邻的2个面片，并删掉1个顶点及3条边；一条边缩成点的位置有多种情况，例如可以由边的两个顶点移动至该边上的任一点上以进行删除，点的具体位置是在该步骤中按照误差值最小进行选择的。
(3)重复以上步骤，不断计算误差值以及删除对应的边，直到边数达到指定边数量。至此，完成对初始牙齿模型的边的数量的调整。
在一些实施例中，按照预设方式将初始牙齿模型中的边的数量减少到预设数量，得到目标牙齿模型包括：对于待处理牙齿模型的每条边，计算按照多种收缩方式中的每种对每条边进行边收缩处理后，每条边对应的顶点到待处理牙齿模型中与每条边对应的顶点的相邻的多边形面片的距离之和；其中，在首次处理中，待处理牙齿模型为初始牙齿模型；确定计算得到的多个距离之和之中的最小值对应的边及收缩方式；对最小值对应的边按照最小值对应的收缩方式进行边收缩处理，以得到目标牙齿模型。
在一些实施例中，对最小值对应的边按照最小值对应的收缩方式进行边收缩处理，以得 到目标牙齿模型包括：在进行边收缩处理后，若得到的牙齿模型中的边数量不大于预设数量，则得到的牙齿模型为目标牙齿模型；若边数量大于预设数量，则将得到的牙齿模型确定为待处理牙齿模型，以继续进行边收缩处理。
本实施例中，在对初始牙齿模型进行边收缩处理时，还要确定要收缩哪一条边。该过程可以为一个循环的过程。即每次收缩一条边并循环或者每次收缩一批边并循环，直到收缩后的待处理牙齿模型符合条件。因此，在每一次的收缩时，要确定到底收缩哪一条边或者哪一批边。
本实施例中，可以将初始模型作为待处理模型，为待处理模型确定所要收缩的边。确定方法可以为先模拟或者先计算，计算假如收缩某一条边，那么收缩后的模型中每条边对应的顶点到待处理牙齿模型中与该条边对应的顶点的相邻的多边形面片的距离之和，如果该和最小，那么就按照该策略，收缩该条边。每收缩一次，重复进行计算，并重复进行边的收缩，直到待处理牙齿模型的边的数量符合预设数量的要求。至此，边收缩完毕，得到目标牙齿模型。
步骤三，基于前处理后的数字化的牙齿模型进行成型处理，得到成型牙齿模型，即实体牙模。
在一实施例中，基于目标分割牙齿打印得到成型牙齿模型，其中，成型牙齿模型用于得到牙科器械。也即将目标分割牙齿排列在一起得到数字数字化的牙齿模型，并基于成型处理(也称打印)得到实体牙模。其中，该3D打印可采用光固化3D打印，如SLA、DLP、LCD，或者其它3D打印方式，如3DP、MJF、FDM、Polyjet等。
在一些实施例中，基于前述的前处理操作，所得到的实体牙模可具有镂空的底板、固定附件、标识信息等。
进一步地，3D打印后还包括后处理步骤，后处理步骤可根据所采用的3D打印的方式进行选择。举例而言，当采用光固化3D打印技术时，后处理步骤可选择后固化、清洗等步骤。
当然，在其它实施例中，成型的方式并不限定于3D打印，还可采用其它成型方式，如注塑等，此处不做限定。
步骤四、对实体牙模进行压膜处理。
得到实体牙模后，可将经预热的高分子材料膜片在该实体牙模上进行压膜处理，得到覆盖该实体牙模的壳状膜片，也即尚未进行切割的初始器材。具体地，该壳状膜片至少覆盖实体牙模的牙齿部分。
在一实施例中，可通过图像采集终端，如CCD图像传感器等，采集实体牙模的图像，进而通过OCR等识别技术识别出该标识信息，然后基于标识信息在数据库中调出对应的压膜指令，并基于该牙模指令进行压膜处理，其中，压膜指令可包括压力参数、膜片预热时间、压膜温度等。
步骤五、对壳状膜片进行打标操作。
该步骤中，主要是基于前处理操作中所添加的标识信息对壳状膜片进行打标操作，以在壳状膜片上形成标识，得到具有标识的初始器材。
具体地，与前述步骤类似，可通过图像采集终端，如CCD图像传感器等，采集实体牙模的图像，然后通过OCR等识别技术识别出该标识信息，然后基于该标识信息调出对应的打标指令，以基于打标指令中所包含的参数信息对壳状膜片进行打标。
需要说明的是，在牙科器械的制作方法中，不限定该步骤的执行，即，可根据实际需求选择是否需要打标操作。
