WO2021016996A1 - 重构点云模型的方法、装置和系统 - Google Patents

重构点云模型的方法、装置和系统 Download PDF

Info

Publication number
WO2021016996A1
WO2021016996A1 PCT/CN2019/098910 CN2019098910W WO2021016996A1 WO 2021016996 A1 WO2021016996 A1 WO 2021016996A1 CN 2019098910 W CN2019098910 W CN 2019098910W WO 2021016996 A1 WO2021016996 A1 WO 2021016996A1
Authority
WO
WIPO (PCT)
Prior art keywords
reconstruction
point
points
cloud model
point cloud
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/098910
Other languages
English (en)
French (fr)
Inventor
王海峰
邹文超
贾绍图
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Siemens Ltd China
Original Assignee
Siemens Ltd China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens Ltd China filed Critical Siemens Ltd China
Priority to EP19915575.5A priority Critical patent/EP3796266A4/en
Priority to PCT/CN2019/098910 priority patent/WO2021016996A1/zh
Priority to CN201980008746.8A priority patent/CN112602121B/zh
Priority to US16/970,691 priority patent/US11410379B2/en
Publication of WO2021016996A1 publication Critical patent/WO2021016996A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three-dimensional [3D] modelling for computer graphics
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • G06T9/001—Model-based coding, e.g. wire frame
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2210/00—Indexing scheme for image generation or computer graphics
    • G06T2210/56—Particle system, point based geometry or rendering

