WO2020042970A1 - 一种三维建模的方法及其装置 - Google Patents

一种三维建模的方法及其装置 Download PDF

Info

Publication number
WO2020042970A1
WO2020042970A1 PCT/CN2019/101613 CN2019101613W WO2020042970A1 WO 2020042970 A1 WO2020042970 A1 WO 2020042970A1 CN 2019101613 W CN2019101613 W CN 2019101613W WO 2020042970 A1 WO2020042970 A1 WO 2020042970A1
Authority
WO
WIPO (PCT)
Prior art keywords
dimensional
image
projection
spatial
monocular camera
Prior art date
Application number
PCT/CN2019/101613
Other languages
English (en)
French (fr)
Inventor
杨伟樑
高志强
林杰勇
林清云
Original Assignee
广景视睿科技（深圳）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
Application filed by 广景视睿科技（深圳）有限公司 filed Critical 广景视睿科技（深圳）有限公司
Publication of WO2020042970A1 publication Critical patent/WO2020042970A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2200/00—Indexing scheme for image data processing or generation, in general
    • G06T2200/08—Indexing scheme for image data processing or generation, in general involving all processing steps from image acquisition to 3D model generation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20068—Projection on vertical or horizontal image axis

Definitions

  • the present application relates to the field of image processing, and in particular, to a method and a device for three-dimensional modeling.
  • the 3D modeling method based on a monocular camera does not have a high-quality modeling solution, and the error rate of the calculated point cloud model is extremely high.
  • the modeling steps need to be repeated for 3D reconstruction.
  • the 3D modeling method based on a monocular camera simply defines the layout of the indoor scene as a single cube primitive, and it does not perform color, position, Analysis of size, contour, etc. (ie analysis of object information).
  • the size, color, and angle of the virtual object cannot be changed according to the object information of the real object to make the combination of the virtual object image and the real object more consistent and realistic.
  • the embodiment of the present application provides a method and a device for three-dimensional modeling, which can quickly and accurately construct a three-dimensional model of a projection space, and can identify real objects in the projection space and calculate each real object from the three-dimensional model of the space Object information.
  • a method for three-dimensional modeling is provided, which is applied to a monocular camera, the monocular camera being connected to a multi-dimensional rotating motor, including:
  • a real object in the projection space is identified and object information of each of the real objects is calculated.
  • controlling the monocular camera to acquire an image of the projection space in a preset manner includes:
  • the method further includes:
  • the points of the corrected image are normalized by focal length division.
  • obtaining the projection space parameters of the projection space and determining a three-dimensional space model according to the image includes:
  • obtaining the spatial feature points and spatial lines of the projection space according to the image includes:
  • a spatial feature point and a spatial line of the projection space are acquired.
  • matching the spatial feature points with the spatial lines includes:
  • the grouping the matched spatial lines to determine a three-dimensional space model includes:
  • All image lines of the image are acquired, the image lines are integrated in the three-dimensional space model, and line segment consistency is used to estimate a three-dimensional indoor Manhattan scene.
  • identifying a real object in the projection space and calculating object information of each of the real objects includes:
  • the real object information includes an outline, a shape, a size, and a color of the object.
  • the real object information includes an outline of an object
  • the method further includes:
  • a three-dimensional modeling device which is applied to a monocular camera.
  • the monocular camera is connected to a multi-dimensional rotating motor, and is characterized by including:
  • An acquisition unit configured to control the monocular camera to acquire an image of a projection space in a preset manner
  • a determining unit configured to obtain a projection space parameter of the projection space according to the image, and determine a three-dimensional space model
  • a recognition and calculation unit configured to recognize a real object in the projection space and calculate object information of each of the real objects according to the three-dimensional space model and a preset image recognition algorithm of the monocular camera.
  • the embodiments of the present application provide a method and a device for three-dimensional modeling.
  • the three-dimensional modeling method is applied to a monocular camera, which is connected to a multi-dimensional rotating motor, and controls a monocular camera to collect an image of a projection space in a preset manner; according to the image, to obtain a projection space parameter of the projection space, And determine the three-dimensional space model; according to the three-dimensional space model and the preset image recognition algorithm of the monocular camera, identify the real objects in the projection space and calculate the object information of each real object.
  • the present application can quickly and accurately construct a spatial three-dimensional model of a projection space, and can recognize a real object in the projection space and calculate object information of each real object from the three-dimensional model of the space, so that it can be changed according to the object information of the real object.
  • the size, color, and angle of the virtual object make the combination of the virtual object image and the real object more consistent and more realistic.
  • FIG. 1 is a schematic structural diagram of a projection space according to an embodiment of the present application.
  • FIG. 2 is a schematic diagram of combining virtual reality in a projection space according to an embodiment of the present application
  • FIG. 3 is a flowchart of a three-dimensional modeling method according to an embodiment of the present application.
  • step S11 in FIG. 3 is a flowchart of a method in step S11 in FIG. 3;
  • FIG. 5 is a flowchart of a three-dimensional modeling method according to another embodiment of the present application.
  • FIG. 6 is a flowchart of a method in step S12 in FIG. 3;
  • FIG. 7 is a flowchart of the method in step S121 in FIG. 6;
  • FIG. 8 is a schematic diagram of an application scenario of step S1212 in FIG. 7 according to an embodiment of the present application.
  • FIG. 9 is a schematic diagram of a result of step S1212 in FIG. 7 provided by an embodiment of the present application.
  • FIG. 10 is a flowchart of the method in step S122 in FIG. 7;
  • FIG. 11 is a flowchart of the method in step S123 in FIG. 7;
  • FIG. 12 is a flowchart of the method in step S13 in FIG. 3;
  • step S13 is a flowchart of a method of step S13 according to another embodiment of the present application.
  • 15 is a schematic diagram of a device for three-dimensional modeling provided by an embodiment of the present application.
  • FIG. 16 is a schematic structural diagram of a smart terminal according to an embodiment of the present application.
  • FIG. 1 is a schematic structural diagram of a projection space according to an embodiment of the present application.
  • the projection space 100 includes a real object 10, a multi-dimensional rotating motor 20, a monocular camera 30, and a projection device 40 located in the projection space 100.
  • the projection space 100 refers to a range that the monocular camera 30 can cover under the driving of the multi-dimensional rotating motor 20, for example, a room, a conference room, an exhibition hall, etc., that is, the projection space
  • the size of 100 may be determined by the image acquisition range of the monocular camera 30. When the number of the monocular cameras 30 is multiple, in a broad sense, the size of the projection space 100 can be infinite.
  • the projection space 100 is spatially continuous, for example, in a room
  • the space is an extended open, semi-closed or closed space.
  • the projection space 100 may also be discontinuous, that is, the discretization of regions, for example, branch office buildings located in different regions, and the projected space image sets in different regions may be collected through mobile wired or wireless networks.
  • An image system is synthesized, and then a three-dimensional model is constructed of projected spatial images of multiple regions within the image system.
  • the real object 10 is an object located in the projection space 100, and may be a display object, such as a table shown in the figure, a vase and a cushion placed on the upper surface of the table, or a movable object, such as a A person walking in the projection space 100, an LED display screen moving in a predetermined direction, and the like are described.
  • the real object 10 has its own object information to distinguish it from other objects, for example, the coordinate position, size, outline, color, etc. of the real object 10 may also be based on the The object information changes the coordinate position, size, outline, color, etc. of the virtual object image projected by the projection device 40, so that the combination of the virtual object image and the real object 10 is more consistent and more realistic.
  • the multi-dimensional rotating motor 20 is connected to the monocular camera 30 and the projection device 40, respectively, so that the monocular camera 30 and the projection device 40 can be rotated around different axes around different angles, that is, the monocular can be changed.
  • the acquisition direction and the acquisition angle of the eyepiece camera 30 may be changed, and at the same time, the projection direction and the projection angle of the projection device 40 may be changed.
  • the monocular camera 30 solves the problem of photo-level accuracy that cannot be achieved by a depth camera.
  • at least one monocular camera 30 may be installed in the projection space 100 according to a certain rule. For example, in a museum hall, it is installed every 5 meters in the vertical or horizontal direction.
  • a monocular camera 30, which can be fixedly mounted on a roof, wall, floor, or surface of a real object. The combination of the multi-dimensional rotating motor 20 and the monocular camera 30 can maximize the collection range of the monocular camera 30, reduce the layout of the monocular camera 30, and thereby reduce the system cost.
  • an integrated monocular camera 30 may be selected instead of the combination of the multi-dimensional rotary motor 20 and the monocular camera 30, such as a hemispherical all-in-one machine, a fast dome-type all-in-one machine, an integrated machine incorporating a gimbal, or All-in-one camera with built-in lens, etc., the above-mentioned all-in-one camera can achieve automatic focusing.
  • a monocular camera 30 having a waterproof function, a small size, a high resolution, a long life, and a universal communication interface is selected.
  • the projection device 40 is connected to the multi-dimensional rotating motor 20. After completing the construction of the three-dimensional space model, the projection device 40 receives preset instructions (such as voice instructions, gesture instructions, keyboard commands, etc.) and sends them to the smart terminal ( A virtual object image 50 (as shown in FIG. 2) is projected onto the real object 10 under the control of a computer, etc., and finally a dynamic display combining virtual reality is realized, wherein the three-dimensional model of the space is usually stored in a computer or other large-scale Storage devices (such as servers).
  • preset instructions such as voice instructions, gesture instructions, keyboard commands, etc.
  • the process of combining virtual reality in the projection space 100 is as follows:
  • Collect at least one monocular camera 30 and multi-dimensional rotating motor 20 images of a table located in the projection space 100, an empty vase on the table, and a cushion to construct a three-dimensional model of the space.
  • the table located in the projection space 100 Images of vacant vases and cushions on the table may be part of the projection space 100, and the three-dimensional model of the space also includes characteristic parameters of images of the table, vacant vases, and cushions on the table.