步骤六、对初始器材进行切割，以得到牙科器械。
该步骤中主要是基于由所提取的牙龈线转换得到的切割线对初始器材进行切割，从而将初始器材中无用部分切除，仅保留有用部分，从而得到牙科器械。具体地，该牙科器械可以 为隐形矫治器。
需要说明的是，该步骤中对于切割线的获取可与前述牙科器械的制作方式实施例中的相同，相关详细内容请参阅前述实施方式，此处不再赘述。另外，本实施例中对于步骤五和步骤六的先后顺序不做限定。
在一实施例中，与前述步骤类似，可通过图像采集终端，如CCD图像传感器等，采集实体牙模的图像，然后通过OCR等识别技术识别出该标识信息，然后基于该标识信息调出对应的切割指令，以基于切割指令中所包含的参数信息对初始器材进行切割。
根据本申请实施例的又一方面，如图18所示，提供了一种三维牙齿模型分割装置，包括：存储器1801和处理器1803，存储器存储有计算机程序，计算机程序被处理器运行时执行上述的三维牙齿模型分割方法。
此处需要说明的是，上述模块与对应的步骤所实现的示例和应用场景相同，但不限于上述实施例所公开的内容。需要说明的是，上述模块作为装置的一部分可以运行在如图1所示的硬件环境中，可以通过软件实现，也可以通过硬件实现。上述装置还可以包括其他部分，如通信接口1805以及连接线1807等。
根据本申请实施例的又一方面还提供了一种计算机程序产品或计算机程序，该计算机程序产品或计算机程序包括计算机指令，该计算机指令存储在计算机可读存储介质中。计算机设备的处理器从计算机可读存储介质读取该计算机指令，处理器执行该计算机指令，使得该计算机设备执行上述任一实施例的步骤。
本申请实施例在具体实现时，可以参阅上述各个实施例，具有相应的技术效果。
可以理解的是，本文描述的这些实施例可以用硬件、软件、固件、中间件、微码或其组合来实现。对于硬件实现，处理单元可以实现在一个或多个专用集成电路(Application Specific Integrated Circuits，ASIC)、数字信号处理器(Digital Signal Processing，DSP)、数字信号处理设备(DSP Device，DSPD)、可编程逻辑设备(Programmable Logic Device，PLD)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)、通用处理器、控制器、微控制器、微处理器、用于执行本申请所述功能的其它电子单元或其组合中。
对于软件实现，可通过执行本文所述功能的单元来实现本文所述的技术。软件代码可存储在存储器中并通过处理器执行。存储器可以在处理器中或在处理器外部实现。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个模块或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实施例的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例上述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。需要说明的是，在本文中，诸如“第一”和“第二”等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个......”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。以上所述仅是本申请的具体实施方式，使本领域技术人员能够理解或实现本申请。对这些实施例的多种修改对本领域的技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本申请的精神或范围的情况下，在其它实施例中实现。因此，本申请将不会被限制于本文所示的这些实施例，而是要符合与本文所申请的原理和新颖特点相一致的最宽的范围。