Definitions

  • the present invention relates to the field of point cloud models, in particular to a method, device and system for reconstructing a point cloud model.
  • the problems to be solved for reconstructing the point cloud model include controlling the resolution of the point cloud model reconstruction, including defining the resolution of the point cloud model based on a specific application. For example, in the ultrasonic mode haptic interface, the center point must be maintained at a specific density To ensure a good sense of touch relative to humans.
  • the control of the number of points in point cloud reconstruction also needs to be considered. Among them, some edge computing devices have limited computing efficiency. The user must control the number of points in the point cloud model to ensure that the equipment can operate properly.
  • how to efficiently extract the main features with acceptable accuracy in the point cloud reconstruction also needs to be considered. However, it is more difficult to efficiently extract the main features from the point cloud model based on less than the possible number of points using the prior art algorithm.
  • point cloud filters there are several point cloud filters in the prior art, such as straight-pass filters, Voxel filters, statistical filters, and conditional filters. These filters can reduce the number of points and Maintain accuracy. However, users cannot use these point cloud filters to clearly define the resolution, quantity, and precision of the point cloud structure.
  • the first aspect of the present invention provides a point cloud model reconstruction method, which includes the following steps: randomly selecting four reconstruction points in the point cloud model that are not coplanar; continuing to iteratively select other reconstruction points until the reconstruction point is satisfied. Constructing conditions, and reconstructing the point cloud model based on all reconstructed points, where the restoration degree of reconstructing the point cloud model is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape surrounded by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • g represents the proportion of the volume ratio in the reduction degree
  • the reconstruction method further includes the following steps: adjusting the ratio of g and k based on customer needs to adjust the reconstruction conditions.
  • the reconstruction condition includes when the reduction degree is the highest among a combination of a limited total number of reconstruction points, wherein the reconstruction method further includes: limiting the total number of the reconstruction points; continue to select other reconstruction points in sequence, and The reduction degree of different reconstruction point combinations of the total number is calculated respectively, and the reconstruction point combination with the highest reduction degree is selected.
  • the reconstruction condition includes when the volume ratio of the volume of the three-dimensional figure surrounded by all the reconstruction points and the original volume of the point cloud model reaches a first threshold, wherein the reconstruction method further includes: The first threshold; continue to sequentially select other reconstruction points, and respectively calculate the volume ratio of the volume of the three-dimensional figure surrounded by all current reconstruction points and the original volume of the point cloud model, until the first threshold is reached.
  • the reconstruction condition includes when the reconstruction point combination that defines the selection range of other reconstruction points based on the four reconstruction points has the highest reduction degree, wherein the reconstruction method further includes: The reconstruction points limit the selection range of other reconstruction points in the point cloud model; continue to select other reconstruction points in the selection range, and calculate the reduction degree of different reconstruction point combinations respectively, and select the highest reduction degree The combination of reconstruction points.
  • each reconstruction point can only be selected once; as the path depth for selecting the reconstruction point increases, The reconstruction points as child nodes of each layer of reconstruction point in the path are decreasing; if the first reconstruction point and the first two reconstruction points of the first reconstruction point are collinear, the first reconstruction is invalid Point and its subsequent path branches; if the second reconstruction point and the first three reconstruction points of the second reconstruction point are coplanar, the second reconstruction point and its subsequent path branches are invalid; if selected At the third reconstruction point, the reduction degree based on the combination of all reconstruction points is lower than the previous reduction degree, and the third reconstruction point and subsequent path branches are invalidated.
  • the second aspect of the present invention provides a point cloud model reconstruction device, which includes: a selection device that randomly selects four reconstruction points in the point cloud model that are not coplanar; the reconstruction device continues to iteratively select other points Reconstruct the points until the reconstruction conditions are met, reconstruct the point cloud model based on all reconstruction points, and adjust the ratio of g and k based on customer needs to adjust the reconstruction conditions, where the point cloud model is reconstructed
  • the reduction degree is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape enclosed by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • g represents the specific gravity of the volume ratio in the degree of reduction.
  • the reconstruction condition includes when the reduction degree is the highest among the combination of the limited total number of reconstruction points, wherein the reconstruction device is also used to: limit the total number of the reconstruction points; continue to select other reconstruction points in sequence, The reduction degree of different reconstruction point combinations of the total number is calculated respectively, and the reconstruction point combination with the highest reduction degree is selected.
  • the reconstruction condition includes when the volume ratio of the volume of the three-dimensional figure enclosed by all current reconstruction points to the original volume of the point cloud model reaches a first threshold, wherein the reconstruction device is also used for: The first threshold; continue to sequentially select other reconstruction points, and respectively calculate the volume ratio of the volume of the three-dimensional figure surrounded by all current reconstruction points and the original volume of the point cloud model, until the first threshold is reached.
  • the reconstruction condition includes when the reconstruction point combination that defines the selection range of other reconstruction points based on the four reconstruction points has the highest degree of reduction, wherein the reconstruction device is also used to: Each reconstruction point defines the selection range of other reconstruction points in the point cloud model; continue to select other reconstruction points in the selection range, and calculate the reduction degree of different reconstruction point combinations respectively, and select the reduction degree The highest combination of reconstruction points.
  • each reconstruction point can only be selected once; as the path depth for selecting the reconstruction point increases, The reconstruction points as child nodes of each layer of reconstruction point in the path are decreasing; if the first reconstruction point and the first two reconstruction points of the first reconstruction point are collinear, the first reconstruction is invalid Point and its subsequent path branches; if the second reconstruction point and the first three reconstruction points of the second reconstruction point are coplanar, the second reconstruction point and its subsequent path branches are invalid; if selected At the third reconstruction point, the reduction degree based on the combination of all reconstruction points is lower than the previous reduction degree, and the third reconstruction point and subsequent path branches are invalidated.
  • the third aspect of the present invention provides a point cloud model reconstruction system, including: a processor; and a memory coupled with the processor, the memory having instructions stored therein, the instructions when being executed by the processor
  • the electronic device performs an action, and the action includes: randomly selecting four reconstruction points that are not coplanar in the point cloud model; continuing to iteratively select other reconstruction points in sequence until a reconstruction condition is met, and based on all reconstruction points
  • the point cloud model is reconstructed, where the reduction degree of reconstructing the point cloud model is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape surrounded by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • the specific gravity in the degree, g represents the proportion of the volume ratio in the reduction degree, and the reconstitution condition is adjusted by adjusting the ratio of g and k based on customer needs.
  • the fourth aspect of the present invention provides a computer program product, which is tangibly stored on a computer-readable medium and includes computer-executable instructions, which when executed, cause at least one processor to execute The method described in the first aspect of the present invention.
  • the fifth aspect of the present invention provides a computer-readable medium on which computer-executable instructions are stored, and when executed, the computer-executable instructions cause at least one processor to perform the method according to the first aspect of the present invention.
  • the reconstructed point cloud model mechanism provided by the present invention can improve the resolution of the reconstructed point cloud model, can control the quality of points in the reconstructed point cloud model, and can also extract feature points from the 3D point cloud model.
  • the present invention can also perform point cloud model reconstruction based on the user's selection and input information.
  • the present invention can improve the display capability and utilization of computing resources, and the present invention can optimize the display capability of the point cloud model on the edge device.
  • the present invention can represent a point cloud model based on a small number of points, which can generate a data set for machine learning, for example, in 3D object recognition and positioning.
  • Fig. 1 is a schematic structural diagram of a point cloud model reconstruction device according to a specific embodiment of the present invention
  • FIG. 2 is a schematic diagram of a point cloud model of a point cloud model reconstruction method according to a specific embodiment of the present invention
  • Fig. 3 is a schematic diagram of a point cloud model of a point cloud model reconstruction method according to another specific embodiment of the present invention.
  • Fig. 4 is a schematic diagram of reconstructed point growth of a point cloud model reconstruction method according to a specific implementation of the present invention
  • Fig. 5 is a flow chart of three implementation modes of a point cloud model reconstruction method in a specific implementation of the present invention.
  • Fig. 1 is a schematic structural diagram of a point cloud structure reconstruction device according to a specific embodiment of the present invention.
  • the point cloud structure reconstruction device includes a user interaction device 100, a restoration degree calculation device 120, a point processing device 130, a path switching device 140 and a mode switching device 150.
  • the user enters relevant information in the user interaction device 100 and imports the 3D model to obtain the point cloud model that needs to be processed.
  • the selected reconstruction point is sent to the point processing device 130 to determine whether the selection of the reconstruction point is a composite rule, and then the reduction degree is calculated
  • the device 120 calculates the restoration degree of the reconstructed point for the complex rule reconstruction point, and controls the growth and cutting of the path branch of the iterative reconstruction point in the path switching device 140, and finally the mode switching device 150 serves as a termination output
  • the means for judging the trigger condition of the means 166 outputs the final selected reconstruction point and its growth path and degree of restoration.
  • the first aspect of the present invention provides a point cloud model reconstruction method, which includes the following steps.