  • the characteristic parameters of the vase on the table and the cushion on the table (including the dimensions, spatial coordinates, shape and color characteristics of the vase and cushion), determine the characteristic parameters of a bunch of flowers according to the characteristic parameters of the vase on the table, and The characteristic parameters determine the characteristic parameters of a cat, match a picture of a bouquet of flowers according to the characteristic parameters of a bouquet of flowers, match the picture of a cat according to the characteristic parameters of a cat, and compare the picture of a bouquet of flowers and a cat's Pictures are image processed.
  • the projection direction of the projection device 40 is adjusted, and a picture of a bouquet of flowers and a picture of a cat are projected onto the vase and the table on the table, respectively.
  • the upper cushion makes the combination of virtual reality more consistent and more realistic.
  • FIG. 3 is a flowchart of a three-dimensional modeling method according to an embodiment of the present application.
  • the three-dimensional modeling method is applied to a monocular camera 30, and the monocular camera 30 is connected to a multi-dimensional rotating motor 20.
  • the three-dimensional modeling method includes:
  • Step S11 controlling the monocular camera 30 to capture an image of the projection space 100 in a preset manner.
  • three-dimensional scene information provides more possibilities for various computer vision applications such as image segmentation, object detection, and object tracking, and images have been widely used as a general way of expressing three-dimensional scene information.
  • the gray value of each pixel point in the image can be used to represent the distance of a certain point in the scene from the monocular camera 30.
  • passive ranging sensing there are two methods for acquiring images: passive ranging sensing and active sensing.
  • the most commonly used method in the passive ranging sensing is a binocular stereo frustum.
  • the monocular camera 30 obtains two images in the same scene at the same time, finds the corresponding pixel points in the two images through the stereo matching algorithm, and then calculates the time difference information according to the triangle principle, and the time difference information can be used to represent the real object in the scene through conversion. 10 messages.
  • active sensing the most obvious feature of active sensing is that the device itself needs to emit energy to complete the collection of information, which ensures that the acquisition of images is independent of the acquisition of color images.
  • step S11 specifically includes:
  • Step S111 Calibrate the multi-dimensional rotary electric machine 20 so that the multi-dimensional rotary electric machine 20 is at a starting point.
  • the multi-dimensional rotary electric machine 20 can be a servo motor. Before each image is turned on, the initial state of the multi-dimensional rotary electric machine 20 is adjusted to the starting point of the multi-dimensional rotary electric machine 20, so that the multi-dimensional rotary electric machine 20 that starts working at different times can collect data. The data is comparable, so that the multi-dimensional rotary electric machine 20 that starts working at the same time has synchronization.
  • Step S112 Calculate the number of images to be collected in the horizontal and vertical directions of the monocular camera 30 according to the preset number of shots, horizontal viewing angle, and vertical viewing angle of each three-dimensional area of the monocular camera 30. And the interval angle between every two images.
  • the preset number of shots for each three-dimensional area is n
  • the horizontal viewing angle is (HFOV)
  • the vertical viewing angle is (VFOV).
  • Nh 360.
  • the position of the monocular camera 30 is unchanged, and the dimensions and angles of the collection can be changed by adjusting the multi-dimensional rotating motor 20 connected to it.
  • the three-dimensional area is the camera field of view of the monocular camera 30.
  • the camera field of view is related to the angle of view.
  • the angle of view refers to the angle that a human, animal, or lens can be involved in, shooting, and seeing.
  • the angle formed by the center point of the lens to the diagonal ends of the imaging plane is the lens angle of view. For the same imaging area, the shorter the focal length of the lens, the larger the angle of view.
  • the angle of view mainly refers to the range of angles that it can achieve.
  • the focal length becomes shorter, the angle of view becomes larger, and a wider range can be shot, but this will affect the sharpness of the distant subject.
  • the focal length becomes longer, the angle of view becomes smaller, and distant objects can be made clear, but the range of widths that can be taken becomes narrower.
  • FIG. 5 is a flowchart of a three-dimensional modeling method according to another embodiment of the present application. As shown in FIG. 5, the method further includes:
  • Step S14 Obtain light distortion parameters and offset parameters of the monocular camera 30.
  • the monocular camera 30 projected by omnidirectional light makes the image of the real object 10 appear distorted, and the offset parameter refers to the displacement in the two images, so that the overlapping area of the two images can be Completely coincide.
  • Step S15 Correct the image according to the light distortion parameter and the offset parameter.
  • Step S16 Normalize the points of the image after correction by focal length division.
  • the so-called normalization is to convert the original image to be processed into a corresponding unique standard form through a series of changes.
  • the standard form image has invariant characteristics for affine transformations such as translation, rotation, and scaling.
  • Step S12 Obtain a projection space parameter of the projection space 100 according to the image, and determine a three-dimensional space model.
  • step S12 specifically includes:
  • Step S121 Obtain a spatial feature point and a spatial line of the projection space 100 according to the image.
  • Step S121 specifically includes:
  • Step S1211 Calculate the relative three-dimensional coordinates between the two pairs of image lines through the conversion relationship between any two of the images.
  • Step S1212 Acquire a spatial feature point and a spatial line of the projection space according to the three-dimensional coordinates.
  • the spatial feature points and spatial lines of the telephone (ie, the real object 10) in FIG. 8 can be obtained (as shown in FIG. 9).
  • the camera attitude of the monocular camera 30 is R
  • the calibration matrix of the monocular camera 30 is C
  • the two-dimensional coordinates of the spatial lines in the acquired image are [u, v].
  • Step S122 specifically includes:
  • Step S1221 The image is applied to the spatial line based on the stereoscopic vertebra where it is located.
  • the vertebral body is a three-dimensional body whose position is related to the monocular camera 30.
  • the shape of the vertebral body determines how the model is projected from the camera space onto the screen.
  • the most common type of projection is perspective projection, so that the The real object 10 near the eye camera 30 is larger after projection, and the object farther from the monocular camera 30 is smaller after projection.
  • the perspective projection uses a pyramid as the viewing cone.
  • the monocular camera 30 is located at the apex of the pyramid.
  • the pyramid is truncated by two front and rear planes to form a pyramid.
  • Step S1222 Match the spatial feature points and the spatial lines of the projection space 100 according to any two overlapping stereoscopic vertebrae of the image.
  • Step S123 group the matching spatial lines to determine the three-dimensional space model.
  • the three-dimensional coordinates of the projected spatial lines of the two images in the overlapping area are defined as L1 and L2.
  • the camera attitude compensation of the two images of L1 and L2 is obtained to obtain the space.
  • L1i and L2i are matched in sequence, and L1i and L2i with coverage ratios reaching a preset ratio (for example, more than 60%) will be defined as the same spatial lines in space, there will be no overlapping or spatial lines with coverage lower than the preset ratio L1i or L2i is deleted.
  • the matched spatial lines are grouped, and then the layout that can be converted into a physically reasonable 3D model is searched.
  • Step S123 specifically includes:
  • Step S1231 Converge the spatial lines according to parallelism or orthogonality to obtain orthogonal vanishing points.
  • Step S1232 Determine a boundary of the projection space 100 according to the spatial line and the orthogonal vanishing point, and determine a three-dimensional space model.
  • the underground boundary and ceiling boundary of the wall are determined according to the detected spatial lines and orthogonal vanishing points.
  • the end of the end of the spatial line connecting another spatial line is defined as the connecting line of different walls.
  • Parallel lines of space can be thought of as lying on the same plane.
  • Many physically valid layout hypotheses are generated from the order of spatial lines, and each generated hypothesis will get the score of the best matching spatial line.
  • the best spatial line combination corresponding to the best hypothesis is converted into a three-dimensional spatial model.
  • Step S1233 Obtain all the image lines of the image, integrate the image lines in the three-dimensional space model, and estimate the three-dimensional indoor Manhattan scene using line segment consistency.
  • Step S13 According to the three-dimensional space model and a preset image recognition algorithm of the monocular camera, identify a real object in the projection space and calculate object information of each of the real objects.
  • step S13 specifically includes:
  • Step S131 Identify the real object 10 in the projection space 100 and the three-dimensional coordinates of the position of the real object 10 according to the three-dimensional space model.
  • Step S132 Calculate information of each of the real objects 10 according to a preset image recognition algorithm of the monocular camera 30.
  • the real object 10 information includes an outline, a shape, a size, and a color of the object.
  • the panoramic view of the projection space 100 uses the starting point of the multi-dimensional motor 20 as an origin.
  • the identification of the real object 10 uses an indoor object data set and a learning algorithm to identify the position where the real object 10 appears. Convert the position (two-dimensional coordinates) of the real object 10 in the image into three-dimensional coordinates in the projection space 100, and store the position information (that is, the three-dimensional coordinates) and the real object information in a database for other applications to query and For updating the three-dimensional space model.
  • step S13 specifically includes:
  • Step S133 Recognize each real object 10 from the image according to the preset image recognition algorithm, and extract an object contour of each real object 10.
  • the contour of the object can be obtained by edge detection, and the method of edge detection includes a gradient operator, a Robert operator, a Sauber operator, a second-order micro operator, and the like.
  • Step S134 identify whether there is occlusion in each of the real objects 10.
  • Step S135 If it exists, perform occlusion compensation on the outline of the occluded real object 10 according to a preset compensation algorithm.
  • the present application synchronizes the construction of the three-dimensional space model during the process of collecting images by the monocular camera 30.
  • the three-dimensional space model of the projection space 100 can be obtained, avoiding the need to repeat the inaccurate three-dimensional model Constructing the three-dimensional space model.
  • FIG. 14 is a flowchart of a three-dimensional modeling method according to another embodiment of the present application. As shown in FIG. 14, the method further includes:
  • Step S17 Receive a virtual reality projection instruction.
  • the virtual reality projection instruction is a voice instruction
  • a virtual object and a real object combined with the virtual object are determined by a voice recognition technology.
  • the voice recognition technology involves signal processing, pattern recognition, probability theory and information theory, and sound generation. Mechanism and auditory mechanism, artificial intelligence and other fields.