Claims (34)

  1. 一种三维牙齿模型分割方法，其中，包括：
    获取三维牙齿模型的二维投影图像，并对所述二维投影图像进行识别，得到多个牙齿区域；
    在所述三维牙齿模型中，确认与所述多个牙齿区域对应的原始种子点；
    在预设范围内对所述原始种子点进行扩展，得到所述三维牙齿模型中牙齿的目标种子点；
    基于各所述牙齿的目标种子点对所述三维牙齿模型进行分割，得到分割牙齿。
  2. 根据权利要求1所述的方法，其中，在预设范围内对所述原始种子点进行扩展，得到所述三维牙齿模型中牙齿的目标种子点包括：
    根据所述原始种子点确定牙齿边缘，所述牙齿边缘为覆盖所述牙齿区域的所有的所述原始种子点的所述三维牙齿模型的表面的边缘；
    在所述三维牙齿模型中，确认与多个所述牙齿区域以外的区域对应的非目标种子点，并根据所述非目标种子点确定牙龈边缘，所述牙龈边缘为覆盖所有的所述非目标种子点的所述三维牙齿模型的表面的边缘；
    在所述三维牙齿模型的表面上，将远离所述牙龈边缘的所述牙齿边缘的一侧的任意一点确定为源点，将远离所述牙齿边缘的所述牙龈边缘的一侧的任意一点确定为汇点；
    根据各所述源点和所述汇点构建多个流网络，使得所述源点和所述汇点之间的种子点均为所述流网络的节点，任意两个相邻的所述节点连接，所述种子点为三角面片的顶点；
    采用最小割算法对各所述流网络进行分割，并将分割后的各所述流网络中与所述源点连接的所述节点确定为对应的各所述牙齿的所述目标种子点。
  3. 根据权利要求2所述的方法，其中，在所述三维牙齿模型中，确认与多个所述牙齿区域以外的区域对应的非目标种子点，并根据所述非目标种子点确定牙龈边缘，包括：
    根据多个所述牙齿区域对所述三维牙齿模型进行初步分牙，得到多个初步分割区域，所述初步分割区域一一对应地覆盖所述牙齿区域；
    在各所述初步分割区域中，确认与各所述牙齿区域以外的区域对应的各组非目标种子点；
    根据各组所述非目标种子点确定对应的各牙龈边缘，所述牙龈边缘为覆盖一组所述非目标种子点的所述三维牙齿模型的表面的边缘，所述牙龈边缘与所述牙齿一一对应。
  4. 根据权利要求2所述的方法，其中，采用最小割算法对所述流网络进行分割，并将分割后的所述流网络中与所述源点连接的所述节点确定为所述目标种子点，包括：
    获取各相邻两个所述节点对应的所述三角面片的几何特征，所述几何特征为目标边的边长、相邻两个所述节点对应的法线与所述目标边的夹角、相邻两个所述节点之间的平均曲率以及所述目标边与所述三角面片的第三个顶点的距离中的一个或者多个，所述目标边为所述三角面片中所述相邻两个所述节点所在的边；
    计算各所述几何特征的加权平均值，得到各所述目标边的容量；
    采用所述最小割算法，根据各所述目标边的容量对所述流网络进行分割，并将分割后的所述流网络中与所述源点连接的所述节点确定为所述目标种子点。
  5. 根据权利要求1所述的方法，其中，所述在预设范围内对所述原始种子点进行扩展，得到所述三维牙齿模型中牙齿的目标种子点包括：
    按照预设曲率阈值对所述原始种子点进行扩展，得到所述目标种子点。
  6. 根据权利要求5所述的方法，其中，所述按照预设曲率阈值对所述原始种子点进行扩展，得到所述目标种子点包括：
    按照初始曲率阈值对所述原始种子点进行扩展，得到第一种子点；
    按照目标曲率阈值对所述第一种子点进行扩展，得到所述目标种子点，其中，所述目标 曲率阈值根据所述初始曲率阈值得到。
  7. 根据权利要求6所述的方法，其中，所述按照初始曲率阈值对所述原始种子点进行扩展，得到第一种子点包括：
    将与所述原始种子点相邻的种子点作为当前种子点；
    在所述当前种子点的曲率小于或等于所述初始曲率阈值的情况下，将所述当前种子点和所述原始种子点作为所述第一种子点。
  8. 根据权利要求6所述的方法，其中，所述按照目标曲率阈值对所述第一种子点进行扩展，得到所述目标种子点包括：
    将与所述第一种子点相邻的种子点作为当前种子点；
    在所述当前种子点的曲率小于或等于所述目标曲率阈值的情况下，将所述第一种子点和所述当前种子点作为所述目标种子点。
  9. 根据权利要求6所述的方法，其中，在按照目标曲率阈值对所述第一种子点进行扩展，得到所述目标种子点之前，所述方法还包括：