  • step S1 is performed to randomly select four reconstruction points in the point cloud model that are not coplanar.
  • the user interaction device includes a model introduction device 112, a mode setting device 114, a parameter input device 116, and a point initialization device 118.
  • the model importing device 112 is used to import a 3D model to obtain a point cloud model for processing, and the user selects four reconstruction points that are not coplanar through the point initialization device 118.
  • the present invention may randomly select four non-coplanar reconstruction points in the point cloud model according to the information input by the user in the user interaction device 100.
  • the four non-coplanar reconstruction points selected by the user through the point initialization device 118 are the first reconstruction point p1, the second reconstruction point p2, and the third reconstruction point.
  • the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, and the fourth reconstruction point p4 are not coplanar.
  • the present invention can also randomly designate four non-coplanar initial reconstruction points without relying on the user's designation.
  • step S2 the point iterative device 162 continues to iteratively select other reconstruction points until the reconstruction conditions are met, and reconstruct the point cloud model based on all reconstruction points, wherein the restoration degree of the reconstructed point cloud model is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape surrounded by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • g represents the proportion of the volume ratio in the reduction degree
  • the reconstruction method further includes the following steps: adjusting the ratio of g and k based on customer needs to adjust the reconstruction conditions.
  • the negative number calculation device 122 in the reduction degree calculation device 120 is used to increase the weight of the negative number factor (ie, k) by increasing the number of reconstruction points of iteration, so as to reduce the reduction degree.
  • the volume ratio calculation device 124 is used to calculate the volume ratio by reconstructing the spatial information from the point processing device 130, and to increase the authority of a positive factor (ie, g) to increase the degree of reduction.
  • the negative number calculation device 122 and the volume ratio calculation device 124 in the calculation device 120 can weigh the weights of g and k, and can also configure g and k according to reconstruction conditions.
  • the parameter input device 116 can control the weights of g and k based on the user's selection.
  • the point iteration device 162 is then triggered to continue to select other points in the point cloud model.
  • Each reconstructed point selected afterwards will be sent to the point processing device 130 to use the spatial information calculating device 132 to calculate the spatial information of all reconstructed points.
  • the calibration device 134 is used to check the spatial information and determine whether the reconstructed point meets the rule 1 and the rule 2. If rule 1 and rule 2 are met, the reconstruction point will change the volume ratio based on the spatial information, and send the volume ratio to the reduction degree calculation device 120 to calculate the reduction degree of the reconstruction point.
  • This reconstruction point is recorded in the path growth device 142 of the path switching device 140, and the path cutting device 144 is used to control the growth and cutting of the path branch of the iterative reconstruction point.
  • the recorded reconstruction point will be sent and stored in the point and path storage device 163, and the branch from this reconstruction point will further grow. If not, the branch from this reconstruction point will be cut by the path cutting device 144 in the path switching device 140, which means that the branch starting from this reconstruction point no longer grows.
  • the negative number calculation device 122 in the reduction degree calculation device 120 is used to increase the weight of the negative number factor by increasing the number of reconstruction points of the iteration to reduce the reduction degree.
  • the volume ratio calculation device 124 is used to calculate the volume ratio by reconstructing the spatial information from the point processing device 130, and to increase the authority of the positive factor to increase the degree of reduction. In this way, the restoration degree of the reconstruction point of each iteration will be calculated.
  • Each calculated reduction degree is compared with the previous reduction degree value in the comparison calculation device 164. If the current restoration degree is greater, the restoration degree will be sent to the temporary storage device 165 to replace the previous restoration degree. If the current reduction degree is smaller, the branch from the reconstruction point of this iteration will be cut by the path cutting device 144. Through the initially set mode, the threshold of this mode triggers the termination of the output device 166.
  • the volume ratio calculated by the volume ratio calculation device 124 is sent to the volume ratio comparison device 152 in the mode switching device 150 to compare the volume ratio and the volume ratio threshold, and the comparison result serves as a trigger condition for terminating the output device 166.
  • the number of stored reconstruction points is sent to the depth comparison device 154 to compare the number of stored reconstruction points with the depth threshold, or the reconstruction point comparison device 156 is used to compare the number of stored reconstruction points And the threshold value of the number of reconstruction points are used as the two trigger conditions for triggering the termination output device 166. Once the termination output device 166 is triggered, the reconstructed points and paths stored in the point and path storage device 163 will be converted into output together with the final restoration degree.
  • iteratively selecting other reconstruction points is the growth process of reconstruction points.
  • four reconstruction points are initially selected, namely the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, and the fourth reconstruction point p4.
  • other reconstruction points are selected from the fourth reconstruction point P4 to form a growth path of the reconstruction point, and the path depth indicates how many layers the reconstruction point has.
  • the path depth indicates how many layers the reconstruction point has.
  • the next depth of each reconstruction point or the reconstruction point of the next layer is the child node of the reconstruction point, for example, the fifth reconstruction point P5, the sixth reconstruction point P6, and the seventh reconstruction point P7, the eighth reconstruction point P8, and the nth reconstruction point Pn are child nodes of the fourth reconstruction point P4.
  • the reconstruction method provided by the present invention should be executed based on certain rules.
  • the basic rule is: in the reconstruction method executed under each reconstruction condition, each reconstruction point can only be selected once; as the reconstruction point is selected The depth of the path increases, and the reconstruction points as child nodes of each layer of the path are decreasing. Regardless of whether the reconstruction method provided by the present invention is based on the first mode, the low second mode, and the third mode, the above-mentioned rules should be observed.
  • rule 1 is: if the first reconstruction point and the first two reconstruction points of the first reconstruction point are collinear, then the first reconstruction point and subsequent path branches are invalid;
  • rule 2 is: if the second reconstruction point If the reconstruction point is coplanar with the first three reconstruction points of the second reconstruction point, then the second reconstruction point and subsequent path branches will be invalid;
  • rule 3 is: if the third reconstruction point is selected based on all The reduction degree of the reconstruction point combination is lower than the previous reduction degree, and the third reconstruction point and subsequent path branches are invalidated.
  • the reconstruction method provided by the present invention can be executed in different modes.
  • the following describes the reconstruction method provided by the present invention according to three different implementation modes.
  • the first implementation When the first implementation is selected, it is set to the first implementation in the mode setting device 114 Way and set relevant thresholds.
  • the parameter input device 116 calculates the input parameters of the degree of restoration, and the parameters related to the first implementation manner are stored in the parameter storage device 161, and the above parameters are read when needed.
  • the reconstruction condition includes when the reduction degree is the highest among a combination of a limited total number of reconstruction points, wherein the reconstruction method further includes: limiting the total number of reconstruction points; continue to select other reconstruction points in sequence, and respectively Calculate the reduction degree of the total number of different reconstruction point combinations, and select the reconstruction point combination with the highest reduction degree. For example, if the total number of reconstruction points is limited to 8, then the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, and the fourth reconstruction point p4 are selected as shown in 2. The fifth reconstruction point, the sixth reconstruction point, the seventh reconstruction point and the eighth reconstruction point are selected. Every time a new reconstruction point is selected, the current restoration degree is recalculated.
  • the fourth reduction degree of reconstructing the point cloud model is:
  • VolRate 4 is the volume ratio between the three-dimensional shape enclosed by the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, and the fourth reconstruction point p4 and the original volume of the point cloud model 200, where ,
  • the three-dimensional shape enclosed by the first reconstructed point p1, the second reconstructed point p2, the third reconstructed point p3, and the fourth reconstructed point p4 is the area of the shaded part shown in FIG. 2, and the original volume of the point cloud model 200 It is the volume of the cube.
  • the original volume of the point cloud model 200 can be acquired when the model introduction device 112 imports the 3D model and acquires the point cloud model for processing.
  • the fourth restoration degree of reconstructing the point cloud model is:
  • VolRate 5 is the three-dimensional shape and point cloud model 200 enclosed by the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, the fourth reconstruction point p4, and the fifth reconstruction point p5.
  • the fifth select another reconstruction point p5 ' VolRate 5' as a first reconstruction points p1, a second reconstruction point p2, the third reconstruction point p3, p4 fourth and fifth reconstruction point
  • the volume ratio between the three-dimensional shape enclosed by the reconstructed point p5' and the original volume of the point cloud model 200 is reconstructed. Therefore, if the randomly selected fifth reconstruction point is different, the volume of the three-dimensional shape enclosed by the five reconstruction points is different.
  • the randomly selected sixth reconstruction point is different, the volume of the three-dimensional shape enclosed by the six reconstruction points is also different... so that the combination of the 8 reconstruction points is different, the reduction degree value is also different.
  • the seventh reconstruction point and the eighth reconstruction point can randomly select all points in the point cloud model 200 except for the first reconstruction point p1, the second reconstruction point p2, the third reconstruction point p3, and the fourth reconstruction point p4.
  • the combination of other points therefore, after permutation and combination, whenever a reconstruction point is selected, the current restoration degree is calculated, and all points in the point cloud model 200 except the first reconstruction point p1, the second reconstruction point p2, and the third reconstruction point are exhausted.
  • a combination of 8 points other than the construction point p3 and the fourth reconstruction point p4 until the highest degree of reduction Reward 8 is selected .
  • the second mode which defines the volume ratio threshold of the volume of the three-dimensional figure surrounded by all current reconstruction points and the original volume of the point cloud model, and once the threshold is reached, the current reconstruction point combination is output.
  • the second implementation manner is selected, the second implementation manner is set in the mode setting device 114, and the volume ratio threshold is set.