  • the voice recognition can be divided into desktop voice recognition, telephone voice recognition, and embedded voice recognition.
  • determining the virtual object to be cast and the real object 10 to be projected and combined by the virtual object are only one of the methods disclosed in the embodiments of the present application through voice instructions.
  • Touch control keyboard (Including physical keyboard and soft keyboard) input, gesture control, virtual space instruction input and other methods.
  • Step S18 Determine a projection position and a projection object according to the virtual reality projection instruction.
  • the projection position is three-dimensional coordinates of a certain position of a three-dimensional space model corresponding to the projection space 100.
  • the projection object should be adapted to the characteristic parameters of the real object 10 in terms of position coordinates, size, and color characteristics, so that the image projected by the projection object and the real object 10 look comfortable, coordinated, and real.
  • the projection object is a virtual object in a virtual object picture.
  • the virtual object picture may be derived from a local database or the Internet. A preliminary screening is performed on the matching of the characteristic parameters, and the local database or the Internet is searched for according to the virtual object determined in the voice command. There are theoretically the best matching picture resources.
  • Step S19 Control the projection device 40 to project the projection object to the projection position.
  • the real object 10 is located in the projection space 100 and is an arbitrary one or more.
  • the selected number of selected projection objects and the real object 10 are the same, and the number may be different.
  • a virtual knife and fork picture is projected next to a plate in the projection space 100, and a virtual art photo is projected into the projection space 100.
  • a virtual object picture can also be projected on the same real object 10 in real space, for example, a virtual blue and white porcelain picture is projected on the vase body, a virtual bouquet is projected on the vase, or Two virtual fruit pictures are projected onto the fruit plate of the projection space 100.
  • the embodiment of the present application provides a method for three-dimensional modeling.
  • the three-dimensional modeling method is applied to a monocular camera, which is connected to a multi-dimensional rotating motor, and controls a monocular camera to collect an image of a projection space in a preset manner; according to the image, to obtain a projection space parameter of the projection space, And determine the three-dimensional space model; according to the three-dimensional space model and the preset image recognition algorithm of the monocular camera, identify the real objects in the projection space and calculate the object information of each real object.
  • the present application can quickly and accurately construct a spatial three-dimensional model of a projection space, and can recognize a real object in the projection space and calculate object information of each real object from the three-dimensional model of the space, so that it can be changed according to the object information of the real object.
  • the size, color, and angle of the virtual object make the combination of the virtual object image and the real object more consistent and more realistic.
  • FIG. 15 is a schematic diagram of a three-dimensional modeling apparatus according to an embodiment of the present application.
  • the three-dimensional modeling device 200 is applied to a monocular camera, and the monocular camera is connected to a multi-dimensional rotating motor.
  • the three-dimensional modeling device 200 includes:
  • An acquisition unit 201 is configured to control the monocular camera to acquire an image of a projection space in a preset manner.
  • the acquisition unit 201 is specifically configured to calibrate the multi-dimensional rotary motor, so that the multi-dimensional rotary motor is at a starting point; according to the preset number of shots and level of each three-dimensional area of the monocular camera For the viewing angle and the vertical viewing angle, the number of the images to be acquired in the horizontal and vertical directions of the monocular camera and the interval angle between each two images are calculated respectively.
  • a determining unit 202 is configured to obtain a projection space parameter of the projection space according to the image, and determine a three-dimensional space model.
  • the determining unit 202 is specifically configured to obtain a spatial feature point and a spatial line of the projection space according to the image; match the spatial feature point and the spatial line; Grouping the space lines to determine the three-dimensional space model.
  • acquiring the spatial feature points and the spatial lines of the projection space according to the image includes: calculating a relative three-dimensional coordinate between the pair of image lines by using a transformation relationship between any two of the images; The three-dimensional coordinates are used to obtain spatial feature points and spatial lines of the projection space.
  • Matching the spatial feature points with the spatial lines includes: paving the spatial lines based on the stereoscopic vertebra where the image is located; performing the mapping based on the stereoscopic vertebrae of any two overlapping images. The matching of spatial feature points and spatial lines of the projection space is described.
  • the grouping the matched spatial lines to determine the three-dimensional spatial model includes: converging the spatial lines according to parallelism or orthogonality to obtain an orthogonal vanishing point; according to the spatial lines and the The orthogonal vanishing point determines the boundary of the projection space and determines the three-dimensional space model; obtains all the image lines of the image, integrates the image lines in the three-dimensional space model, and estimates the three-dimensional indoor Manhattan using line segment consistency Scenes.
  • the recognition and calculation unit 203 is configured to recognize a real object in the projection space and calculate object information of each of the real objects according to the three-dimensional space model and a preset image recognition algorithm of the monocular camera.
  • the identification and calculation unit 203 is specifically configured to identify a real object in the projection space and a three-dimensional coordinate of the position of the real object according to the three-dimensional space model; according to the monocular camera
  • the preset image recognition algorithm calculates each of the real object information, and the real object information includes an outline, a shape, a size, and a color of the object.
  • each real object is identified from the image, and the object outline of each of the real object is extracted; whether or not each of the real object exists is identified Occlusion; if present, occlusion compensation is performed on the object contour of the real object that is occluded according to a preset compensation algorithm.
  • the units of the above device cooperate with each other.
  • the acquisition unit 201 controls the monocular camera to acquire an image of a projection space in a preset manner.
  • the projection space parameter of the projection space and a three-dimensional space model are determined.
  • the recognition and calculation unit 203 recognizes a real object in the projection space and calculates the calculation based on the three-dimensional space model and a preset image recognition algorithm of the monocular camera. Object information of each of the real objects.
  • the device embodiment and the method embodiment are based on the same concept, as long as the contents do not conflict with each other, the content of the device embodiment may refer to the method embodiment, and details are not described herein.
  • An embodiment of the present application provides a device for three-dimensional modeling.
  • the three-dimensional modeling device is applied to a monocular camera, which is connected to a multi-dimensional rotating motor, and controls the monocular camera to acquire an image of a projection space in a preset manner; according to the image, to obtain a projection space parameter of the projection space, And determine the three-dimensional space model; according to the three-dimensional space model and the preset image recognition algorithm of the monocular camera, identify the real objects in the projection space and calculate the object information of each real object.
  • the present application can quickly and accurately construct a spatial three-dimensional model of a projection space, and can recognize a real object in the projection space and calculate object information of each real object from the three-dimensional model of the space, so that it can be changed according to the object information of the real object.
  • the size, color, and angle of the virtual object make the combination of the virtual object image and the real object more consistent and more realistic.
  • FIG. 16 is a schematic structural diagram of a smart terminal according to an embodiment of the present application.
  • the smart terminal 300 includes one or more processors 301 and a memory 302. Among them, one processor 301 is taken as an example in FIG. 16.
  • the processor 301 and the memory 302 may be connected through a bus or in other manners. In FIG. 16, the connection through the bus is taken as an example.
  • the memory 302 is a non-volatile computer-readable storage medium, and may be used to store non-volatile software programs, non-volatile computer executable programs, and modules, as corresponding to the three-dimensional modeling method in the embodiment of the present application.
  • Program instructions / modules for example, the acquisition unit 201, the determination unit 202, and the identification and calculation unit 203 shown in FIG. 15).
  • the processor 301 executes various functional applications and data processing of the three-dimensional modeling device by running non-volatile software programs, instructions, and modules stored in the memory 302, that is, a method for implementing the three-dimensional modeling of the foregoing method embodiment And the functions of the various modules and units of the above device embodiments.
  • the memory 302 may include a storage program area and a storage data area, where the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the three-dimensionally modeled device, and the like.
  • the memory 302 may include a high-speed random access memory, and may further include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage device.
  • the memory 302 may optionally include a memory remotely set relative to the processor 301, and these remote memories may be connected to the processor 301 through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • the program instructions / modules are stored in the memory 302, and when executed by the one or more processors 301, perform the method of three-dimensional modeling in any of the above method embodiments, for example, execute FIG. 3 described above.
  • the method includes steps S11 to S13; the functions of each module or unit described in FIG. 15 may also be implemented.
  • the embodiments of the present application further provide a non-volatile computer-readable storage medium.
  • the non-volatile computer-readable storage medium stores executable instructions of the smart terminal, which are used to cause the smart terminal to execute the method of the three-dimensional modeling of the foregoing embodiment, so as to achieve a space capable of quickly and accurately constructing a projection space.
  • a three-dimensional model, and the real objects in the projection space can be identified from the three-dimensional model of the space and the object information of each real object can be calculated, so that the size, color, and angle of the virtual object can be changed according to the object information of the real object to make the virtual object image
  • the combination with real objects is more consistent and more realistic.
  • the above product can execute the method provided in the embodiment of the present application, and has the corresponding functional modules and beneficial effects of executing the method.
  • the above product can execute the method provided in the embodiment of the present application, and has the corresponding functional modules and beneficial effects of executing the method.