    将所述初始曲率阈值与预设值的和作为所述目标曲率阈值，其中，所述预设值为正数。
  10. 根据权利要求6所述的方法，其中，在按照初始曲率阈值对所述原始种子点进行扩展得到第一种子点，或者按照目标曲率阈值对所述第一种子点进行扩展得到所述目标种子点之后，所述方法还包括：
    在所述第一种子点在所述二维投影图像上的对应点未落入对应的牙齿区域的情况下，将所述第一种子点调整为非第一种子点；或者
    在所述目标种子点在所述二维投影图像上的对应点未落入对应的牙齿区域的情况下，将所述目标种子点调整为非目标种子点。
  11. 根据权利要求5至10任意一项所述的方法，其中，在按照曲率阈值对所述原始种子点进行扩展，得到所述目标种子点之后，在基于各所述牙齿的目标种子点对所述三维牙齿模型进行分割，得到分割牙齿之前，所述方法还包括：
    对所述多个牙齿区域进行扩大后得到的区域，得到目标区域；
    在所述目标区域中，按照初始曲率阈值和高度对所述目标种子点进行扩展。
  12. 根据权利要求11所述的方法，其中，所述按照初始曲率阈值和高度对所述目标种子点进行扩展包括：
    将与所述目标种子点相邻的种子点作为当前种子点；
    在所述当前种子点的高度大于预设标准高度，且所述当前种子点的曲率小于或等于目标曲率阈值的情况下，将所述当前种子点作为所述目标种子点。
  13. 根据权利要求12所述的方法，其中，所述在所述当前种子点的高度大于预设标准高度，且所述当前种子点的曲率小于或等于所述目标曲率阈值的情况下，将所述当前种子点作为所述目标种子点包括：
    在所述当前种子点的高度大于标准高度，且所述当前种子点的曲率小于或等于所述目标曲率阈值，且所述当前种子点在所述二维投影图像上的对应点位于目标区域内的情况下，将所述当前种子点作为所述目标种子点，其中，所述目标区域为对所述多个牙齿区域进行扩大后得到的区域；
    在所述当前种子点在所述二维投影图像上的对应点位于所述目标区域外的情况下，将所述当前种子点作为非目标种子点。
  14. 根据权利要求11所述的方法，其中，在按照初始曲率阈值和高度对所述目标种子点进行扩展之后，所述方法还包括：
    将与扩展后的目标种子点相邻的种子点作为当前种子点；
    将所述当前种子点同样作为所述目标种子点。
  15. 根据权利要求1所述的方法，其中，在预设范围内对所述原始种子点进行扩展之前，所述方法还包括：
    将所述多个牙齿区域进行扩大，得到目标区域；
    将所述目标区域在所述三维牙齿模型上的点之外的点标记为第三种子点；
    按照曲率阈值对所述第三种子点进行扩展。
  16. 根据权利要求15所述的方法，其中，所述按照曲率阈值对所述第三种子点进行扩展包括：
    将与所述第三种子点相邻的种子点作为当前种子点；
    在所述当前种子点的曲率大于所述曲率阈值的情况下，将所述当前种子点作为所述第三种子点。
  17. 根据权利要求16所述的方法，其中，在将与所述第三种子点相邻的种子点作为当前种子点之后，所述方法还包括：
    在所述当前种子点在所述二维投影图像中对应的点位于所述目标区域内的情况下，将所述当前种子点作为非第三种子点。
  18. 根据权利要求15所述的方法，其中，在按照曲率阈值对所述第三种子点进行扩展之后，所述方法还包括：
    确定所述第三种子点组成的第一区域；
    从所述第一区域中确定出子区域；
    在所述子区域被所述目标种子点包围的情况下，将所述子区域中的种子点确定为所述目标种子点。
  19. 根据权利要求1所述的方法，其中，所述获取三维牙齿模型的二维投影图像包括：
    将所述三维牙齿模型的朝向由初始朝向调整为目标朝向；
    将目标朝向的所述三维牙齿模型投影到目标面上，得到所述二维投影图像。
  20. 根据权利要求1所述的方法，其中，所述对所述二维投影图像进行识别，得到多个牙齿区域包括：
    将所述二维投影图像输入到识别模型中，由所述识别模型在所述二维投影图像上标记出所述多个牙齿区域。
  21. 根据权利要求1所述的方法，其中，所述在所述三维牙齿模型中，确认与所述多个牙齿区域对应的原始种子点包括：
    将所述三维牙齿模型中的每一个三角面的顶点作为当前顶点；
    在所述当前顶点在所述二维投影图像中的对应点位于所述多个牙齿区域内的情况下，将所述当前顶点作为一个所述原始种子点。