  • the parameter input device 116 calculates the input parameters of the degree of restoration, and the parameters related to the third implementation manner are stored in the parameter storage device 161, and the aforementioned parameters are read when needed.
  • the reconstruction condition includes when the volume ratio of the volume of the three-dimensional figure enclosed by all reconstruction points and the original volume of the point cloud model reaches a first threshold
  • the reconstruction method further includes: defining the The first threshold; continue to sequentially select other reconstruction points, and respectively calculate the volume ratio of the volume of the three-dimensional figure surrounded by all current reconstruction points to the original volume of the point cloud model, until the first threshold is reached. Assuming that the first threshold is 95%, other reconstruction points are always grown from the fourth reconstruction point p4, and the volume ratio is always calculated. Once it reaches 95%, the current reconstruction point combination and its reduction degree are output.
  • the third mode which limits the selection range of other points after the fourth reconstruction point is selected, and selects the reconstruction point combination with the highest degree of restoration in the range.
  • the third implementation manner is set in the mode setting device 114, and the selection range of other reconstruction points is set.
  • the parameter input device 116 calculates the input parameters of the degree of restoration, and the parameters related to the third implementation mode are stored in the parameter storage device 161, and the above parameters are read when needed.
  • the reconstruction condition includes when the reconstruction point combination that defines the selection range of other reconstruction points based on the four reconstruction points has the highest reduction degree
  • the reconstruction method further includes: using the four reconstruction points
  • the structure points limit the selection range of other reconstruction points in the point cloud model; continue to select other reconstruction points in the selection range, and calculate the reduction degrees of different reconstruction point combinations, and select the highest reduction point Refactor the point combination.
  • this embodiment limits the selection range of four other reconstruction points.
  • the range delineated in the first selection range surrounds the first reconstruction point p1, which means that the points available for selection are reconstruction points.
  • the selection range of reconstruction points may be based on the user's selection.
  • Fig. 5 is a flow chart of three implementation modes of a point cloud model reconstruction method in a specific implementation of the present invention.
  • the user can input the parameters g and k via the parameter input device 116, or can select the first mode and the second mode via the mode setting device 114 As with the third mode, four reconstruction points that are not coplanar can also be selected by the point initialization device 118.
  • the user can also based on, as shown in Figure 5, first determine whether the mode is the third mode, if it is, follow M3 to execute the step process, if not continue to determine whether the mode is the second mode, if it is, follow M1 to execute the step process , Otherwise follow M2 to execute the step flow.
  • the point iterative device 162 sequentially executes iterative selection of other reconstruction points starting from the fifth reconstruction point, and uses the spatial information calculating device 132 to obtain the spatial information of the selected reconstruction point. Then judge whether the reconstructed point meets the rules 1, 2 and 3, and then judge whether the rules 1, 2 and 3 are passed, if yes, start the path growth of the reconstructed point from the fourth reconstruction point and store the reconstructed point and its path In the path growth device 142, otherwise, the path clipping device 144 controls the growth and shearing of the path branch of the iterative reconstruction point.
  • the reconstruction point comparison device 156 determines whether the reconstruction point combination reaches the maximum number of reconstruction points, and if so, the termination output device 166 outputs the reconstruction point combination and its degree of reduction. If not, the point iteration device 162 is triggered again to perform iteration selection of other reconstruction points.
  • the point iterative device 162 sequentially executes the iterative selection of other reconstruction points starting from the fifth reconstruction point, and uses the spatial information calculating device 132 to obtain the spatial information of the selected reconstruction point. Then judge whether the reconstructed point meets the rules 1, 2 and 3, and then judge whether the rules 1, 2 and 3 are passed, if yes, start the path growth of the reconstructed point from the fourth reconstruction point and store the reconstructed point and its path In the path growth device 142, otherwise, the path clipping device 144 controls the growth and shearing of the path branch of the iterative reconstruction point. The number of stored reconstruction points will be sent to the depth comparison device 154.
  • the depth comparison device 154 is used to compare the number of stored reconstruction points and the depth threshold to determine whether the depth has reached the total number of reconstruction points, if so
  • the terminal output device 166 outputs the reconstructed point combination and its restoration degree. If not, the point iteration device 162 is triggered again to perform iteration selection of other reconstruction points.
  • the point iterative device 162 sequentially executes the iterative selection of other reconstruction points starting from the fifth reconstruction point, and uses the spatial information calculating device 132 to obtain the spatial information of the selected reconstruction point. Then it is judged whether the reconstructed point meets the rules 1, 2 and 3, and then it is judged to pass the rules 1, 2 and 3, otherwise, the path clipping device 144 controls the growth and cutting of the path branch of the iterative reconstruction point.
  • the volume ratio calculation device 124 calculates the current volume ratio of each reconstruction point, and will be sent to the volume ratio comparison device 152 in the mode switching device 150 to determine whether the volume ratio reaches the first threshold, or if it is not.
  • the output device 166 outputs the reconstructed point combination and its degree of reduction. If not, the point iteration device 162 is triggered again to perform iteration selection of other reconstruction points.
  • the second aspect of the present invention provides a point cloud model reconstruction device, which includes: a selection device that randomly selects four reconstruction points in the point cloud model that are not coplanar; the reconstruction device continues to iteratively select other points Reconstruct the points until the reconstruction conditions are met, reconstruct the point cloud model based on all reconstruction points, and adjust the ratio of g and k based on customer needs to adjust the reconstruction conditions, where the point cloud model is reconstructed
  • the reduction degree is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape surrounded by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • g represents the specific gravity of the volume ratio in the degree of reduction.
  • the reconstruction condition includes when the reduction degree is the highest among the combination of the limited total number of reconstruction points, wherein the reconstruction device is also used to: limit the total number of the reconstruction points; continue to select other reconstruction points in sequence, The reduction degree of different reconstruction point combinations of the total number is calculated respectively, and the reconstruction point combination with the highest reduction degree is selected.
  • the reconstruction condition includes when the volume ratio of the volume of the three-dimensional figure enclosed by all current reconstruction points to the original volume of the point cloud model reaches a first threshold, wherein the reconstruction device is also used for: The first threshold; continue to sequentially select other reconstruction points, and respectively calculate the volume ratio of the volume of the three-dimensional figure surrounded by all current reconstruction points and the original volume of the point cloud model, until the first threshold is reached.
  • the reconstruction condition includes when the reconstruction point combination that defines the selection range of other reconstruction points based on the four reconstruction points has the highest degree of reduction, wherein the reconstruction device is also used to: Each reconstruction point defines the selection range of other reconstruction points in the point cloud model; continue to select other reconstruction points in the selection range, and calculate the reduction degree of different reconstruction point combinations respectively, and select the reduction degree The highest combination of reconstruction points.
  • each reconstruction point can only be selected once; as the path depth for selecting the reconstruction point increases, The reconstruction points as child nodes of each layer of reconstruction point in the path are decreasing; if the first reconstruction point and the first two reconstruction points of the first reconstruction point are collinear, the first reconstruction is invalid Point and its subsequent path branches; if the second reconstruction point and the first three reconstruction points of the second reconstruction point are coplanar, the second reconstruction point and its subsequent path branches are invalid; if selected At the third reconstruction point, the reduction degree based on the combination of all reconstruction points is lower than the previous reduction degree, and the third reconstruction point and subsequent path branches are invalidated.
  • the third aspect of the present invention provides a point cloud model reconstruction system, including: a processor; and a memory coupled with the processor, the memory having instructions stored therein, the instructions when being executed by the processor
  • the electronic device performs an action, and the action includes: randomly selecting four reconstruction points that are not coplanar in the point cloud model; continuing to iteratively select other reconstruction points in sequence until a reconstruction condition is met, and based on all reconstruction points
  • the point cloud model is reconstructed, where the reduction degree of reconstructing the point cloud model is:
  • PointNum represents the number of currently selected reconstruction points
  • VolRate represents the volume ratio of the volume of the three-dimensional shape surrounded by all current reconstruction points to the original volume of the point cloud model
  • k represents the number of selected points in the restoration
  • the specific gravity in the degree, g represents the proportion of the volume ratio in the reduction degree, and the reconstitution condition is adjusted by adjusting the ratio of g and k based on customer needs.
  • the fourth aspect of the present invention provides a computer program product, which is tangibly stored on a computer-readable medium and includes computer-executable instructions, which when executed, cause at least one processor to execute The method described in the first aspect of the present invention.
  • the fifth aspect of the present invention provides a computer-readable medium on which computer-executable instructions are stored, and when executed, the computer-executable instructions cause at least one processor to perform the method according to the first aspect of the present invention.
  • the reconstructed point cloud model mechanism provided by the present invention can improve the resolution of the reconstructed point cloud model, can control the quality of points in the reconstructed point cloud model, and can also extract feature points from the 3D point cloud model.
  • the present invention can also perform point cloud model reconstruction based on the user's selection and input information.
  • the present invention can improve the display capability and utilization of computing resources, and the present invention can optimize the display capability of the point cloud model on the edge device.
  • the present invention can represent a point cloud model based on a small number of points, which can generate a data set for machine learning, for example, in 3D object recognition and positioning.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Processing Or Creating Images (AREA)
  • Image Processing (AREA)