Landscapes

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

Abstract

本申请涉及图像处理领域，公开了一种三维建模的方法及其装置。所述三维建模的方法应用于单目相机，所述单目相机与多维旋转电机连接，通过控制单目相机按预设方式采集投影空间的图像；根据图像，获取投影空间的投影空间参数，并确定三维空间模型；根据三维空间模型和单目相机的预设图像识别算法，识别投影空间中的现实物体以及计算各现实物体的物体信息。本申请能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息，从而可以根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。

Description

一种三维建模的方法及其装置
本申请要求于2018年08月29日提交中国专利局，申请号为201810996321X，发明名称为“一种三维建模的方法及其装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图像处理领域，特别是涉及一种三维建模的方法及其装置。
背景技术
随着科学技术的发展和人民生活水平的不断提高，人们对视觉感受方面的要求越来越高。在显示效果上，人们又倾向于追求增强现实、身临其境的视觉享受。但是，在实际的动态增强现实投影应用中，若虚拟对象和现实物体的空间匹配度存在误差，由于人眼对于这种误差非常的敏感，这种误差会让用户产生虚拟现实结合不协调、不逼真的感觉，使得用户体验较差。因此，准确快速的对投影空间进行三维建模，并基于投影空间的三维空间模型获取投影空间中的现实物体的物体信息(包括现实物体的位置、轮廓信息等)是虚拟现实结合的基础，直接影响虚拟现实结合的真实感、实时性和交互性。
目前，基于单目相机的三维建模方法没有优质的建模方案，所计算出的点云模型的误差率极高，在出现错误时，还需要重复建模的步骤进行三维重构。同时，基于单目相机的三维建模方法只是简单的将室内场景布局定义为单个立方体基元，也并没有基于所建立的投影空间的三维空间模型对投影空间内的现实物体进行颜色、位置、大小、轮廓等分析(即物体信息的分析)。对要求较高的虚拟现实结合动态投影，无法根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。
发明内容
本申请实施例提供了一种三维建模的方法及其装置，其能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息。
为解决上述技术问题，本申请实施例采用的以下技术方案：
在第一方面，提供一种三维建模的方法，应用于单目相机，所述单目相机与多维旋转电机连接，包括：
控制所述单目相机按预设方式采集投影空间的图像；根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型；
根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
进一步的，所述控制所述单目相机按预设方式采集投影空间的图像包括：
校准所述多维旋转电机，使所述多维旋转电机处于起始点；
根据所述单目相机的每一个三维区域预设的拍照次数、水平视角和垂直视角，分别计算所述单目相机水平方向和垂直方向上需采集的所述图像的张数和每两张图像的间隔角度。
进一步的，所述方法还包括：
获取所述单目相机的光线扭曲参数及偏移量参数；
根据所述光线扭曲参数和偏移量参数，校正所述图像；
通过焦距除法来对校正后的所述图像的点进行归一化。
进一步的，所述根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型包括：
根据所述图像，获取所述投影空间的空间特征点和空间线条；
将所述空间特征点和所述空间线条进行匹配；
将匹配后的所述空间线条进行分组，确定所述三维空间模型。
进一步的，所述根据所述图像，获取所述投影空间的空间特征点和空间线条包括：
通过任意两个所述图像间的转换关系，计算图像线条两两之间相对 的三维坐标；
根据所述三维坐标，获取所述投影空间的空间特征点和空间线条。
进一步的，所述将所述空间特征点和所述空间线条进行匹配包括：
所述图像基于其所在的立体视椎铺贴所述空间线条；
根据任意两个重叠的所述图像的立体视椎，进行所述投影空间的空间特征点和空间线条的匹配。
进一步的，所述将匹配后的所述空间线条进行分组，确定三维空间模型包括：
将所述空间线条根据并行性或正交性进行收敛，得到正交消失点；
根据所述空间线条和所述正交消失点确定所述投影空间的边界，确定三维空间模型；
获取所述图像的所有图像线条，将所述图像线条整合在所述三维空间模型里，并使用线段一致性估测三维室内曼哈顿场景。
进一步的，所述根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息包括：
根据所述三维空间模型识别所述投影空间中的现实物体以及所述现实物体所处位置的三维坐标；
根据所述单目相机的预设图像识别算法计算各所述现实物体信息，所述现实物体信息包括物体的轮廓、形状、大小和颜色。
进一步的，当所述现实物体信息包括物体的轮廓时，
根据所述预设图像识别算法，从所述图像识别出各个现实物体，并且提取各所述现实物体的物体轮廓；
识别各所述现实物体是否存在遮挡；
若存在，则根据预设补偿算法，对存在遮挡的现实物体的物体轮廓进行遮挡补偿。
进一步的，所述方法还包括：
接收虚拟现实投影指令；
根据所述虚拟现实投影指令，确定投影位置以及投影物体；
控制投影设备向所述投影位置投影所述投影物体。
在第二方面，提供一种三维建模的装置，应用于单目相机，所述单目相机与多维旋转电机连接，其特征在于，包括：
采集单元，用于控制所述单目相机按预设方式采集投影空间的图像；
确定单元，用于根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型；
识别与计算单元，用于根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
本申请实施方式的有益效果是：区别于现有技术的情况，本申请实施例提供一种三维建模的方法及其装置。所述三维建模的方法应用于单目相机，所述单目相机与多维旋转电机连接，通过控制单目相机按预设方式采集投影空间的图像；根据图像，获取投影空间的投影空间参数，并确定三维空间模型；根据三维空间模型和单目相机的预设图像识别算法，识别投影空间中的现实物体以及计算各现实物体的物体信息。因此，本申请能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息，从而可以根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。
附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定，附图中具有相同参考数字标号的元件表示为类似的元件，除非有特别申明，附图中的图不构成比例限制。
图1是本申请实施例提供的一种投影空间的结构示意图；
图2是本申请实施例提供的一种投影空间中虚拟现实结合的示意图；
图3是本申请实施例提供的一种三维建模的方法流程图；
图4是图3中步骤S11的方法流程图；
图5是本申请另一实施例提供的一种三维建模的方法流程图；
图6是图3中步骤S12的方法流程图；
图7是图6中步骤S121的方法流程图；
图8是本申请实施例提供的图7中步骤S1212的应用场景示意图；
图9是本申请实施例提供的图7中步骤S1212的结果示意图；
图10是图7中步骤S122的方法流程图；
图11是图7中步骤S123的方法流程图；
图12是图3中步骤S13的方法流程图；
图13是本申请另一实施例提供的步骤S13的方法流程图；
图14是本申请又一实施例提供的一种三维建模的方法流程图；
图15是本申请实施例提供的一种三维建模的装置示意图；
图16是本申请实施例提供的一种智能终端的结构示意图。
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。
此外，下面所描述的本申请各个实施方式中所涉及到的技术特征只要彼此之间未构成冲突就可以相互组合。
请参阅图1，图1为本申请实施例提供的一种投影空间的结构示意图。如图1所示，所述投影空间100包括位于所述投影空间100内的现实物体10、多维旋转电机20、单目相机30和投影设备40。
可以理解，所述投影空间100指的是所述单目相机30在所述多维旋转电机20的驱动下可以覆盖的范围，例如，一个房间、一个会议室、一个展厅等，即所述投影空间100的大小可以由所述单目相机30的图像采集范围确定。当所述单目相机30的数量为多个时，广义上，所述 投影空间100的大小可以做到无限大，此时，所述投影空间100在空间上是连续的，例如，一个房间内的空间是延展的开放、半封闭或封闭空间。在一些实施例中，所述投影空间100还可以是断续的，即区域的离散化，例如，位于不同地区的分公司大楼建筑，通过移动有线或无线网络可以将不同区域的投影空间图像集合成一个图像***，然后，对该图像***内的多个区域的投影空间图像构建三维模型。
所述现实物体10为位于所述投影空间100中的物品，其可以是摆设物件，例如图中所示的桌子、放置于桌子上表面的花瓶和坐垫，还可以是活动的物体，例如在所述投影空间100中走动的人、沿预设方向运动的LED显示屏等。需要说明的是，所述现实物体10具有标识其自身的物体信息，用以区别于其他物件，比如，所述现实物体10的坐标位置，尺寸大小、轮廓线条、颜色等，还可以根据所述物体信息改变所述投影设备40投影的虚拟物体图像坐标位置，尺寸大小、轮廓线条、颜色等，使虚拟物体图像与所述现实物体10的结合更吻合、更逼真。
所述多维旋转电机20分别与所述单目相机30和所述投影设备40连接，可以使得所述单目相机30和投影设备40分别沿着不同的角度绕轴旋转，即可以改变所述单目相机30的采集方向和采集角度，同时，也可以改变所述投影设备40的投影方向和投影角度。
所述单目相机30解决了深度摄像机无法达到的照片级精度的问题。在本申请实施例中，至少一台所述单目相机30可以按照一定的规则安装在所述投影空间100，比如，在一个博物馆的大厅内，在竖直或水平方向上每隔5米安装一个单目相机30，所述单目相机30可以固定安装在屋顶、墙壁、地面或现实物体的表面上。通过多维旋转电机20与单目相机30结合可以最大化地增加单目相机30的采集范围，减少单目相机30的布设，进而减少***成本。
在一些实施例中，可以选择一体化的单目相机30替代多维旋转电机20与单目相机30结合的方式，比如，半球形一体机、快速球型一体机、结合云台的一体化机或镜头内置于云台的一体机等，上述的一体机可以实现自动聚焦。优选的，选择具有防水功能、体积较小、分辨率高、 高寿命以及具有通用通信接口等的单目相机30。
所述投影设备40与所述多维旋转电机20连接，在完成三维空间模型的构建之后，所述投影设备40接收预设的指令(比如语音指令、手势指令、键盘命令等)，在智能终端(比如计算机等)的控制下向所述现实物体10投影虚拟物体图像50(如图2所示)，最终实现虚拟现实结合的动态显示，其中，所述空间三维模型通常存储在计算机或其他大型的存储设备中(比如服务器等)。
如图2所示，在本申请实施例中，所述投影空间100中虚拟现实结合的过程具体如下：
通过至少一台所述单目相机30与多维旋转电机20采集位于投影空间100的桌子、位于桌子上的空置的花瓶以及坐垫的图像，构建空间三维模型，可以理解，位于投影空间100的桌子、位于桌子上的空置的花瓶以及坐垫的图像可能是投影空间100的一部分，所述空间三维模型还包括桌子、位于桌子上的空置的花瓶以及坐垫的图像的特征参数。
获取用户语音指令“搜索一束鲜花放置于花瓶内，搜索一只猫咪放置于坐垫上”，并对该语音指令进行识别，最终确认所需被投的虚拟物体分别为一束鲜花和一只猫咪，虚拟物体所要投影结合的现实物体10分别为桌子上的花瓶和桌子上的坐垫。
由于在同一投影空间100内的同一种类的现实物体10可能存在多个，比如，同一投影空间100内可能存在多张桌子，可以选择参照物进行区分，比如“请选择桌面上放置有化妆品的桌子”、“请选择旁边只有一张凳子的桌子”；可以通过物品本身的特性进行区分，比如，“请选择具有三条腿的桌子”、“请选择红色的桌子”、“请选择尺寸最小的桌子”，可以通过相对位置的方式进行区分，比如，“请选择单目相机30正前方60度方向的桌子”，或通过其他方式进行区分。
获取桌子上的花瓶和桌子上的坐垫的特征参数(包括花瓶和坐垫的尺寸、空间坐标、形状和颜色特性等)，根据桌子上的花瓶的特征参数确定一束鲜花的特征参数，根据坐垫的特征参数确定一只猫咪的特征参数，根据一束鲜花的特征参数匹配一束鲜花的图片，根据一只猫咪的特 征参数匹配一只猫咪的图片，并对一束鲜花的图片和一只猫咪的图片进行图像处理。最终，根据投影设备40与桌子上的花瓶和桌子上的坐垫的位置关系调整所述投影设备40的投影方向，将一束鲜花的图片和一只猫咪的图片分别投影到桌子上的花瓶和桌子上的坐垫，使虚拟现实的结合更吻合、更逼真。
请参阅图3，图3为本申请实施例提供的一种三维建模的方法流程图。如图3所示，所述三维建模的方法应用于单目相机30，所述单目相机30与多维旋转电机20连接，所述三维建模的方法包括：
步骤S11：控制所述单目相机30按预设方式采集投影空间100的图像。
在计算机视觉***中，三维场景信息为图像分割、目标检测、物体跟踪等各类计算机视觉应用提供了更多的可能性，而图像作为一种普遍的三维场景信息表达方式得到了广泛的应用。所述图像中的每一个像素点的灰度值可用于表征场景中某一点距离所述单目相机30的远近。
一般的，获取图像的方式包括被动测距传感和主动传感两种，其中，所述被动测距传感中最常采用的方法是双目立体视锥，该方法通过两个相隔一定具体的单目相机30同时获取同一场景中的两幅图像，通过立体匹配算法找到两幅图像中对应的像素点，随后根据三角形原理计算出时差信息，而时差信息通过转换可用于表征场景中现实物体10的信息。主动传感相较于被动测距传感最明显的特征是设备本身需要发射能量来完成信息的采集，保证了图像的获取独立于彩色图像的获取。
如图4所示，步骤S11具体包括：
步骤S111：校准所述多维旋转电机20，使所述多维旋转电机20处于起始点。
所述多维旋转电机20可以采用伺服电机，在每一次的开启采集图像前，调整所述多维旋转电机20的初始状态为多维旋转电机20的起始点，使得不同时间开启工作的多维旋转电机20采集的数据具有可比性，使得相同时间开始工作的多维旋转电机20具有同步性。
步骤S112：根据所述单目相机30的每一个三维区域预设的拍照次 数、水平视角和垂直视角，分别计算所述单目相机30水平方向和垂直方向上需采集的所述图像的张数和每两张图像的间隔角度。
假设每一个三维区域预设的拍照次数为n，水平视角为(HFOV)，垂直视角为(VFOV)，那么，所述单目相机30水平方向上需采集的所述图像的张数Nh＝360°/HFOV*n，所述单目相机30水平方向上每两张图像的间隔角度Dh＝360°/Nh，所述单目相机30垂直方向上需采集的所述图像的张数Nh＝270°/VFOV*n，由于常规条件下，所述单目相机30采集到其正下方的区域，故而，在所述单目相机30垂直方向上采用270°进行计算，所述单目相机30垂直方向上每两张图像的间隔角度Dv＝360°/Nv。
在本申请实施例中，所述单目相机30的位置是不变的，通过调节与之相连的多维旋转电机20可以改变采集的维度和角度。所述三维区域即所述单目相机30的摄像机视野，所述摄像机视野与视角有关，视角是指人类、动物或者镜头所能涉及、拍摄、看到的角度，视度愈窄(角度愈小)在不动眼睛的情况下看到的视野或景物愈小，也就代表小的景物也能看到清楚，视角宽，则相反。镜头中心点到成像平面对角线两端所形成的夹角就是镜头视角，对于相同的成像面积，镜头焦距越短，其视角就越大。对于镜头来说，视角主要是指它可以实现的视角范围，当焦距变短时视角就变大了，可以拍出更宽的范围，但这样会影响较远拍摄对象的清晰度。当焦距变长时，视角就变小了，可以使较远的物体变得清晰，但是能够拍摄的宽度范围就变窄了。
请参阅图5，图5为本申请另一实施例提供的一种三维建模的方法流程图。如图5所示，所述方法还包括：
步骤S14：获取所述单目相机30的光线扭曲参数及偏移量参数。
可以理解，全方位光线投射的所述单目相机30使得所述现实物体10的图像呈现扭曲现象，所述偏移量参数指的是两幅图像中的位移，使得两幅图像的重叠区域能够完全重合。
步骤S15：根据所述光线扭曲参数和偏移量参数，校正所述图像。
步骤S16：通过焦距除法来对校正后的所述图像的点进行归一化。
所谓的归一化就是通过一系列的变化，将待处理的原始图像转换成相应的唯一标准形式，该标准形式的图像对平移、旋转、缩放等仿射变换具有不变特性。
步骤S12：根据所述图像，获取所述投影空间100的投影空间参数，并确定三维空间模型。
如图6所示，步骤S12具体包括：
步骤S121：根据所述图像，获取所述投影空间100的空间特征点和空间线条。
请一并参阅图7，步骤S121具体包括：
步骤S1211：通过任意两个所述图像间的转换关系，计算图像线条两两之间相对的三维坐标。
步骤S1212：根据所述三维坐标，获取所述投影空间的空间特征点和空间线条。
利用以上的2个步骤可以获取图8中的电话机(即现实物体10)的空间特征点和空间线条(如图9所示)。
假设所述单目相机30的相机姿态为R，所述单目相机30的标定矩阵为C，空间线条在所采集的图像中的二维坐标为[u，v]，所述空间线条的在三维空间模型中的三维坐标为[x，y，z]，可以得到[u，v]＝RC[x，y，z]。
请一并参阅图10，步骤S122具体包括：
步骤S1221：所述图像基于其所在的立体视椎铺贴所述空间线条。
视椎体是一个三维体，其位置与所述单目相机30相关，视椎体的形状决定了模型如何从相机空间投影到屏幕上，最常见的投影类型为透视投影，使得离所述单目相机30近的现实物体10投影后较大，而离所述单目相机30较远的物体投影后较小。透视投影使用棱锥作为视锥体，所述单目相机30位于棱锥的椎顶，该棱锥被前后两个平面截断，形成一个棱台。
步骤S1222：根据任意两个重叠的所述图像的立体视椎，进行所述投影空间100的空间特征点和空间线条的匹配。
步骤S123：将匹配后的所述空间线条进行分组，确定所述三维空间模型。
已知每张照片的相机变换矩阵Ti，在重叠区域里两张图像的投影空间线条的三维坐标定义为L1和L2，基于极线约束对L1和L2的两张图像的相机姿态补偿，获得空间线条L1i和L2i的三维位置，其中，投影空间线条L1包括所述空间线条L1i，投影空间线条L2包括所述空间线条L2i。顺序把L1i和L2i进行匹配，覆盖率达到预设比例(比如百分之六十以上)的L1i和L2i将定义为空间里同一空间线条，将没有重叠或者覆盖率低于预设比例的空间线条L1i或L2i删除，完成所有线条的匹配过程后，将匹配后的所述空间线条进行分组，然後搜索可以转换为物理上合理的3D模型的布局。
请一并参阅图11，步骤S123具体包括：
步骤S1231：将所述空间线条根据并行性或正交性进行收敛，得到正交消失点。
步骤S1232：根据所述空间线条和所述正交消失点确定所述投影空间100的边界，确定三维空间模型。根据检测到的所述空间线条和正交消失点确定墙面的地下边界和天花板边界，基于空间设计布局，空间线条尾端连接另一空间线条的首端定义为不同墙面的连接线，两条平行的空间线条可认为其位于同一平面。从空间线条顺序生成许多物理上有效的布局假设，每个生成的假设都会得到最佳匹配空间线条的评分，最后将最佳的假设对应的最佳空间线条组合转换为三维空间模型。
步骤S1233：获取所述图像的所有图像线条，将所述图像线条整合在所述三维空间模型里，并使用线段一致性估测三维室内曼哈顿场景。
由于室内场景内普遍存在物体之间的相互遮挡以及较为明显的光照变化，加之通常情况下难以预知摄像机的参数，使得传统的算法很难通过单幅图像对室内场景进行重建。加之，大多数场景都满足的“曼哈顿特性”，即场景内的大部分平面都分布在三个相互正交的方向上，基于投票机制和灭点的二维信息与三维信息之间的交互关系逐次得到曼哈顿方向的正交灭点，摆脱了传统方法求取正交灭点时对摄像机参数的 依赖，同时约束了直线检测结果的误差、直线的长度以及候选灭点与约束直线之间的位置关系对灭点检测精度造成的影响。故而，基于上述的三个步骤可以得到投影空间100对应的较为精准的三维空间模型，且可以呈现三维空间模型中的三维室内曼哈顿场景，在视觉上更为直观，提升了用户体验。
步骤S13：根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
如图12所示，步骤S13具体包括：
步骤S131：根据所述三维空间模型识别所述投影空间100中的现实物体10以及所述现实物体10所处位置的三维坐标。
步骤S132：根据所述单目相机30的预设图像识别算法计算各所述现实物体10信息，所述现实物体10信息包括物体的轮廓、形状、大小和颜色。
所述投影空间100的全景图以所述多维电机20的起始点为原点，现实物体10的识别使用了室内物件数据集和学习的算法，识别出现实物体10的位置。将所述现实物体10位于图像中位置(二维坐标)转换为投影空间100中的三维坐标，并把位置信息(即所述三维坐标)和现实物体信息存于资料库，以便其他应用查询以及用于所述三维空间模型的更新。
请参阅图13，当所述现实物体信息包括物体的轮廓时，步骤S13具体包括：
步骤S133：根据所述预设图像识别算法，从所述图像识别出各个现实物体10，并且提取各所述现实物体10的物体轮廓。
其中，所述物体轮廓可以通过边缘检测得出，所述边缘检测的方法包括梯度算子、罗伯特算子、索伯算子、二阶微算子等。
步骤S134：识别各所述现实物体10是否存在遮挡。
步骤S135：若存在，则根据预设补偿算法，对存在遮挡的现实物体10的物体轮廓进行遮挡补偿。
综上，本申请在单目相机30采集图像的过程中同步进行三维空间模型的构建，当拍摄完成时即可以得出投影空间100的三维空间模型，避免当出现不准确的三维模型时需重构所述三维空间模型。
请参阅图14，图14为本申请又一实施例提供的一种三维建模的方法流程图。如图14所示，所述方法还包括：
步骤S17：接收虚拟现实投影指令。
在本申请实施例中，所述虚拟现实投影指令为语音指令，通过语音识别技术来确定虚拟物体和与之结合的现实物体10，语音识别技术涉及信号处理、模式识别、概率论和信息论、发声机理和听觉机理、人工智能等领域。根据语音设备和通道，所述语音识别可以分为桌面语音识别、电话语音识别和嵌入式语音识别等。
可以理解的是，通过语音指令确定所需被投的虚拟物体，以及所述虚拟物体所要投影结合的现实物体10仅是本申请实施例所公开的其中一种方式，还可通过触摸控制、键盘(包括实体键盘和软键盘)输入、手势控制、虚拟空间指令输入等方式确定。
步骤S18：根据所述虚拟现实投影指令，确定投影位置以及投影物体。
在本申请实施例中，所述投影位置是所述投影空间100所对应的三维空间模型的某一个位置的三维坐标，通过识别所述语音指令中的独立词、关键词或连续语音等，可以确定所需被投的投影物体，以及所述投影物体所要投影结合的现实物体10。在一些实施例中，在提取所述语音指令中的特征之前，应当进行一定程度的降噪处理，同时，还可以进行语音增强。
所述投影物体应当在包括位置坐标、尺寸大小以及颜色特性等方面适配所述现实物体10的特征参数，使得所述投影物体投射的影像与现实物体10结合起来看起来舒适、协调、真实。所述投影物体为虚拟对象图片中的虚拟物体，虚拟对象图片可以来源于本地数据库或互联网，从特征参数的匹配上进行初步的筛选，依据语音指令中确定的虚拟物体查找本地数据库或互联网中现有的理论上最为匹配的图片资源。
步骤S19：控制投影设备40向所述投影位置投影所述投影物体。
所述现实物体10位于所述投影空间100，且是任意的一个或者多个，相对应的，被选择的所述投影物体与所述现实物体10的被选择次数一致，数量可以不一致。可以理解的是，所述投影物体与所述现实物10体存在一一对应的关系，比如，将虚拟的刀叉图片投射到投影空间100的盘子旁边，将虚拟的艺术照片投射到投影空间100的相框内。当然，也可以将多个虚拟对象图片投射于现实空间中的同一个现实物体10，比如，将虚拟的青花瓷图片投射于花瓶的瓶身，将虚拟的花束投射于花瓶的上方，或者，将至少两张虚拟水果图片投射到投影空间100的果盘中。
本申请实施例提供了一种三维建模的方法。所述三维建模的方法应用于单目相机，所述单目相机与多维旋转电机连接，通过控制单目相机按预设方式采集投影空间的图像；根据图像，获取投影空间的投影空间参数，并确定三维空间模型；根据三维空间模型和单目相机的预设图像识别算法，识别投影空间中的现实物体以及计算各现实物体的物体信息。因此，本申请能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息，从而可以根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。
请参阅图15，为本申请实施例提供的一种三维建模的装置示意图。如图15所示，所述三维建模的装置200应用于单目相机，所述单目相机与多维旋转电机连接，所述三维建模的装置200包括：
采集单元201，用于控制所述单目相机按预设方式采集投影空间的图像。
在本申请实施例中，所述采集单元201具体用于校准所述多维旋转电机，使所述多维旋转电机处于起始点；根据所述单目相机的每一个三维区域预设的拍照次数、水平视角和垂直视角，分别计算所述单目相机水平方向和垂直方向上需采集的所述图像的张数和每两张图像的间隔角度。
确定单元202，用于根据所述图像，获取所述投影空间的投影空间 参数，并确定三维空间模型。
在本申请实施例中，所述确定单元202具体用于根据所述图像，获取所述投影空间的空间特征点和空间线条；将所述空间特征点和所述空间线条进行匹配；将匹配后的所述空间线条进行分组，确定所述三维空间模型。
其中，所述根据所述图像，获取所述投影空间的空间特征点和空间线条包括：通过任意两个所述图像间的转换关系，计算图像线条两两之间相对的三维坐标；根据所述三维坐标，获取所述投影空间的空间特征点和空间线条。
所述将所述空间特征点和所述空间线条进行匹配包括：所述图像基于其所在的立体视椎铺贴所述空间线条；根据任意两个重叠的所述图像的立体视椎，进行所述投影空间的空间特征点和空间线条的匹配。
所述将匹配后的所述空间线条进行分组，确定所述三维空间模型包括：将所述空间线条根据并行性或正交性进行收敛，得到正交消失点；根据所述空间线条和所述正交消失点确定所述投影空间的边界，确定三维空间模型；获取所述图像的所有图像线条，将所述图像线条整合在所述三维空间模型里，并使用线段一致性估测三维室内曼哈顿场景。
识别与计算单元203，用于根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
在本申请实施例中，所述识别与计算单元203具体用于根据所述三维空间模型识别所述投影空间中的现实物体以及所述现实物体所处位置的三维坐标；根据所述单目相机的预设图像识别算法计算各所述现实物体信息，所述现实物体信息包括物体的轮廓、形状、大小和颜色。
当所述现实物体信息包括物体的轮廓时，根据所述预设图像识别算法，从所述图像识别出各个现实物体，并且提取各所述现实物体的物体轮廓；识别各所述现实物体是否存在遮挡；若存在，则根据预设补偿算法，对存在遮挡的现实物体的物体轮廓进行遮挡补偿。
为实现本申请实施例的基于三维建模的方法，上述装置的各个单元 相互配合，采集单元201控制所述单目相机按预设方式采集投影空间的图像，确定单元202根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型，识别与计算单元203根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
由于装置实施例和方法实施例是基于同一构思，在内容不互相冲突的前提下，装置实施例的内容可以引用方法实施例的，在此不赘述。
本申请实施例提供了一种三维建模的装置。所述三维建模的装置应用于单目相机，所述单目相机与多维旋转电机连接，通过控制单目相机按预设方式采集投影空间的图像；根据图像，获取投影空间的投影空间参数，并确定三维空间模型；根据三维空间模型和单目相机的预设图像识别算法，识别投影空间中的现实物体以及计算各现实物体的物体信息。因此，本申请能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息，从而可以根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。
请参阅图16，图16为本申请实施例提供的一种智能终端的结构示意图。如图16所示，该智能终端300包括一个或多个处理器301以及存储器302。其中，图16中以一个处理器301为例。
处理器301和存储器302可以通过总线或者其他方式连接，图16中以通过总线连接为例。
存储器302作为一种非易失性计算机可读存储介质，可用于存储非易失性软件程序、非易失性计算机可执行程序以及模块，如本申请实施例中的三维建模的方法对应的程序指令/模块(例如，附图15所示的采集单元201、确定单元202以及识别与计算单元203)。处理器301通过运行存储在存储器302中的非易失性软件程序、指令以及模块，从而执行三维建模的装置的各种功能应用以及数据处理，即实现上述方法实施例的三维建模的方法以及上述装置实施例的各个模块和单元的功能。
存储器302可以包括存储程序区和存储数据区，其中，存储程序区 可存储操作***、至少一个功能所需要的应用程序；存储数据区可存储根据三维建模的装置的使用所创建的数据等。此外，存储器302可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实施例中，存储器302可选包括相对于处理器301远程设置的存储器，这些远程存储器可以通过网络连接至处理器301。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
所述程序指令/模块存储在所述存储器302中，当被所述一个或者多个处理器301执行时，执行上述任意方法实施例中的三维建模的方法，例如，执行以上描述的图3中的方法步骤S11至步骤S13；也可实现附图15所述的各个模块或单元的功能。
作为本申请实施例的另一方面，本申请实施例还提供一种非易失性计算机可读存储介质。非易失性计算机可读存储介质存储有智能终端可执行指令，所述计算机可执行指令用于使智能终端执行上述实施例的三维建模的方法，以达到能够快速准确地构建投影空间的空间三维模型，且能从该空间三维模型中识别投影空间中的现实物体以及计算各现实物体的物体信息，从而可以根据现实物体的物体信息来改变虚拟物体的大小、颜色、角度等使虚拟物体图像与现实物体的结合更吻合、更逼真。
上述产品可执行本申请实施例所提供的方法，具备执行方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节，可参见本申请实施例所提供的方法。
以上所描述的***或设备实施例仅仅是示意性的，其中所述作为分离部件说明的单元模块可以是或者也可以不是物理上分开的，作为模块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模块单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对相关技术做出贡 献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如ROM/RAM、磁碟、光盘等，包括若干指令用至少一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行各个实施例或者实施例的某些部分所述的方法。
最后应说明的是：以上实施例仅用以说明本申请的技术方案，而非对其限制；在本申请的思路下，以上实施例或者不同实施例中的技术特征之间也可以进行组合，步骤可以以任意顺序实现，并存在如上所述的本申请的不同方面的许多其它变化，为了简明，它们没有在细节中提供；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (11)

  1. 一种三维建模的方法，应用于单目相机，所述单目相机与多维旋转电机连接，其特征在于，包括：
    控制所述单目相机按预设方式采集投影空间的图像；
    根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型；
    根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
  2. 根据权利要求1所述的方法，其特征在于，
    所述控制所述单目相机按预设方式采集投影空间的图像包括：
    校准所述多维旋转电机，使所述多维旋转电机处于起始点；
    根据所述单目相机的每一个三维区域预设的拍照次数、水平视角和垂直视角，分别计算所述单目相机水平方向和垂直方向上需采集的所述图像的张数和每两张图像的间隔角度。
  3. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    获取所述单目相机的光线扭曲参数及偏移量参数；
    根据所述光线扭曲参数和偏移量参数，校正所述图像；
    通过焦距除法来对校正后的所述图像的点进行归一化。
  4. 根据权利要求1所述的方法，其特征在于，所述根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型包括：
    根据所述图像，获取所述投影空间的空间特征点和空间线条；
    将所述空间特征点和所述空间线条进行匹配；
    将匹配后的所述空间线条进行分组，确定所述三维空间模型。
  5. 根据权利要求4所述的方法，其特征在于，所述根据所述图像，获取所述投影空间的空间特征点和空间线条包括：
    通过任意两个所述图像间的转换关系，计算图像线条两两之间相对的三维坐标；
    根据所述三维坐标，获取所述投影空间的空间特征点和空间线条。
  6. 根据权利要求4所述的方法，其特征在于，所述将所述空间特征点和所述空间线条进行匹配包括：
    所述图像基于其所在的立体视椎铺贴所述空间线条；
    根据任意两个重叠的所述图像的立体视椎，进行所述投影空间的空间特征点和空间线条的匹配。
  7. 根据权利要求4所述的方法，其特征在于，所述将匹配后的所述空间线条进行分组，确定三维空间模型包括：
    将所述空间线条根据并行性或正交性进行收敛，得到正交消失点；
    根据所述空间线条和所述正交消失点确定所述投影空间的边界，确定三维空间模型；
    获取所述图像的所有图像线条，将所述图像线条整合在所述三维空间模型里，并使用线段一致性估测三维室内曼哈顿场景。
  8. 根据权利要求1所述的方法，其特征在于，所述根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息包括：
    根据所述三维空间模型识别所述投影空间中的现实物体以及所述现实物体所处位置的三维坐标；
    根据所述单目相机的预设图像识别算法计算各所述现实物体信息，所述现实物体信息包括物体的轮廓、形状、大小和颜色。
  9. 根据权利要求1所述的方法，其特征在于，
    当所述现实物体信息包括物体的轮廓时，
    根据所述预设图像识别算法，从所述图像识别出各个现实物体，并且提取各所述现实物体的物体轮廓；
    识别各所述现实物体是否存在遮挡；
    若存在，则根据预设补偿算法，对存在遮挡的现实物体的物体轮廓进行遮挡补偿。
  10. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    接收虚拟现实投影指令；
    根据所述虚拟现实投影指令，确定投影位置以及投影物体；
    控制投影设备向所述投影位置投影所述投影物体。
  11. 一种三维建模的装置，应用于单目相机，所述单目相机与多维旋转电机连接，其特征在于，包括：
    采集单元，用于控制所述单目相机按预设方式采集投影空间的图像；
    确定单元，用于根据所述图像，获取所述投影空间的投影空间参数，并确定三维空间模型；
    识别与计算单元，用于根据所述三维空间模型和所述单目相机的预设图像识别算法，识别所述投影空间中的现实物体以及计算各所述现实物体的物体信息。
PCT/CN2019/101613 2018-08-29 2019-08-20 一种三维建模的方法及其装置 WO2020042970A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810996321.X 2018-08-29
CN201810996321.XA CN109242958A (zh) 2018-08-29 2018-08-29 一种三维建模的方法及其装置

Publications (1)

Publication Number Publication Date
WO2020042970A1 true WO2020042970A1 (zh) 2020-03-05

Family

ID=65069781

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/101613 WO2020042970A1 (zh) 2018-08-29 2019-08-20 一种三维建模的方法及其装置

Country Status (2)

Country Link
CN (1) CN109242958A (zh)
WO (1) WO2020042970A1 (zh)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109242958A (zh) * 2018-08-29 2019-01-18 广景视睿科技（深圳）有限公司 一种三维建模的方法及其装置
CN111666792B (zh) * 2019-03-07 2023-04-28 阿里巴巴集团控股有限公司 图像识别方法、图像采集和识别方法以及商品识别方法
CN110930518A (zh) * 2019-08-29 2020-03-27 广景视睿科技（深圳）有限公司 基于增强现实技术的投影方法及投影设备
CN110969905A (zh) * 2019-11-29 2020-04-07 塔普翊海（上海）智能科技有限公司 混合现实的远程教学互动、教具互动***及其互动方法
CN111161173B (zh) * 2019-12-18 2020-11-06 北京城市网邻信息技术有限公司 图像的校正信息获取、模型构建方法和装置、介质
CN113240768B (zh) * 2019-12-18 2022-03-15 北京城市网邻信息技术有限公司 户型图处理方法、***、装置和计算机可读存储介质
CN111010561A (zh) * 2019-12-20 2020-04-14 上海沃咨信息科技有限公司 一种基于vr技术的虚拟现实投影***
CN111476894A (zh) * 2020-05-14 2020-07-31 小狗电器互联网科技（北京）股份有限公司 三维语义地图构建方法、装置、存储介质及电子设备
CN111951404A (zh) * 2020-07-24 2020-11-17 广东交通职业技术学院 一种船舶管控方法、***、装置及存储介质
CN112530005B (zh) * 2020-12-11 2022-10-18 埃洛克航空科技（北京）有限公司 一种三维模型直线结构识别与自动修复方法
CN113327329B (zh) * 2020-12-15 2024-06-14 广州富港生活智能科技有限公司 基于三维模型的室内投影方法、装置及***
CN118233606A (zh) * 2022-12-09 2024-06-21 吉猪信息科技(深圳)有限公司 一种智能投影对象控制方法、***及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103714571A (zh) * 2013-09-23 2014-04-09 西安新拓三维光测科技有限公司 一种基于摄影测量的单像机三维重建方法
CN104183019A (zh) * 2013-05-20 2014-12-03 三菱电机株式会社 用于重建3d线的方法
CN107292956A (zh) * 2017-07-12 2017-10-24 杭州电子科技大学 一种基于曼哈顿假设的场景重建方法
CN108225218A (zh) * 2018-02-07 2018-06-29 苏州镭图光电科技有限公司 基于光学微机电***的三维扫描成像方法及成像装置
CN108320333A (zh) * 2017-12-29 2018-07-24 ***股份有限公司 场景适应型虚拟现实转换设备和虚拟现实的场景适应方法
CN109242958A (zh) * 2018-08-29 2019-01-18 广景视睿科技（深圳）有限公司 一种三维建模的方法及其装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10198865B2 (en) * 2014-07-10 2019-02-05 Seiko Epson Corporation HMD calibration with direct geometric modeling
CN104932698B (zh) * 2015-06-30 2018-03-27 广景视睿科技（深圳）有限公司 一种手持交互设备装置及其投影交互方法
CN107016704A (zh) * 2017-03-09 2017-08-04 杭州电子科技大学 一种基于增强现实的虚拟现实实现方法
CN107860337B (zh) * 2017-10-11 2020-03-24 华天科技（昆山）电子有限公司 基于阵列相机的结构光三维重建方法与装置
CN107945268B (zh) * 2017-12-15 2019-11-29 深圳大学 一种基于二元面结构光的高精度三维重建方法及***

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104183019A (zh) * 2013-05-20 2014-12-03 三菱电机株式会社 用于重建3d线的方法
CN103714571A (zh) * 2013-09-23 2014-04-09 西安新拓三维光测科技有限公司 一种基于摄影测量的单像机三维重建方法
CN107292956A (zh) * 2017-07-12 2017-10-24 杭州电子科技大学 一种基于曼哈顿假设的场景重建方法
CN108320333A (zh) * 2017-12-29 2018-07-24 ***股份有限公司 场景适应型虚拟现实转换设备和虚拟现实的场景适应方法
CN108225218A (zh) * 2018-02-07 2018-06-29 苏州镭图光电科技有限公司 基于光学微机电***的三维扫描成像方法及成像装置
CN109242958A (zh) * 2018-08-29 2019-01-18 广景视睿科技（深圳）有限公司 一种三维建模的方法及其装置

Also Published As

Publication number Publication date
CN109242958A (zh) 2019-01-18

Similar Documents

Publication Publication Date Title
WO2020042970A1 (zh) 一种三维建模的方法及其装置
US11869205B1 (en) Techniques for determining a three-dimensional representation of a surface of an object from a set of images
US11170561B1 (en) Techniques for determining a three-dimensional textured representation of a surface of an object from a set of images with varying formats
US9779512B2 (en) Automatic generation of virtual materials from real-world materials
US10551913B2 (en) Virtual 3D methods, systems and software
CN107004275B (zh) 确定实物至少一部分的3d重构件空间坐标的方法和***
CN104781849B (zh) 单眼视觉同时定位与建图(slam)的快速初始化
KR101323966B1 (ko) 이미지 프로세싱을 기반으로 한 3ｄ 공간 차원용 시스템 및 방법
WO2020063139A1 (zh) 脸部建模方法、装置、电子设备和计算机可读介质
US10855932B2 (en) Processing holographic videos
US10580205B2 (en) 3D model generating system, 3D model generating method, and program
CN110392251B (zh) 一种基于虚拟现实的动态投影方法及***
JP2021530817A (ja) 画像ディスプレイデバイスの位置特定マップを決定および／または評価するための方法および装置
TW202205059A (zh) 虛擬對象的控制方法、電子設備及電腦可讀儲存介質
CN110889890A (zh) 图像处理方法及装置、处理器、电子设备及存储介质
CN108369653A (zh) 使用眼睛特征的眼睛姿态识别
CN104881526B (zh) 一种基于3d的物品穿戴方法及眼镜试戴方法
WO2019035155A1 (ja) 画像処理システム、画像処理方法、及びプログラム
JP2024056955A (ja) 光学式捕捉によるパーソナライズされたｈｒｔｆ
US11508141B2 (en) Simple environment solver using planar extraction
JPH09139956A (ja) 電子シーン分析および強調のための装置および方法
TW202025719A (zh) 圖像處理方法及裝置、電子設備及儲存介質
JP6656382B2 (ja) マルチメディア情報を処理する方法及び装置
CN109064533B (zh) 一种3d漫游方法及***
WO2023024441A1 (zh) 模型重建方法及相关装置、电子设备和存储介质

Legal Events

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

Ref document number: 19856214

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 10/06/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19856214

Country of ref document: EP

Kind code of ref document: A1