  22. 根据权利要求1所述的方法，其中，在基于各所述牙齿的目标种子点对所述三维牙齿模型进行分割，得到分割牙齿之后，所述方法还包括：
    对分割后的牙齿进行排序。
  23. 根据权利要求22所述的方法，其中，所述对分割后的牙齿进行排序包括：
    将所有牙齿的牙中点的平均值作为起始点，将每一颗牙齿的所述牙中点作为终点，形成每一颗牙齿的向量；
    将所有牙齿中，牙中点之间的距离最大的两颗牙齿所在的直线作为目标直线；
    按照所述向量与所述目标直线之间的夹角的大小，对所述所有牙齿进行排序。
  24. 根据权利要求1所述的方法，其中，在基于所述牙齿的目标种子点对所述三维牙齿模型进行分割，得到分割牙齿之后，所述方法还包括：
    对所述牙齿的边缘进行平滑，得到平滑后的边缘。
  25. 根据权利要求24所述的方法，其中，所述对所述牙齿的边缘进行平滑，得到平滑后的 边缘包括：
    将所述牙齿的边缘上的顶点进行排序；
    将排序后的顶点中，第二个顶点作为当前顶点，对所述当前顶点执行如下操作，直到所述当前顶点不包括后顶点：将所述当前顶点的前顶点和后顶点的中心点作为平滑点；每得到一个所述平滑点，将所述当前顶点的后顶点作为新的当前顶点，将得到的所述平滑点作为新的所述当前顶点的所述前顶点；
    将所述牙齿的边缘上的第一个顶点与得到的所述平滑点按照先后顺序相连，得到所述牙齿平滑后的边缘。
  26. 一种牙科器械的制作方法，其中，包括：
    如权利要求1-22任一项所述的方法得到分割牙齿；
    基于所述分割牙齿打印得到成型牙齿模型；其中，所述成型牙齿模型用于得到牙科器械。
  27. 根据权利要求26所述的牙科器械的制作方法，其中，还包括：
    获取三维牙齿模型的牙龈线，将所述牙龈线转换为切割线；
    基于所述切割线，切割初始器材，得到所述牙科器械；其中，所述初始器材与所述成型牙齿模型具有关联关系。
  28. 根据权利要求27所述的牙科器械的制作方法，其中，还包括：
    对所述成型牙齿模型进行压膜处理，得到覆盖所述成型牙齿模型的壳状膜片；
    对所述壳状膜片进行打标操作，以得到具有标识的所述初始器材。
  29. 根据权利要求27所述的牙科器械的制作方法，其中，获取三维牙齿模型的牙龈线的步骤，包括：
    获取目标牙齿模型，其中，所述目标牙齿模型包括多个多边形面片；
    确定所述多个多边形面片的边的边类别，其中，所述边类别包括牙齿边、牙龈边；
    根据所述边的所述边类别，确定所述目标牙齿模型的牙龈线。
  30. 根据权利要求29所述的牙科器械的制作方法，其中，所述确定所述多个多边形面片的边的边类别包括：
    获取所述多个多边形面片的边的目标特征，其中，所述目标特征包括几何特征；
    根据所述目标特征，确定所述边的所述边类别。
  31. 根据权利要求30所述的牙科器械的制作方法，其中，所述根据所述目标特征，确定所述边的所述边类别包括：
    对所述边的初始维度的所述目标特征进行升维，得到第一维度的第一特征；
    对所述第一特征进行降维，得到目标维度的第二特征；
    根据所述第二特征的值的大小确定所述边为牙齿边或牙龈边。
  32. 根据权利要求29所述的牙科器械的制作方法，其中，所述根据所述边的所述边类别，确定所述目标牙齿模型的牙龈线包括：
    根据所述边类别，确定所述多边形面片的面片类别，其中，所述面片类别包括牙齿面片、牙龈面片；
    基于所述面片类别不同且相邻的多边形面片的共用边提取所述牙龈线。
  33. 一种三维牙齿模型分割装置，其中，包括：
    存储器和处理器，所述存储器存储有计算机程序，所述计算机程序被处理器运行时执行所述权利要求1至32任一项中所述的方法。
  34. 一种计算机可读的存储介质，所述计算机可读的存储介质存储有计算机程序，其中，所述计算机程序被处理器运行时执行所述权利要求1-32任一项中所述的方法。
PCT/CN2023/135600 2022-12-02 2023-11-30 三维牙齿模型分割方法和装置 WO2024114758A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211543154.6A CN116168185B (zh) 2022-12-02 三维牙齿模型分割方法和装置
CN202211543154.6 2022-12-02