Abstract

本发明提供了重构点云模型的方法、装置和系统，其中，包括如下步骤：随机选择所述点云模型中不共面的四个重构点；继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：Reward＝-k·(PointNum-4)+g·VolRate，其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，其中，所述重构方法还包括如下步骤：基于客户需求调整g和k的比例来调整所述重构条件。本发明提供的重构点云模型机制能够改善重构的点云模型的分辨率，能够控制在重构的点云模型中的点的质量。

Description

重构点云模型的方法、装置和系统 技术领域
本发明涉及点云模型领域，尤其涉及一种重构点云模型的方法、装置和系统。
背景技术
现在，重构点云模型待解决的问题包括控制点云模型重构的分辨率，包括基于特定申请定义点云模型的分辨率，例如，在超声波模式触觉接口中，中心点必须保持在特定密度以保证相对于人来说好的触觉感。另一方面，点云重构中点的数量控制也需要考虑。其中，一些边缘计算设备具有有限的计算效率。用户必须控制在点云模型中的点的数量来保证设备可以适当运行。其次，如何高效提取点云重建中具有可接受精确度的主要特征也是需要考虑的，然而利用现有技术算法基于少于可能数量的点的点云模型中高效提取主要特征是比较困难的。
现有技术有几个点云过滤器，例如直通器(straight-pass filter)、Voxel过滤器、统计滤波器(statistic filters)和条件滤波器(Conditional filter)，这些过滤器可以减少点的数量并保持精确度。然而，用户并不能利用这些点云过滤器明确定义点云结构的分辨率、数量和精度。
发明内容
本发明第一方面提供了点云模型的重构方法，其中，包括如下步骤：随机选择所述点云模型中不共面的四个重构点；继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，其中，所述重构方法还包括如下步骤：基于客户需求 调整g和k的比例来调整所述重构条件。
进一步地，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构方法还包括：限定所述重构点的总数；继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构方法还包括：定义所述第一阈值；继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。
进一步地，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构方法还包括：以所述四个重构点在所述点云模型中限定其他重构点的选择范围；在所述选择范围中继续依次选择其他重构点，并分别计算不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构方法基于以下规则执行：在每个重构条件下执行的重构方法中，每个重构点只能选择一次；随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少；如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；如果选择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
本发明第二方面提供了点云模型的重构装置，其中，包括：选择装置，其随机选择所述点云模型中不共面的四个重构点；重构装置，继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，并基于客户需求调整g和k的比例来调整所述重构条件，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例， k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重。
进一步地，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构装置还用于：限定所述重构点的总数；继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构装置还用于：定义所述第一阈值；继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。
进一步地，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构装置还用于：以所述四个重构点在所述点云模型中限定其他重构点的选择范围；在所述选择范围中继续依次选择其他重构点，并分别计算不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构装置基于以下规则执行：在每个重构条件下执行的重构方法中，每个重构点只能选择一次；随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少；如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；如果选择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
本发明第三方面提供了点云模型的重构系统，包括：处理器；以及与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述电子设备执行动作，所述动作包括：随机选择所述点云模型中不共面的四个重构点；继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，并且基于客户需求调整g和k的比例来调整所述重构条件。
本发明第四方面提供了计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明第五方面提供了计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明提供的重构点云模型机制能够改善重构的点云模型的分辨率，能够控制在重构的点云模型中的点的质量，还能够从3D点云模型中提取特征点。本发明还能够基于用户的选择和输入的信息执行点云模型重构。此外，本发明能够提高显示能力和计算资源的利用，本发明能够优化点云模型在边缘设备上的显示能力。另外，本发明能够基于少量点来表现点云模型，其能够为机器学习产生数据集，例如在3D物体识别和定位中。
附图说明
图1是根据本发明一个具体实施例的点云模型重构装置的结构示意图；
图2是根据本发明一个具体实施例的点云模型重构方法的点云模型示意图；
图3是根据本发明又一具体实施例的点云模型重构方法的点云模型示意图；
图4是根据本发明一个具体实施里的点云模型重构方法的重构点生长示意图；
图5是根据本发明一个具体实施里的点云模型重构方法的三种实现模式的步骤流程图。
具体实施方式
以下结合附图，对本发明的具体实施方式进行说明。
图1是根据本发明一个具体实施例的点云结构重构装置的结构示意图。如图1所示，点云结构重构装置包括用户交互装置100、还原度计算装置120、点处理装置130、路径切换装置140和模式切换装置150。用户在用户交互装置100中输入相关信息并导入3D模型来获取需要处理的点云模型，被选择的重构点发送给点处理装置130来判断重构点的选取是否复合规则，然后还原度计算装置120针对复合规则的重构点计算该重构点的还原度，并在路径切换装置140中控制迭代的重构点的路径支路的生长和剪切，最后模式切换装置150中充当终结输出装置166的触发条件的判断装置，输出最终选择重构点及其生长路径和还原度。
本发明第一方面提供了一种点云模型的重构方法，其中，包括如下步骤。
首先执行步骤S1，随机选择所述点云模型中不共面的四个重构点。具体地，所述用户交互装置包括模型导入装置112、模式设置装置114、参数输入装置116和点初始化装置118。其中，模型导入装置112用于导入3D模型以获取用于处理的点云模型，用户通过点初始化装置118选择不共面的四个重构点。
可选地，本发明可以根据用户在用户交互装置100中输入的信息随机选择所述点云模型中不共面的四个重构点。例如，基于如图2所示的点云模型200，用户通过点初始化装置118选择的不共面的四个重构点，分别为第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4。其中，如图2所示，第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4不共面。
当然，本发明也可以随机指定四个不共面的初始重构点，而无需依赖于用户的指定。
然后执行步骤S2，点迭代装置162继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，其中，所述重构方法还包括如下步骤：基于客户需求调整g和k的比例来调整所述重构条件。具体地，如图2所示，在还原度计算装置120中的负数计算装置122用于通过增加迭代的重构点的数量增加负数因素(即k)的权重，以减少还原度。体积比例计算装置124用于通过重构点处理装置130来的空间信息计算体积比例，并增加正数因素(即g)的权限以增加还原度。计算装置120中的负数计算装置122和体积比例计算装置124可以权衡g和k的权重，也可以根据重构条件来配置g和k。参数输入装置116可以基于用户的选择控制g和k的权重。
具体地，如图1所示，在选择了不共面的四个重构点之后，随即触发点迭代装置162继续在点云模型中选择其他点。之后选择的每个重构点都会发送给点处理装置130，以用空间信息计算装置132来计算所有重构点的空间信息。校准装置134用于检查空间信息，判断重构点是否符合规则1和规则2。如果符合规则1和规则2，重构点会基于空间信息来改变体积比例，并将所述体积比例发送给还原度计算装置120以计算该重构点的还原度。这个重构点会记录在路径切换装置140的路径生长装置142中，路径剪切装置144用于控制迭代的重构点的路径支路的生长和剪切。记录下来的重构点会被发送并存储于点及路径存储装置163中，而从这个重构点开始的支路会进一步生长。如果不是，从这个重构点来的这个支路会被路径切换装置140中的路径剪切装置144剪切，这意味着从这个重构点开始的支路不再生长。在还原度计算装置120中的负数计算装置122用于通过增加迭代的重构点的数量增加负数因素的权重，以减少还原度。体积比例计算装置124用于通过重构点处理装置130来的空间信息计算体积比例，并增加正数因素的权限以增加还原度。这样，每个迭代的重构点的还原度都会被计算。每个计算出的还原度会被与前一个还原度值在比较计算装置164中比较。如果当下的还原度更大，该还原度则会被发送给暂时存储装置165中来替代前一还原度。如果当下的还原度更小，从这个迭代的重构点来的支路就会被路径剪切装置144 剪切。通过初始设置的模式，该模式的阈值触发终结输出装置166。由体积比例计算装置124计算的体积比例则会被发给模式切换装置150中的体积比例比较装置152，以比较体积比例和体积比例阈值，比较结果充当终结输出装置166的触发条件。存储的重构点的个数会被发送给深度比较装置154，以比较存储的重构点的个数和深度阈值，或者，重构点比较装置156用于比较存储的重构点的个数和重构点的个数阈值，从而作为触发终结输出装置166的两个触发条件。一旦终结输出装置166触发，存储在点及路径存储装置163中的重构点和路径会和最终还原度一起转化为输出。
具体地，迭代依次选择其他重构点即是重构点的生长过程。如图4所示，初始选择了四个重构点，分别为第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4。接下来，从第四重构点P4开始选择其他重构点，以形成一条重构点的生长路径，路径深度则表示该重构点有多少层。具体地，如图4所示，在分别为第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4、第五重构点p5、第九重构点p9……第n重构点构成的生长路径中，第六重构点P6、第七重构点P7、第八重构点P8和第n重构点Pn是深度为5，第九重构点p9、第十重构点P10、第十一重构点P11和第十二重构点P12则深度为6……第n1重构点pn1、第n2重构点pn2、第n3重构点pn3、第nn重构点pnn则是深度为n。其中，每个重构点的下一个深度或者下一层的重构点则为该重构点的子节点，例如，第五重构点P5、第六重构点P6、第七重构点P7、第八重构点P8和第n重构点Pn为第四重构点P4的子节点。
进一步地，本发明提供的重构方法应基于一定规则执行，基本规则为：在每个重构条件下执行的重构方法中，每个重构点只能选择一次；随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少。不论本发明提供的重构方法是基于第一模式、低二模式和第三模式都应当遵守上述规则。
此外，本发明还应当基于以下规则执行。规则1为：如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；规则2为：如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；规则3为：如果选 择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
按照重构条件不同，本发明提供的重构方法可以按照不同的模式执行，下面按照三种不同的实现模式介绍本发明提供的重构方法。
首先介绍第一模式，其限定了重构点的总数并选出还原度最高的重构点组合作为输出，当选择第一种实现方式时，则在模式设置装置114中设置为第一种实现方式，并设置相关阈值。参数输入装置116还原度的计算输入参数，第一种实现方式相关的参数会被存储在参数存储装置161中，并在需要时读取上述参数。
其中，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构方法还包括：限定所述重构点的总数；继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。例如，将重构点的总数限定为8个，则在如2所示选择了第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4以后继续选定第五重构点、第六重构点、第七重构点和第八重构点。每新选定一个重构点，就重新计算一次当下的还原度。
具体地，当选择了第四重构点p4之后，重构所述点云模型的第四还原度为：
Reward 4＝-k·(8-4)+g·VolRate＝-4k+g·VolRate 4，
其中，VolRate 4为第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4围起来的立体形状和点云模型200的原始体积的体积比例，其中，第一重构点p1、第二重构点p2、第三重构点p3和第四重构点p4围起来的立体形状为图2所示的阴影部分面积，点云模型200的原始体积则为正方体的体积。需要说明的是，在模型导入装置112导入3D模型并获取用于处理的点云模型时则可以获取点云模型200的原始体积。
如图2所示，当选择了第五重构点p5以后，重构所述点云模型的第四还原度为：
Reward 5＝-4k+g·VolRate 5，
其中，VolRate 5为第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4和第五重构点p5围起来的立体形状和点云模型200的原始体积的体积比例。可选地，如果选择另一个第五重构点p5’，VolRate 5′为 第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4和第五重构点p5’围起来的立体形状和点云模型200的原始体积的体积比例。因此，随机选择的第五重构点不同，则五个重构点围起来的立体形状的体积不同。以此类推，随机选择的第六重构点不同，则六个重构点围起来的立体形状的体积也不同……从而8个重构点的组合不同，则还原度值也不同。
因此，在本实施例里，第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4确定以后，第五重构点、第六重构点、第七重构点和第八重构点可以随机选择点云模型200中所有除了第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4以外的其他点的组合，因此，经过排列组合每当选取一个重构点则计算当前的还原度，穷尽所有点云模型200中所有除了第一重构点p1、第二重构点p2、第三重构点p3、第四重构点p4以外的其他点的8个点组合，直至选出最高的还原度Reward 8。
然后介绍第二模式，其限定了当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例阈值，一旦达到阈值则输出此时的重构点组合。当选择第二种实现方式时，则在模式设置装置114中设置为第二种实现方式，并设置体积比例阈值。参数输入装置116还原度的计算输入参数，第三种实现方式相关的参数会被存储在参数存储装置161中，并在需要时读取上述参数。
其中，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构方法还包括：定义所述第一阈值；继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。假设第一阈值为95％，因此一直从第四重构点p4开始生长其他重构点，一直计算体积占比，一旦达到95％则输出此时的重构点组合以及其还原度。
然后介绍第三模式，其限定了选定第四个重构点以后的其他点的选择范围，选择范围内还原度最高的重构点组合。当选择第三种实现方式时，则在模式设置装置114中设置为第三种实现方式，并设置其他重构点的选择范围。参数输入装置116还原度的计算输入参数，第三种实现 方式相关的参数会被存储在参数存储装置161中，并在需要时读取上述参数。
其中，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构方法还包括：以所述四个重构点在所述点云模型中限定其他重构点的选择范围；在所述选择范围中继续依次选择其他重构点，并分别计算不同重构点组合的还原度，选出还原度最高的重构点组合。如图3所示，本实施例限定了四个其他重构点的选择范围，第一个选择范围中圈定的范围围绕着第一重构点p1，这意味着可供选择的点为重构点p11和p12；第二个选择范围中圈定的范围围绕着第一重构点p2，这意味着可供选择的点为点p21、p22和p23；第三个选择范围中圈定的范围围绕着第三重构点p3，这意味着可供选择的点为点p31；第四个选择范围中圈定的范围围绕着第四重构点p4，这意味着可供选择的点为点p41、p42、p43和p44。在重构点p11、p12、p21、p22、p23、p41、p42、p43和p44中迭代选择还原度最高的重构点组合即可。
需要说明的是，可选地，重构点选择范围可基于用户的选择。
图5是根据本发明一个具体实施里的点云模型重构方法的三种实现模式的步骤流程图。根据本发明一个具体实施例，在模型导入装置112导入3D模型获取点云模型以后，用户可以通过参数输入装置116输入参数g和k，也可以通过模式设置装置114选择第一模式、第二模式和第三模式中的一种，还可以通过点初始化装置118选择不共面的四个重构点。用户也可以基于，如图5所示，首先判断模式是否为第三模式，如果是的话按照M3来执行步骤流程，如果不是继续判断模式是否为第二模式，如果是的话按照M1来执行步骤流程，否则就按照M2来执行步骤流程。
当模式为第三模式M3时，点迭代装置162则从第五重构点开始依次执行其他重构点的迭代选择，并且利用空间信息计算装置132获取所选择的重构点的空间信息。然后判断重构点是否符合规则1、2和3，接着判断通过规则1、2和3，如果是的话就从第四重构点开始进行重构点的路径生长并且存储重构点及其路径在路径生长装置142中，否则就通过路径剪切装置144控制迭代的重构点的路径支路的生长和剪切。最后 通过重构点比较装置156判断重构点组合是否达到最大重构点数量，如果是的话则通过终结输出装置166输出重构点组合及其还原度。如果不是的话则重新触发点迭代装置162进行其他重构点的迭代选择。
当模式为第一模式M1时，点迭代装置162则从第五重构点开始依次执行其他重构点的迭代选择，并且利用空间信息计算装置132获取所选择的重构点的空间信息。然后判断重构点是否符合规则1、2和3，接着判断通过规则1、2和3，如果是的话就从第四重构点开始进行重构点的路径生长并且存储重构点及其路径在路径生长装置142中，否则就通过路径剪切装置144控制迭代的重构点的路径支路的生长和剪切。存储的重构点的个数会被发送给深度比较装置154，深度比较装置154用于比较存储的重构点的个数和深度阈值，判断深度是否达到了重构点限制总数，如果是的话则通过终结输出装置166输出重构点组合及其还原度。如果不是的话则重新触发点迭代装置162进行其他重构点的迭代选择。
当模式为第二模式M2时，点迭代装置162则从第五重构点开始依次执行其他重构点的迭代选择，并且利用空间信息计算装置132获取所选择的重构点的空间信息。然后判断重构点是否符合规则1、2和3，接着判断通过规则1、2和3，否则就通过路径剪切装置144控制迭代的重构点的路径支路的生长和剪切。由体积比例计算装置124计算出每个重构点当下的体积比例，并会被发给模式切换装置150中的体积比例比较装置152，用以判断体积比例是否达到第一阈值，如果不是的话就继续进行重构点的路径生长并且存储重构点及其路径在路径生长装置142中，然后通过重构点比较装置156判断重构点组合是否达到最大重构点数量，如果是的话则通过终结输出装置166输出重构点组合及其还原度。如果不是的话则重新触发点迭代装置162进行其他重构点的迭代选择。
本发明第二方面提供了点云模型的重构装置，其中，包括：选择装置，其随机选择所述点云模型中不共面的四个重构点；重构装置，继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，并基于客户需求调整g和k的比例来调整所述重构条件，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所 有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重。
进一步地，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构装置还用于：限定所述重构点的总数；继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构装置还用于：定义所述第一阈值；继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。
进一步地，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构装置还用于：以所述四个重构点在所述点云模型中限定其他重构点的选择范围；在所述选择范围中继续依次选择其他重构点，并分别计算不同重构点组合的还原度，选出还原度最高的重构点组合。
进一步地，所述重构装置基于以下规则执行：在每个重构条件下执行的重构方法中，每个重构点只能选择一次；随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少；如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；如果选择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
本发明第三方面提供了点云模型的重构系统，包括：处理器；以及与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述电子设备执行动作，所述动作包括：随机选择所述点云模型中不共面的四个重构点；继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
Reward＝-k·(PointNum-4)+g·VolRate，
其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，并且基于客户需求调整g和k的比例来调整所述重构条件。
本发明第四方面提供了计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明第五方面提供了计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据本发明第一方面所述的方法。
本发明提供的重构点云模型机制能够改善重构的点云模型的分辨率，能够控制在重构的点云模型中的点的质量，还能够从3D点云模型中提取特征点。本发明还能够基于用户的选择和输入的信息执行点云模型重构。此外，本发明能够提高显示能力和计算资源的利用，本发明能够优化点云模型在边缘设备上的显示能力。另外，本发明能够基于少量点来表现点云模型，其能够为机器学习产生数据集，例如在3D物体识别和定位中。
尽管本发明的内容已经通过上述优选实施例作了详细介绍，但应当认识到上述的描述不应被认为是对本发明的限制。在本领域技术人员阅读了上述内容后，对于本发明的多种修改和替代都将是显而易见的。因此，本发明的保护范围应由所附的权利要求来限定。此外，不应将权利要求中的任何附图标记视为限制所涉及的权利要求；“包括”一词不排除其它权利要求或说明书中未列出的装置或步骤；“第一”、“第二”等词语仅用来表示名称，而并不表示任何特定的顺序。