Publications (1)

Publication Number Publication Date
WO2024114758A1 true WO2024114758A1 (zh) 2024-06-06

Family

ID=86412107

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/135600 WO2024114758A1 (zh) 2022-12-02 2023-11-30 三维牙齿模型分割方法和装置

Country Status (1)

Country Link
WO (1) WO2024114758A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108986111A (zh) * 2018-07-02 2018-12-11 西安增材制造国家研究院有限公司 一种用于计算机辅助隐形正畸的三维牙颌模型分割方法
CN109903396A (zh) * 2019-03-20 2019-06-18 洛阳中科信息产业研究院(中科院计算技术研究所洛阳分所) 一种基于曲面参数化的牙齿三维模型自动分割方法
US20210200188A1 (en) * 2018-10-23 2021-07-01 Ormco Corporation Systems and methods for designing and manufacturing an orthodontic appliance
CN113876444A (zh) * 2020-07-02 2022-01-04 广州黑格智造信息科技有限公司 一种牙科器械的3d打印方法、系统、装置及介质
WO2022092627A1 (ko) * 2020-10-28 2022-05-05 주식회사 메디트 3차원 모델로부터 대상체 영역을 결정하는 방법 및 3차원 모델 처리 장치
CN114897924A (zh) * 2022-04-24 2022-08-12 雅客智慧（北京）科技有限公司 牙齿模型分割方法及装置
CN116168185A (zh) * 2022-12-02 2023-05-26 广州黑格智造信息科技有限公司 三维牙齿模型分割方法和装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108986111A (zh) * 2018-07-02 2018-12-11 西安增材制造国家研究院有限公司 一种用于计算机辅助隐形正畸的三维牙颌模型分割方法
US20210200188A1 (en) * 2018-10-23 2021-07-01 Ormco Corporation Systems and methods for designing and manufacturing an orthodontic appliance
CN109903396A (zh) * 2019-03-20 2019-06-18 洛阳中科信息产业研究院(中科院计算技术研究所洛阳分所) 一种基于曲面参数化的牙齿三维模型自动分割方法
CN113876444A (zh) * 2020-07-02 2022-01-04 广州黑格智造信息科技有限公司 一种牙科器械的3d打印方法、系统、装置及介质
WO2022092627A1 (ko) * 2020-10-28 2022-05-05 주식회사 메디트 3차원 모델로부터 대상체 영역을 결정하는 방법 및 3차원 모델 처리 장치
CN114897924A (zh) * 2022-04-24 2022-08-12 雅客智慧（北京）科技有限公司 牙齿模型分割方法及装置
CN116168185A (zh) * 2022-12-02 2023-05-26 广州黑格智造信息科技有限公司 三维牙齿模型分割方法和装置

Also Published As

Publication number Publication date
CN116168185A (zh) 2023-05-26

Similar Documents

Publication Publication Date Title
US11194312B2 (en) Orthodontic appliances including at least partially un-erupted teeth and method of forming them
US11960795B2 (en) Neural network-based generation and placement of tooth restoration dental appliances
US20230281943A1 (en) Systems and methods for constructing a three-dimensional model from two-dimensional images
CN109310488B (zh) 用于估计牙齿修复体的形状、位置以及取向中的至少一个的方法
US11900538B2 (en) Systems and methods for constructing a dental arch image using a machine learning model
CN106504331B (zh) 基于三维模型检索的牙齿建模方法
US11850113B2 (en) Systems and methods for constructing a three-dimensional model from two-dimensional images
EP3975920A1 (en) Automated creation of tooth restoration dental appliances
WO2022001001A1 (zh) 一种3d牙模牙龈线的识别方法、系统、装置和存储介质
WO2023242757A1 (en) Geometry generation for dental restoration appliances, and the validation of that geometry
CN116246046A (zh) 三维牙齿模型分割方法和装置
WO2024114758A1 (zh) 三维牙齿模型分割方法和装置
CN116168808A (zh) 牙龈线提取方法、牙科器械的制作方法、设备及介质
CN114663621A (zh) 一种基于几何信息的三维牙冠模型优化方法
CN116168185B (zh) 三维牙齿模型分割方法和装置
CN116196123A (zh) 正畸导板的生成方法和装置
WO2023185405A1 (zh) 义齿3d打印支架的设计方法、装置及可存储介质
WO2024114343A1 (zh) 牙龈线提取方法、牙科器械的制作方法及电子设备
CN112206064B (zh) 一种数字化牙模生成方法及系统
WO2023151549A1 (zh) 面中线及牙齿矫治位置的确定方法、制造方法及系统
CN113139908A (zh) 一种三维牙列分割与标注方法
WO2024141074A1 (zh) 面中线的生成方法、矫治器及其制造方法
EP4307229A1 (en) Method and system for tooth pose estimation
TWI802309B (zh) 臨床相片智能分類與導正的方法及影像處理裝置、電腦可讀取的記錄媒體
CN117115352A (zh) 一种口腔正畸效果预览图像生成方法及装置