Claims (13)

  1. 点云模型的重构方法，其中，包括如下步骤：
    随机选择所述点云模型中不共面的四个重构点；
    继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
    Reward＝-k·(PointNum-4)+g·VolRate，
    其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，
    其中，所述重构方法还包括如下步骤：基于客户需求调整g和k的比例来调整所述重构条件。
  2. 根据权利要求1所述的点云模型的重构方法，其特征在于，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构方法还包括：
    限定所述重构点的总数；
    继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。
  3. 根据权利要求1所述的点云模型的重构方法，其特征在于，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构方法还包括：
    定义所述第一阈值；
    继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。
  4. 根据权利要求1所述的点云模型的重构方法，其特征在于，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构方法还包括：
    以所述四个重构点在所述点云模型中限定其他重构点的选择范围；
    在所述选择范围中继续依次选择其他重构点，并分别计算不同重构 点组合的还原度，选出还原度最高的重构点组合。
  5. 根据权利要求1所述的点云模型的重构方法，其特征在于，所述重构方法基于以下规则执行：
    -在每个重构条件下执行的重构方法中，每个重构点只能选择一次；
    -随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少；
    -如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；
    -如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；
    -如果选择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
  6. 点云模型的重构装置，其中，包括：
    选择装置，其随机选择所述点云模型中不共面的四个重构点；
    重构装置，继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，并基于客户需求调整g和k的比例来调整所述重构条件，其中，重构所述点云模型的还原度为：
    Reward＝-k·(PointNum-4)+g·VolRate，
    其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例，k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重。
  7. 根据权利要求1所述的点云模型的重构装置，其特征在于，所述重构条件包括限定总数的重构点组合中还原度最高时，其中，所述重构装置还用于：
    限定所述重构点的总数；
    继续依次选择其他重构点，并分别计算该总数的不同重构点组合的还原度，选出还原度最高的重构点组合。
  8. 根据权利要求1所述的点云模型的重构装置，其特征在于，所述重构条件包括当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例达到第一阈值时，其中，所述重构装置还用于：
    定义所述第一阈值；
    继续依次选择其他重构点，并分别计算当前所有重构点围起来的立体图形的体积和所述点云模型原始体积的体积比例，直至达到所述第一阈值。
  9. 根据权利要求1所述的点云模型的重构装置，其特征在于，所述重构条件包括基于所述四个重构点限定其他重构点选择范围的重构点组合中还原度最高时，其中，所述重构装置还用于：
    以所述四个重构点在所述点云模型中限定其他重构点的选择范围；
    在所述选择范围中继续依次选择其他重构点，并分别计算不同重构点组合的还原度，选出还原度最高的重构点组合。
  10. 根据权利要求1所述的点云模型的重构装置，其特征在于，所述重构装置基于以下规则执行：
    -在每个重构条件下执行的重构方法中，每个重构点只能选择一次；
    -随着选择重构点的路径深度增加，在该路径中的每一层重构点的作为子节点的重构点在减少；
    -如果第一重构点和该第一重构点的前两个重构点共线，则无效该第一重构点及其以后的路径支路；
    -如果第二重构点和该第二重构点的前三个重构点共面，则无效该第二重构点及其以后的路径支路；
    -如果选择第三重构点时基于所有重构点组合的还原度较之前的还原度降低，则无效该第三重构点及其以后的路径支路。
  11. 点云模型的重构系统，包括：
    处理器；以及
    与所述处理器耦合的存储器，所述存储器具有存储于其中的指令，所述指令在被处理器执行时使所述电子设备执行动作，所述动作包括：
    随机选择所述点云模型中不共面的四个重构点；
    继续依次迭代选择其他重构点直至满足重构条件，并基于所有重构点重构该点云模型，其中，重构所述点云模型的还原度为：
    Reward＝-k·(PointNum-4)+g·VolRate，
    其中，PointNum表示当前所选择重构点的数量，VolRate表示当前所有重构点围起来的立体形状的体积和所述点云模型原始体积的体积比例， k表示所选点的数目在所述还原度中的比重，g表示所述体积比例在所述还原度中的比重，并且基于客户需求调整g和k的比例来调整所述重构条件。
  12. 计算机程序产品，所述计算机程序产品被有形地存储在计算机可读介质上并且包括计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据权利要求1至5中任一项所述的方法。
  13. 计算机可读介质，其上存储有计算机可执行指令，所述计算机可执行指令在被执行时使至少一个处理器执行根据权利要求1至5中任一项所述的方法。
PCT/CN2019/098910 2019-08-01 2019-08-01 重构点云模型的方法、装置和系统 Ceased WO2021016996A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP19915575.5A EP3796266A4 (en) 2019-08-01 2019-08-01 POINT CLOUD MODEL RECONSTRUCTION METHOD AND APPARATUS, AND SYSTEM
PCT/CN2019/098910 WO2021016996A1 (zh) 2019-08-01 2019-08-01 重构点云模型的方法、装置和系统
CN201980008746.8A CN112602121B (zh) 2019-08-01 2019-08-01 重构点云模型的方法、装置和系统
US16/970,691 US11410379B2 (en) 2019-08-01 2019-08-01 Point cloud model reconstruction method, apparatus, and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/098910 WO2021016996A1 (zh) 2019-08-01 2019-08-01 重构点云模型的方法、装置和系统

Publications (1)

Publication Number Publication Date
WO2021016996A1 true WO2021016996A1 (zh) 2021-02-04

Family

ID=74229945

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/098910 Ceased WO2021016996A1 (zh) 2019-08-01 2019-08-01 重构点云模型的方法、装置和系统

Country Status (4)

Country Link
US (1) US11410379B2 (zh)
EP (1) EP3796266A4 (zh)
CN (1) CN112602121B (zh)
WO (1) WO2021016996A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022188172A1 (en) * 2021-03-12 2022-09-15 Siemens Aktiengesellschaft Graph transformation method, apparatus and system of function block chain

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117853484B (zh) * 2024-03-05 2024-05-28 湖南建工交建宏特科技有限公司 一种基于视觉的桥梁损伤智能监测方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103646421A (zh) * 2013-12-13 2014-03-19 贾金原 基于增强型PyrLK光流法的树木轻量化3D重建方法
US20180165004A1 (en) * 2014-05-06 2018-06-14 Conceptualiz Inc. System and method for interactive 3d surgical planning and modelling of surgical implants
CN108759665A (zh) * 2018-05-25 2018-11-06 哈尔滨工业大学 一种基于坐标转换的空间目标三维重建精度分析方法
WO2019012539A1 (en) * 2017-07-13 2019-01-17 Stratasys Ltd. METHOD OF PRINTING A 3D MODEL FROM POINT CLOUD DATA
CN109472861A (zh) * 2018-12-03 2019-03-15 山东大学 交互式树木建模方法、模型生成方法、系统及仿生树木

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8525848B2 (en) * 2009-11-16 2013-09-03 Autodesk, Inc. Point cloud decimation engine
DE102011081987B4 (de) 2011-09-01 2014-05-28 Tomtec Imaging Systems Gmbh Verfahren zur Erzeugung eines Modells einer Oberfläche einer Hohlraumwand
CN104361625B (zh) 2014-07-24 2017-12-22 西北农林科技大学 一种基于射线原理的带边界保留的云数据精简算法
US9582939B2 (en) * 2015-06-11 2017-02-28 Nokia Technologies Oy Structure preserved point cloud simplification
WO2017113260A1 (zh) 2015-12-30 2017-07-06 中国科学院深圳先进技术研究院 一种三维点云模型重建方法及装置
GB2564402B (en) 2017-07-06 2021-03-17 Sony Interactive Entertainment Inc System and method of 3D print modelling

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103646421A (zh) * 2013-12-13 2014-03-19 贾金原 基于增强型PyrLK光流法的树木轻量化3D重建方法
US20180165004A1 (en) * 2014-05-06 2018-06-14 Conceptualiz Inc. System and method for interactive 3d surgical planning and modelling of surgical implants
WO2019012539A1 (en) * 2017-07-13 2019-01-17 Stratasys Ltd. METHOD OF PRINTING A 3D MODEL FROM POINT CLOUD DATA
CN108759665A (zh) * 2018-05-25 2018-11-06 哈尔滨工业大学 一种基于坐标转换的空间目标三维重建精度分析方法
CN109472861A (zh) * 2018-12-03 2019-03-15 山东大学 交互式树木建模方法、模型生成方法、系统及仿生树木

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3796266A4 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022188172A1 (en) * 2021-03-12 2022-09-15 Siemens Aktiengesellschaft Graph transformation method, apparatus and system of function block chain

Also Published As

Publication number Publication date
US11410379B2 (en) 2022-08-09
US20220148258A1 (en) 2022-05-12
EP3796266A4 (en) 2021-11-17
EP3796266A1 (en) 2021-03-24
CN112602121A (zh) 2021-04-02
CN112602121B (zh) 2022-06-24

Similar Documents

Publication Publication Date Title
JP7582592B2 (ja) データ生成方法、データ生成装置、およびプログラム
JP6892424B2 (ja) ハイパーパラメータチューニング方法、装置及びプログラム
CN113420880B (zh) 网络模型训练方法、装置、电子设备及可读存储介质
TW201235865A (en) Data structure for tiling and packetizing a sparse matrix
CN109656798B (zh) 基于顶点重排序的超级计算机大数据处理能力测试方法
CN111670463A (zh) 基于机器学习的几何网格简化
CN112602121B (zh) 重构点云模型的方法、装置和系统
US20200364538A1 (en) Method of performing, by electronic device, convolution operation at certain layer in neural network, and electronic device therefor
CN110704535B (zh) 数据分箱方法、装置、设备及计算机可读存储介质
CN114359447A (zh) 骨骼数据的建模方法、计算机设备及存储介质
CN104183021A (zh) 一种利用可移动空间网格精简点云数据的方法
JP2023123636A (ja) ハイパーパラメータチューニング方法、装置及びプログラム
CN111476872B (zh) 一种图像绘制方法及图像绘制装置
CN113681897B (zh) 切片处理方法、打印方法、系统、设备和存储介质
US20190347080A1 (en) Branch objects for dependent optimization problems
CN114522426A (zh) 游戏ai角色技能确定方法、装置、电子设备及存储介质
Kallis et al. Incremental 2D Delaunay triangulation core implementation on FPGA for surface reconstruction via high-level synthesis
Bouchard et al. On the mathematics and physics of high genus invariants of C^3/Z_3
WO2019191466A1 (en) Real-time spatial authoring in augmented reality using additive and subtractive modeling
CN114969917A (zh) 一种3d植被图的生成方法、装置、电子设备和存储介质
CN113610992A (zh) 骨骼驱动系数确定方法、装置、电子设备及可读存储介质
Chowdhury et al. Non-zero multi-valued decision diagram (NZMDD) based synthesis of multi-valued logic (MVL) functions
CN114237402B (zh) 一种虚拟现实的空间移动控制系统及方法
CN110188773A (zh) 特征提取方法、图像处理方法及装置
Siddiqui et al. Variable Ordering in Binary Decision Diagram using Spider Monkey Optimization for node and path length optimization

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2019915575

Country of ref document: EP

Effective date: 20200825

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19915575

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE