WO2024051783A1 - 一种三维重建方法、装置、设备和存储介质 - Google Patents

一种三维重建方法、装置、设备和存储介质 Download PDF

Info

Publication number
WO2024051783A1
WO2024051783A1 PCT/CN2023/117506 CN2023117506W WO2024051783A1 WO 2024051783 A1 WO2024051783 A1 WO 2024051783A1 CN 2023117506 W CN2023117506 W CN 2023117506W WO 2024051783 A1 WO2024051783 A1 WO 2024051783A1
Authority
WO
WIPO (PCT)
Prior art keywords
reconstructed
images
reconstruction
matching
result
Prior art date
Application number
PCT/CN2023/117506
Other languages
English (en)
French (fr)
Inventor
刘雨
Original Assignee
先临三维科技股份有限公司
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 先临三维科技股份有限公司 filed Critical 先临三维科技股份有限公司
Publication of WO2024051783A1 publication Critical patent/WO2024051783A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/0895—Weakly supervised learning, e.g. semi-supervised or self-supervised learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T15/00—3D [Three Dimensional] image rendering
    • G06T15/10—Geometric effects
    • G06T15/20—Perspective computation
    • 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
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration

Definitions

  • the present disclosure relates to the field of computer technology, and in particular, to a three-dimensional reconstruction method, device, equipment and storage medium.
  • 3D reconstruction technology is widely used in the field of oral medicine.
  • professional dental fields such as implantology and orthodontics
  • 3D tooth model reconstruction is mostly achieved through binocular cameras and structured light projectors.
  • this method requires professionals to operate the equipment.
  • the requirements for the operating environment are relatively high, and the operation process is complicated, so The hardware cost of the equipment used is also relatively high.
  • the technical problems to be solved by this disclosure are the existing dental model reconstruction operation procedures are complicated, the operating environment requirements are high, and the equipment hardware cost is high.
  • embodiments of the present disclosure provide a three-dimensional reconstruction method, including:
  • Three-dimensional reconstruction is performed based on the graphic structure to obtain a first reconstruction result of the target object, where the first reconstruction result includes three-dimensional data of the target object.
  • a three-dimensional reconstruction device including:
  • the acquisition module is used to acquire multiple images to be reconstructed of the target object
  • a matching module configured to perform feature matching based on the extracted feature data of the multiple images to be reconstructed, and obtain a first matching result of the multiple images to be reconstructed;
  • a construction module configured to construct a pattern structure between multiple target images to be reconstructed in the plurality of images to be reconstructed according to the first matching result
  • a reconstruction module configured to perform three-dimensional reconstruction based on the graphic structure to obtain a first reconstruction result of the target object, wherein the first reconstruction result includes three-dimensional data of the target object.
  • an electronic device including:
  • the computer program is stored in the memory and configured to be executed by the processor to implement the above three-dimensional reconstruction method.
  • a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the steps of the above-mentioned three-dimensional reconstruction method are implemented.
  • the three-dimensional reconstruction method includes: acquiring multiple images to be reconstructed of the target object; performing feature matching based on the extracted feature data of the multiple images to be reconstructed, Obtain the first matching results of multiple images to be reconstructed; construct a pattern structure between multiple target images to be reconstructed in the multiple images to be reconstructed based on the first matching results; perform three-dimensional reconstruction based on the pattern structure to obtain the third image of the target object.
  • a reconstruction result, wherein the first reconstruction result includes three-dimensional data of the target object.
  • the method provided by the present disclosure obtains multiple images to be reconstructed of the target object through a movable terminal with a camera function, and completes the three-dimensional reconstruction of the target object based on the multiple images to be reconstructed, which simplifies the operation process, reduces the equipment cost, and achieves three-dimensional reconstruction.
  • the effect is relatively good.
  • Figure 1 is a schematic diagram of an application scenario provided by an embodiment of the present disclosure
  • Figure 2 is a schematic flowchart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure
  • Figure 3 is a schematic flowchart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure
  • Figure 4 is a schematic flowchart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure
  • Figure 5 is a schematic structural diagram of a three-dimensional reconstruction device provided by an embodiment of the present disclosure.
  • FIG. 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure.
  • the three-dimensional reconstruction of teeth is mostly achieved through binocular cameras and structured light projectors.
  • the three-dimensional reconstruction of teeth in a single frame is achieved through equipment, and then the three-dimensional reconstruction is completed through a multi-frame point cloud splicing algorithm.
  • the equipment operation process of the above method is relatively complicated.
  • the hardware cost of the equipment is high, it requires a high-configuration terminal to implement, and the requirements for the operating environment are also relatively high.
  • embodiments of the present disclosure provide a three-dimensional reconstruction method.
  • the user uses a mobile terminal to collect dental and jaw data of a sequence of images offline through a fixed device in the mouth.
  • the teeth and jaws in the mouth are the target objects to be reconstructed.
  • the teeth and jaws The data includes multiple images to be reconstructed.
  • the mobile terminal can be a terminal with a shooting function, such as a mobile phone or a simple monocular camera.
  • the dental data includes crown data, gum data, etc., and the dental data is then transmitted to the server.
  • Perform three-dimensional reconstruction that is, reconstruct a three-dimensional model of the target object based on multiple two-dimensional images to be reconstructed.
  • the method provided by this disclosure uses a mobile terminal with a shooting function to collect dental and jaw data.
  • the equipment hardware cost is low, and the use cost and holding cost are also relatively low. It further simplifies the operation process of the equipment.
  • After collecting the dental and jaw data it is directly Offline 3D reconstruction is performed on cloud servers, computers or mobile terminals. It has a simple structure, is easy to operate, has low requirements on the user's operating environment, can be applied to a variety of scenarios, and the accuracy and efficiency of subsequent 3D reconstruction are also relatively high. The details will be described in detail through one or more embodiments below.
  • FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the present disclosure.
  • the terminal 11 collects multiple images to be reconstructed of a target object
  • the server 12 acquires multiple images to be reconstructed for three-dimensional reconstruction.
  • Multiple images to be reconstructed may be captured by the terminal 11.
  • the multiple images to be reconstructed are obtained by the terminal 11 from other devices.
  • the multiple images to be reconstructed are images obtained by the terminal 11 performing image processing on a preset image.
  • the preset image may be photographed by the terminal 11, or the preset image may be obtained by the terminal 11 from other devices. .
  • there are no specific limitations on other devices. It can be understood that the three-dimensional reconstruction method provided by the embodiments of the present disclosure is not limited to the several possible scenarios mentioned above.
  • FIG. 2 is a schematic flow chart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure. Used for the above-mentioned server 12, specifically including the following steps S210 to S240 as shown in Figure 2:
  • the target object may be teeth
  • the images to be reconstructed may be generated by the user using a mobile terminal with a shooting function such as a digital camera or a mobile phone camera to collect dental data by himself.
  • the image to be reconstructed can also be called a tooth image.
  • users can collect dental data including multiple images to be reconstructed by themselves.
  • the collected dental data can be transmitted to the server in real time, or the complete dental data in the user's mouth can be collected offline.
  • users can also directly collect continuous video streams about intraoral teeth and then transmit them to the server offline or in real time.
  • each image to be reconstructed includes at least part of the target object.
  • the target object to be acquired is the teeth, and the teeth include crowns and gingiva, etc., and the image to be reconstructed includes at least part of the crowns and/or gingiva. .
  • S220 Perform feature matching based on the extracted feature data of the multiple images to be reconstructed, and obtain a first matching result of the multiple images to be reconstructed.
  • the n images to be reconstructed can be matched in pairs, n A certain image to be reconstructed among the images to be reconstructed will obtain n-1 related sub-matching results.
  • the multiple images to be reconstructed are dental data collected continuously in the mouth
  • the adjacent images to be reconstructed in the multiple images to be reconstructed can be directly matched in pairs to obtain the two-dimensional feature of the adjacent images to be reconstructed.
  • the specific matching method can be selected according to the user's method of collecting dental and jaw data, and is not limited here.
  • obtaining the first matching result in the above S220 can be achieved through the following steps:
  • Feature data of each two images to be reconstructed among the multiple images to be reconstructed are matched to obtain a first matching result.
  • the features of the target objects included in each image to be reconstructed are extracted through a pre-trained feature extractor, that is, the features of the teeth in each image to be reconstructed are extracted, and the feature data of each image to be reconstructed is obtained.
  • the feature extractor used for feature extraction can be obtained by training a large number of tooth data sets using the built self-supervised deep learning framework (SuperPoint).
  • SuperPoint built self-supervised deep learning framework
  • the feature extractor can be used to accurately extract the feature data of the tooth image.
  • the feature extractor The specific training process is not limited here.
  • the feature matcher used for feature matching can be obtained by training a large number of tooth data sets by a constructed multi-layer neural network structure (SuperGlue).
  • the tooth data set can be the feature data output by the feature extractor that has been trained above. , the specific training process of the feature matcher is not limited here.
  • the two images to be reconstructed included in the image pair are the target images to be reconstructed, that is, through each two images to be reconstructed
  • the matching relationship between images (image pairs) establishes the correlation between multiple images to be reconstructed to complete the conversion from two-dimensional to three-dimensional.
  • S240 Perform three-dimensional reconstruction based on the graphic structure to obtain a first reconstruction result of the target object.
  • the first reconstruction result includes three-dimensional data of the target object.
  • three-dimensional reconstruction is performed based on the pattern structure to obtain the first reconstruction result of the target object.
  • the first reconstruction result includes the target object.
  • Three-dimensional data three-dimensional model.
  • the three-dimensional data can be three-dimensional grid data or three-dimensional point cloud data. The specific type of reconstructed three-dimensional data is not limited.
  • Embodiments of the present disclosure provide a three-dimensional reconstruction method.
  • the user photographs a target object through a movable terminal with a camera function, and generates multiple images to be reconstructed.
  • the operation process is simple, and the requirements for the operating environment and operators are relatively low.
  • the operating equipment The cost is also relatively low.
  • the server obtains multiple to-be-reconstructed images of the target object generated by the terminal, extracts the feature data of each of the multiple to-be-reconstructed images, and performs pairwise matching between the multiple to-be-reconstructed images based on the feature data to obtain A first matching result related to at least part of the plurality of images to be reconstructed, the matching result reflecting the two-dimensional matching relationship of the image pairs; constructing multiple target images to be reconstructed among the plurality of images to be reconstructed according to the first matching result The graphical structure between them; finally, three-dimensional reconstruction is performed based on the graphical structure to obtain the first reconstruction result of the target object, completing the conversion of two-dimensional data to three-dimensional data.
  • the first reconstruction result includes a three-dimensional reconstruction result of the target object with good reconstruction effect. data.
  • the method provided by the present disclosure obtains multiple images to be reconstructed of the target object through a movable terminal with a camera function, and completes the three-dimensional reconstruction of the target object based on the multiple images to be reconstructed, which simplifies the operation process, facilitates implementation, and reduces equipment costs. And effectively solves the problem of poor three-dimensional reconstruction results.
  • FIG. 3 is a schematic flowchart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure.
  • multiple of the multiple images to be reconstructed are constructed.
  • the graphic structure between images to be reconstructed includes: The following steps S310 to S320 shown in Figure 3:
  • the sub-matching results with mismatching relationships among the multiple sub-matching results included in the first matching result are filtered out to obtain the second matching result.
  • the sub-matching results included in the second matching result are The number should be less than or equal to the number of sub-matching results included in the first matching result. Filter sub-matching results with mismatching relationships and improve the matching rate. This can reduce the amount of data for subsequent 3D reconstruction, improve 3D reconstruction efficiency, and improve 3D reconstruction. Effect.
  • obtaining the second matching result in the above S310 can be achieved through the following steps:
  • Matching point information in each sub-matching result included in the first matching result is counted to filter multiple sub-matching results included in the first matching result to obtain a third matching result.
  • the matching point information in each sub-matching result included in the first matching result is counted.
  • the sub-matching result is obtained by matching the feature data of two images (image pairs) to be reconstructed.
  • the sub-matching result includes the image pair.
  • the matching point information is the matching point information.
  • the matching point information includes the number and distribution of matching points.
  • the number and distribution of matching points in multiple sub-matching results are less than or equal to the preset threshold. Delete, for example, if the number and distribution of matching points included in the sub-matching result 1 that reflects the matching relationship between the feature points of image 1 and image 2 is less than the preset threshold, it means that image 1 and image 2 are mismatched, and the sub-matching needs to be Result 1 is deleted to improve the match rate.
  • the multiple sub-matching results included in the first matching result are filtered for the first time through the matching quality and threshold to obtain the third matching result.
  • the number of sub-matching results included in the third matching result must be less than or equal to the number of sub-matching results included in the first matching result. The number of matching results.
  • the binocular epipolar geometric algorithm is used to determine the corresponding sub-matching results included in the third matching result to be reconstructed.
  • the correspondence between images is used to perform a second filtering on multiple sub-matching results included in the third matching result to obtain the second matching result, that is, the mismatched images are further filtered through the calculated correspondence between image pairs, where,
  • the number of sub-matching results included in the second matching result must be less than or equal to the number of sub-matching results included in the third matching result.
  • the number of images to be reconstructed corresponding to the second matching result must also be less than or equal to the number of images to be reconstructed corresponding to the third matching result.
  • the number of reconstructed images, the image to be reconstructed corresponding to the second matching result is the target image to be reconstructed, and the image to be reconstructed corresponding to the second matching result is an image with a more accurate image pair matching.
  • the plurality of target images to be reconstructed are images to be reconstructed corresponding to multiple sub-matching results included in the second matching result.
  • a graph optimization (pose-graph) algorithm is used to construct multiple sub-matching results included in the second matching result.
  • the target is to reconstruct the graphical structure between images.
  • the pose-graph algorithm is an algorithm that locally optimizes multiple frames of images to obtain the pose of each frame, and then calculates the relative pose between adjacent key frames.
  • the camera position of the image to be reconstructed corresponding to each sub-matching result in the second matching result is determined, and the camera position is used as the target vertex.
  • the two-dimensional matching relationship between the images to be reconstructed in each sub-matching result is used as the target edge.
  • the camera position of each image to be reconstructed corresponding to the matching result is determined according to each sub-matching result in the second matching result, that is, the camera position of each target image to be reconstructed is determined through the matching relationship. Specifically, a certain The camera position of a target image pair to be matched is used as the target vertex, and the two-dimensional matching relationship between the images to be reconstructed in each sub-matching result is used as the target edge. Based on the target vertices and target edges, multiple target graphs to be reconstructed are constructed.
  • the graph structure between images starts from the target vertex, and the target image to be reconstructed corresponding to the target vertex is used as the first frame image, and the relative pose between adjacent frame images is gradually calculated based on the matching relationship, that is, one frame Calculate the camera pose of each image frame by image, and construct a graph structure (graph structure) between the target images to be reconstructed with an effective matching relationship according to the estimated camera trajectory.
  • graph structure graph structure
  • Embodiments of the present disclosure provide a three-dimensional reconstruction method.
  • a second matching result reflecting a valid matching relationship is obtained, and then multiple targets are constructed based on the second matching result.
  • the third matching result is obtained. Then, the sub-matching results that still have mismatching relationships in the third matching result are further filtered through the binocular epipolar geometry and pose-graph algorithm to obtain the second matching result.
  • the corresponding camera of the image to be reconstructed is determined based on the sub-matching result. position, and the camera position is used as the target vertex, and the two-dimensional matching relationship between the image and the image is the target edge, and a graph structure between multiple target images to be reconstructed with effective matching relationships is constructed.
  • the method provided by the present disclosure obtains the second matching result by filtering sub-matching results with mismatching relationships multiple times, improves the matching rate, and simultaneously completes the construction of a graph structure between multiple target images to be reconstructed, which facilitates subsequent graph structure-based construction. Efficiently and accurately construct the 3D data of the target object, complete the conversion between 2D data and 3D data, and provide technical support for simplifying the 3D reconstruction operation process.
  • FIG. 4 is a schematic flow chart of a three-dimensional reconstruction method provided by an embodiment of the present disclosure.
  • the three-dimensional reconstruction is performed based on the graphic structure to obtain the first image of the target object.
  • the reconstruction result specifically includes the following steps S410 to S420 as shown in Figure 4:
  • sparse reconstruction of the three-dimensional data of the target object is performed based on the graph structure to obtain the second reconstruction result of the target object.
  • the second reconstruction result includes a large number of three-dimensional points.
  • Sparse reconstruction is a process of determining three-dimensional points based on two-dimensional feature points.
  • the camera parameters include camera internal parameters, camera extrinsic parameters and camera pose, that is, each shot of the image to be reconstructed is There are corresponding camera parameters for the reconstructed images, and the camera's trajectory can be estimated based on the camera parameters.
  • the above-mentioned S410 to obtain the second reconstruction result and estimate the camera parameters of multiple target zone reconstructed images can be achieved through the following steps:
  • An initial image to be reconstructed is determined among multiple target images to be reconstructed corresponding to the pattern structure, and binocular reconstruction is performed on the initial image to be reconstructed to obtain an initial three-dimensional point.
  • Each image pair is composed of two target images to be reconstructed.
  • the multiple image pairs are the multiple image pairs corresponding to the graph structure.
  • the target image to be reconstructed consists of.
  • the target image pair with the highest matching quality is determined among multiple image pairs, and the two images to be reconstructed included in the target image pair are used as the initial images to be reconstructed.
  • the highest matching quality refers to the highest number and distribution of matching points between the image pairs. .
  • binocular reconstruction is performed on the initial image to be reconstructed to obtain the initial three-dimensional point of the tooth three-dimensional model. This initial three-dimensional point can be understood as the starting three-dimensional point.
  • the adjacent edges in the graphic structure are incrementally optimized through the bundle adjustment optimization algorithm (Bundle Adjustment, BA) to add new three-dimensional points, that is, starting from the initial three-dimensional point , based on the two-dimensional feature points in the graphic structure, the three-dimensional points of the three-dimensional tooth model are continuously expanded until the sparse reconstruction of the target object is completed, and the second reconstruction result is obtained.
  • BA bundle Adjustment
  • the first reconstruction result is the complete three-dimensional mesh data of the target object.
  • the above-mentioned S420 obtains the first reconstruction result through the following steps:
  • a target three-dimensional scene of the target object is constructed according to camera parameters of the multiple target images to be reconstructed.
  • a neural radiation scene description about the target object is obtained according to the target three-dimensional target scene.
  • a first reconstruction result of the target object is generated based on the neural radiation scene description.
  • the scene in the user's mouth is non-explicitly modeled through the neural radiation field (Neural Radiance Field, Nerf) according to the estimated camera parameters of the multiple target images to be reconstructed, and the result includes:
  • the target three-dimensional scene of the three-dimensional data of the target object can be understood as a three-dimensional scene of teeth.
  • the three-dimensional scene of teeth in the mouth can be represented as a function input as a 5D vector, and then the neural network about the three-dimensional teeth data in the mouth can be obtained.
  • Radiation field description and then combined with the MachineCube algorithm to extract the three-dimensional data of the teeth from the neural radiation field description.
  • the extracted three-dimensional data of the teeth is the first reconstruction result, that is, the second reconstruction result of the sparse reconstruction is densely reconstructed, and multiple Three-dimensional points are supplemented to describe, and the first reconstruction result with the best reconstruction effect is obtained.
  • Embodiments of the present disclosure provide a three-dimensional reconstruction method that performs sparse reconstruction based on a graph structure, obtains a second reconstruction result of the target object, and estimates camera parameters of multiple target images to be reconstructed, wherein sparse reconstruction mainly reconstructs Two-dimensional matching points (two-dimensional feature points) three-dimensional points to determine a series of three-dimensional points corresponding to the two-dimensional matching points related to the target object; then based on the camera parameters of the multiple target images to be reconstructed, the second reconstruction result is densely reconstructed through the Nerf algorithm to obtain the optimal The first reconstruction result of 3D reconstruction effect.
  • FIG. 5 is a schematic structural diagram of a three-dimensional reconstruction device provided by an embodiment of the present disclosure.
  • the three-dimensional reconstruction device provided by the embodiment of the present disclosure can execute the processing flow provided by the above-mentioned three-dimensional reconstruction method embodiment.
  • the three-dimensional reconstruction device 500 includes an acquisition module 510, a matching module 520, a construction module 530 and a reconstruction module 540, where :
  • the acquisition module 510 is used to acquire multiple images of the target object to be reconstructed
  • the matching module 520 is configured to perform feature matching based on the extracted feature data of the multiple images to be reconstructed, and obtain a first matching result of the multiple images to be reconstructed;
  • a construction module 530 configured to construct a pattern structure between multiple target images to be reconstructed in the multiple images to be reconstructed according to the first matching result
  • the reconstruction module 540 is configured to perform three-dimensional reconstruction based on the graphic structure to obtain a first reconstruction result of the target object, where the first reconstruction result includes the three-dimensional data of the target object.
  • the reconstruction module 540 is specifically used for:
  • the second reconstruction result is optimized to obtain a first reconstruction result of the target object.
  • the matching module 520 is specifically used for:
  • Feature data of each two images to be reconstructed among the multiple images to be reconstructed are matched to obtain a first matching result.
  • building module 530 is specifically used for:
  • the plurality of target images to be reconstructed are images to be reconstructed corresponding to multiple sub-matching results included in the second matching result.
  • building module 530 is specifically used for:
  • building module 530 is specifically used for:
  • the reconstruction module 540 is specifically used for:
  • the reconstruction module 540 is specifically used for:
  • a first reconstruction result of the target object is generated based on the neural radiation scene description.
  • the three-dimensional reconstruction device of the embodiment shown in Figure 5 can be used to perform the technical solution of the above method embodiment. Its implementation principles and technical effects are similar and will not be described again here.
  • FIG. 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure.
  • the electronic device 600 in the embodiment of the present disclosure may include, but is not limited to, mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players), vehicle-mounted terminals ( Mobile terminals such as vehicle navigation terminals), wearable electronic devices, etc., and fixed terminals such as digital TVs, desktop computers, smart home devices, etc.
  • the electronic device shown in FIG. 6 is only an example and should not impose any limitations on the functions and scope of use of the embodiments of the present disclosure.
  • the electronic device 600 may include a processing device (eg, central processing unit, graphics processor, etc.) 601, which may be loaded into a random access device according to a program stored in a read-only memory (ROM) 602 or from a storage device 608.
  • the program in the memory (RAM) 603 performs various appropriate actions and processes to implement the three-dimensional reconstruction method according to the embodiments of the present disclosure.
  • various programs and data required for the operation of the electronic device 600 are also stored.
  • the processing device 601, ROM 602 and RAM 603 are connected to each other via a bus 604.
  • An input/output (I/O) interface 605 is also connected to bus 604.
  • input devices 606 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; including, for example, a liquid crystal display (LCD), speakers, vibration
  • An output device 607 such as a computer
  • a storage device 608 including a magnetic tape, a hard disk, etc.
  • Communication device 609 may allow electronic device 600 to communicate wirelessly or wiredly with other devices to exchange data.
  • FIG. 6 illustrates electronic device 600 with various means, it should be understood that implementation or availability of all illustrated means is not required. More or fewer means may alternatively be implemented or provided.
  • embodiments of the present disclosure include a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, the computer program including program code for executing the method shown in the flowchart, thereby achieving the above The three-dimensional reconstruction method.
  • the computer program may be downloaded and installed from the network via communication device 609, or from storage device 608, or from ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the method of the embodiment of the present disclosure are performed.
  • the computer-readable medium mentioned above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two.
  • the computer-readable storage medium may be, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer readable storage media may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard drive, random access memory (RAM), read only memory (ROM), removable Programmd read-only memory (EPROM or flash memory), fiber optics, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device.
  • a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.
  • a computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device .
  • Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wire, optical cable, RF (radio frequency), etc., or any suitable combination of the above.
  • the client and the server may utilize any currently known protocol such as HTTP (HyperText Transfer Protocol). or network protocols developed in the future, and can be interconnected with any form or medium of digital data communication (e.g., communication network).
  • Examples of communication networks include local area networks (“LAN”), wide area networks (“WAN”), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or developed in the future network of.
  • the above-mentioned computer-readable medium may be included in the above-mentioned electronic device; it may also exist independently without being assembled into the electronic device.
  • the electronic device may also perform other steps described in the above embodiments.
  • Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, including but not limited to object-oriented programming languages—such as Java, Smalltalk, C++, and Includes conventional procedural programming languages—such as "C” or similar programming languages.
  • the program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server.
  • the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as an Internet service provider through Internet connection).
  • LAN local area network
  • WAN wide area network
  • Internet service provider such as an Internet service provider through Internet connection
  • each block in the flowchart or block diagram may represent a module, segment, or portion of code that contains one or more logic functions that implement the specified executable instructions.
  • the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown one after another may actually execute substantially in parallel, or they may sometimes execute in the reverse order, depending on the functionality involved.
  • each block in the block diagram and/or flowchart illustration, and groups of blocks in the block diagram and/or flowchart illustration may be implemented using dedicated hardware-based systems that perform specified functions or operations, or may be implemented using a combination of dedicated hardware and computer instructions.
  • the units involved in the embodiments of the present disclosure can be implemented in software or hardware. Among them, the name of a unit does not constitute a limitation on the unit itself under certain circumstances.
  • FPGAs Field Programmable Gate Arrays
  • ASICs Application Specific Integrated Circuits
  • ASSPs Application Specific Standard Products
  • SOCs Systems on Chips
  • CPLD Complex Programmable Logical device
  • a machine-readable medium may be a tangible medium that may contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.
  • the machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium.
  • Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any suitable combination of the foregoing.
  • machine-readable storage media would include one or more wire-based electrical connections, laptop disks, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • RAM random access memory
  • ROM read only memory
  • EPROM or flash memory erasable programmable read only memory
  • CD-ROM portable compact disk read-only memory
  • magnetic storage device or any suitable combination of the above.
  • embodiments of the present disclosure also provide a computer-readable storage medium on which a computer program is stored, and the computer program is executed by a processor to implement the three-dimensional reconstruction method described in the above embodiments.
  • embodiments of the present disclosure also provide a computer program product, which includes a computer program or instructions. When executed by a processor, the computer program or instructions implement the three-dimensional reconstruction method as described above.
  • the three-dimensional reconstruction method provided by this disclosure can use a mobile terminal with a shooting function to collect dental data.
  • the equipment hardware cost is low, and the use cost and holding cost are also relatively low.
  • the three-dimensional dental data can be well constructed. model, and the accuracy and efficiency of three-dimensional reconstruction are relatively high, which has strong industrial practicability.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Geometry (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Graphics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Processing (AREA)

Abstract

本公开涉及一种三维重建方法、装置、设备和存储介质。三维重建方法包括：获取目标对象的多个待重建图像（S210）；基于提取的多个待重建图像的特征数据进行特征匹配，得到多个待重建图像的第一匹配结果（S220）；根据第一匹配结果，构建多个待重建图像中多个目标待重建图像间的图型结构（S230）；基于图型结构进行三维重建，得到目标对象的第一重建结果（S240），其中，第一重建结果包括目标对象的三维数据。本公开提供的方法，通过具有摄像功能的可移动终端获取目标对象的多个待重建图像，基于多个待重建图像完成目标对象的三维重建，简化了操作流程，降低了设备成本，且三维重建效果比较好。

Description

一种三维重建方法、装置、设备和存储介质
本公开要求于2022年09月09日提交中国专利局、申请号为202211104909.2、发明名称为“一种三维重建方法、装置、设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及计算机技术领域，尤其涉及一种三维重建方法、装置、设备和存储介质。
背景技术
随着三维技术的发展，三维重建技术在口腔医学领域被广泛应用，在专业种植、正畸等牙科领域，需要获得高精度的牙齿模型以进行种植和正畸诊疗。目前，牙齿三维模型重建多是通过双目相机和结构光投射器实现的，但是该种方式需要专业人员来操作设备，在设备的使用过程中对操作环境的要求比较高，操作流程复杂，所使用的设备的硬件成本也比较高。
发明内容
(一)要解决的技术问题
本公开要解决的技术问题是现有牙齿模型重建操作流程复杂，操作环境要求高，以及设备硬件成本高的问题。
(二)技术方案
为了解决上述技术问题，本公开实施例提供了一种三维重建方法，包括：
获取目标对象的多个待重建图像；
基于提取的所述多个待重建图像的特征数据进行特征匹配，得到 所述多个待重建图像的第一匹配结果；
根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构；
基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，其中，所述第一重建结果包括所述目标对象的三维数据。
第二方面，还提供一种三维重建装置，包括：
获取模块，用于获取目标对象的多个待重建图像；
匹配模块，用于基于提取的所述多个待重建图像的特征数据进行特征匹配，得到所述多个待重建图像的第一匹配结果；
构建模块，用于根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构；
重建模块，用于基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，其中，所述第一重建结果包括所述目标对象的三维数据。
第三方面，还提供一种电子设备，包括：
存储器；
处理器；以及
计算机程序；
其中，所述计算机程序存储在所述存储器中，并被配置为由所述处理器执行以实现上述的三维重建方法。
第四方面，还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现上述的三维重建方法的步骤。
(三)有益效果
本公开实施例提供的上述技术方案与现有技术相比具有如下优点：
本公开实施例提供的该三维重建方法，包括：获取目标对象的多个待重建图像；基于提取的多个待重建图像的特征数据进行特征匹配， 得到多个待重建图像的第一匹配结果；根据第一匹配结果，构建多个待重建图像中多个目标待重建图像间的图型结构；基于图型结构进行三维重建，得到目标对象的第一重建结果，其中，第一重建结果包括目标对象的三维数据。本公开提供的方法，通过具有摄像功能的可移动终端获取目标对象的多个待重建图像，基于多个待重建图像完成目标对象的三维重建，简化了操作流程，降低了设备成本，且三维重建效果比较好。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本公开。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本公开的实施例，并与说明书一起用于解释本公开的原理。
为了更清楚地说明本公开实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本公开实施例提供的一种应用场景的示意图；
图2为本公开实施例提供的一种三维重建方法的流程示意图；
图3为本公开实施例提供的一种三维重建方法的流程示意图；
图4为本公开实施例提供的一种三维重建方法的流程示意图；
图5为本公开实施例提供的一种三维重建装置的结构示意图；
图6为本公开实施例提供的一种电子设备的结构示意图。
具体实施方式
为使本公开实施例的目的、技术方案和优点更加清楚，下面将对本公开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公开的一部分实施例，而不是全部的实施例。基于本公开中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本公开保护的范围。
目前，牙齿的三维重建多是通过双目相机和结构光投射器实现的，通过设备实现单帧的牙齿三维重建，再通过多帧点云拼接算法完成三维重建，但是上述方法设备操作流程比较复杂，设备的硬件成本高，需要高配置的终端来实现，对操作环境的要求也比较高。
针对上述技术问题，本公开实施例提供了一种三维重建方法，用户使用可移动终端通过固定设备离线在口内采集序列图像的牙颌数据，口内的牙颌即为待重建的目标对象，牙颌数据包括多个待重建图像，其中，可移动终端可以是具有拍摄功能的终端，例如，手机或者简易单目相机，牙颌数据包括牙冠数据、牙龈数据等，随后将牙颌数据传输至服务器进行三维重建，也就是基于多个二维的待重建图像重建出目标对象的三维模型。本公开提供的方法，使用具有拍摄功能的可移动终端采集牙颌数据，设备硬件成本低，使用成本和持有成本也比较低，进一步简化了设备的操作流程，采集到牙颌数据后直接在云服务器、计算机或者可移动终端上进行离线三维重建，结构简单，操作方便，对用户的操作环境要求低，可适用于多种场景，且后续三维重建的精度和效率也比较高。具体通过下述一个或多个实施例进行详细说明。
具体的，下述实施例以可移动终端采集待重建图像，服务器执行三维重建方法为例进行说明。示例性的，参见图1，图1为本公开实施例提供的一种应用场景的示意图，终端11采集目标对象的多个待重建图像，服务器12获取多个待重建图像进行三维重建。多个待重建图像可以是终端11拍摄获得的。或者，多个待重建图像是终端11从其他设备中获取的。再或者，多个待重建图像是终端11对预设图像进行图像处理后得到的图像，该预设图像可以是终端11拍摄获得的，或者该预设图像可以是终端11从其他设备中获取的。此处，并不对其他设备做具体限定。可以理解的是，本公开实施例提供的三维重建方法并不限于如上所述的几种可能场景。
图2为本公开实施例提供的一种三维重建方法的流程示意图，应 用于上述服务器12，具体包括如图2所示的如下步骤S210至S240：
S210、获取目标对象的多个待重建图像。
可理解的，获取目标对象的多个待重建图像，目标对象可以是牙颌，待重建图像可以是在用户使用数码相机或者手机相机等具有拍摄功能的可移动终端自行采集牙颌数据生成的，该种场景下，待重建图像也可以称为牙齿图像。具体的，用户可以自行采集包括多个待重建图像的牙颌数据，在采集过程中可以将采集的牙颌数据实时传输至服务器，也可以是离线采集完用户口内的完整牙颌数据后，再将牙颌数据传输至服务器，用户还可以直接采集关于口内牙颌的连续视频流，随后离线或实时传输至服务器，随后服务器对接收到的视频流进行抽帧以获取到多个待重建图像，具体应用场景不作限定。其中，每个待重建图像包括至少部分的目标对象，例如，被采集的目标对象是牙颌，牙颌包括牙冠和牙龈等，采集到的待重建图像包括至少部分的牙冠和/或牙龈。
S220、基于提取的所述多个待重建图像的特征数据进行特征匹配，得到所述多个待重建图像的第一匹配结果。
可理解的，在上述S210的基础上，获取到多个待重建图像后，对多个待重建图像进行特征提取，得到多个待重建图像的特征数据。随后，将多个待重建图像的特征数据进行特征匹配，得到和多个待重建图像中至少部分待重建图像相关的第一匹配结果，其中，第一匹配结果包括多个子匹配结果，每个匹配结果反映多个待重建图像中每两个待重建图像之间的二维(2d-2d)匹配关系，例如，获取到n个待重建图像，可以将n个待重建图像进行两两匹配，n个待重建图像中的某1个待重建图像会得到n-1个相关的子匹配结果。或者，若获取的多个待重建图像是在口内连续采集的牙颌数据，则可以直接将多个待重建图像中的相邻待重建图像进行两两匹配，得到相邻待重建图像特征的二维匹配关系，具体的匹配方式可根据用户采集牙颌数据的方式自行选取，在此不作限定。
可选的，上述S220得到第一匹配结果具体可以通过下述步骤实现：
提取所述多个待重建图像中每个待重建图像中包括的目标对象的特征，得到所述每个待重建图像的特征数据。
将所述多个待重建图像中每两个待重建图像的特征数据进行特征匹配，得到第一匹配结果。
可理解的，通过预先训练的特征提取器来提取每个待重建图像中包括的目标对象的特征，也就是提取每个待重建图像中牙颌的特征，得到每个待重建图像的特征数据。其中，用于特征提取的特征提取器可以是由构建的自监督深度学习框架(SuperPoint)通过大量的牙齿数据集训练得到的，利用特征提取器可以准确提取牙齿图像的特征数据，特征提取器的具体训练过程在此不作限定。
可理解的，得到每个待重建图像的特征数据后，基于预先训练的特征匹配器对多个待重建图像中每两个待重建图像的特征数据进行特征匹配，得到多个反应每两个待重建图像之间匹配关系的子匹配结果，进而由多个子匹配结果组成第一匹配结果，具有匹配关系的两个待重建图像可以称为图像对。其中，用于特征匹配的特征匹配器可以是由构建的多层神经网络结构(SuperGlue)通过大量的牙齿数据集训练得到的，该牙齿数据集可以是上述训练完成的特征提取器输出的特征数据，特征匹配器的具体训练过程在此不作限定。
S230、根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构。
可理解的，在上述S220的基础上，得到第一匹配结果后，可以对第一匹配结果包括的多个子匹配结果进行过滤，过滤掉存在误匹配关系的子匹配结果。随后，基于过滤后的多个子匹配结果构建多个目标待重建图像之间的图型结构(graph结构)，多个目标待重建图像为过滤后的多个子匹配结果对应的待重建图像，过滤后的每个子匹配结果反应的是包括有效二维匹配点的图像对之间的匹配关系，该图像对包括的两个待重建图像即为目标待重建图像，也就是通过每两个待重建 图像间(图像对)的匹配关系建立多个待重建图像之间的关联关系，以完成二维到三维的转换。
S240、基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果。
其中，所述第一重建结果包括所述目标对象的三维数据。
可理解的，在上述S230的基础上，得到多个目标待重建图像间的图型结构后，基于图型结构进行三维重建，得到目标对象的第一重建结果，第一重建结果包括目标对象的三维数据(三维模型)，三维数据可以是三维网格数据或者三维点云数据，具体重建出的三维数据类型不做限定。
本公开实施例提供了一种三维重建方法，用户通过具有摄像功能的可移动终端拍摄目标对象，并生成多个待重建图像，操作流程简单，对操作环境以及操作人员的要求比较低，操作设备的成本也比较低。随后，服务器获取终端生成的目标对象的多个待重建图像，提取多个待重建图像中每个待重建图像的特征数据，并在多个待重建图像之间基于特征数据进行两两匹配，得到和多个待重建图像至少部分待重建图像相关的第一匹配结果，匹配结果反应的是图像对的二维匹配关系；根据第一匹配结果，构建多个待重建图像中多个目标待重建图像间的图型结构；最后基于图型结构进行三维重建，得到目标对象的第一重建结果，完成二维数据到三维数据的转换，其中，第一重建结果包括目标对象的重建效果较好的三维数据。本公开提供的方法，通过具有摄像功能的可移动终端获取目标对象的多个待重建图像，基于多个待重建图像完成目标对象的三维重建，简化了操作流程，便于实施，降低了设备成本，且有效解决了三维重建效果不佳的问题。
在上述实施例的基础上，图3为本公开实施例提供的一种三维重建方法的流程示意图，可选的，所述根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构，具体包括如 图3所示的如下步骤S310至S320：
S310、对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果。
可理解的，得到第一匹配结果后，将第一匹配结果包括的多个子匹配结果中存在误匹配关系的子匹配结果过滤出去，得到第二匹配结果，第二匹配结果包括的子匹配结果的数量要小于或等于第一匹配结果包括的子匹配结果的数量，过滤存在误匹配关系的子匹配结果，提高匹配率，可以减少后续进行三维重建的数据量，提高三维重建效率，并提高三维重建的效果。
可选的，上述S310得到第二匹配结果具体可以通过如下步骤实现：
统计所述第一匹配结果包括的每个子匹配结果中的匹配点信息，以对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第三匹配结果。
确定所述第三匹配结果包括的每个子匹配结果对应的待重建图像间的对应关系，以对所述第三匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果。
可理解的，统计第一匹配结果包括的每个子匹配结果中的匹配点信息，子匹配结果是对两个待重建图像(图像对)的特征数据进行匹配后得到的，子匹配结果包括图像对之间匹配点的信息，匹配点也就是匹配上的特征点，匹配点信息包括匹配点的数量和分布，将多个子匹配结果中匹配点的数量和分布小于或等于预设阈值的子匹配结果删除，例如，反应图像1和图像2特征点之间的匹配关系的子匹配结果1中包括的匹配点数量和分布小于预设阈值，则说明图像1和图像2是误匹配，需要将子匹配结果1删除，以提高匹配率。通过匹配质量和阈值对第一匹配结果包括的多个子匹配结果进行第一次过滤，得到第三匹配结果，第三匹配结果包括的子匹配结果的数量要小于或等于第一匹配结果包括的子匹配结果的数量。完成第一次过滤后，通过双目对极几何算法确定第三匹配结果包括的每个子匹配结果对应的待重建 图像间的对应关系，以对第三匹配结果包括的多个子匹配结果进行第二次过滤，得到第二匹配结果，也就是通过计算的图像对之间的对应关系进一步过滤误匹配图像，其中，第二匹配结果包括的子匹配结果的数量要小于或等于第三匹配结果包括的子匹配结果的数量，第二匹配结果对应的待重建图像的数量也要小于或等于第三匹配结果对应的待重建图像的数量，第二匹配结果对应的待重建图像即为目标待重建图像，第二匹配结果对应的待重建图像是图像对匹配较为准确的图像。
S320、基于所述第二匹配结果构建多个目标待重建图像间的图型结构。
其中，所述多个目标待重建图像为所述第二匹配结果包括的多个子匹配结果对应的待重建图像。
可理解的，在上述S310的基础上，第一匹配结果经过两次过滤得到第二匹配结果后，通过图优化(pose-graph)算法，基于第二匹配结果包括的多个子匹配结果构建多个目标待重建图像间的图型结构，其中，pose-graph算法是对多帧图像进行局部优化，得出各帧图像的位姿，再计算得出相邻关键帧间的相对位姿的算法。
可选的，上述S320得到图型结构具体可以通过如下步骤实现：
确定所述第二匹配结果中每个子匹配结果对应的待重建图像的相机位置，并将所述相机位置作为目标顶点。
将所述每个子匹配结果中待重建图像之间的二维匹配关系作为目标边。
基于所述目标顶点和所述目标边，构建所述多个目标待重建图像间的图型结构。
可理解的，根据第二匹配结果中每个子匹配结果确定该匹配结果对应的每个待重建图像的相机位置，也就是通过匹配关系确定每个目标待重建图像的相机位置，具体的可以将某一目标待匹配图像对的相机位置作为目标顶点，将每个子匹配结果中待重建图像之间的二维匹配关系作为目标边，基于目标顶点和目标边，构建多个目标待重建图 像间的图型结构，也就是由目标顶点开始，将目标顶点对应的目标待重建图像作为第一帧图像，根据匹配关系逐渐计算出相邻帧图像之间的相对位姿，也就是一帧图像一帧图像的计算出每帧图像的相机位姿，按照估计出的相机的运行轨迹，构建具有有效匹配关系的目标待重建图像间的graph结构(图型结构)，在构建图型结构的过程中，还会根据每帧图像间的相对位姿进一步过滤出可能存在误匹配关系的图像。
本公开实施例提供了一种三维重建方法，通过对第一匹配结果包括的多个子匹配结果进行多次过滤，得到反应有效匹配关系的第二匹配结果，随后基于第二匹配结果构建多个目标待重建图像间的图型结构，具体的，统计第一匹配结果包括的每个子匹配结果中匹配点数量和分布小于或等于预设阈值的子匹配结果，进行将该复合条件的子匹配结果进行第一次过滤，得到第三匹配结果。随后再通过双目对极几何和pose-graph算法进一步过滤第三匹配结果中尚且存在误匹配关系的子匹配结果，得到第二匹配结果，同时根据子匹配结果确定其对应的待重建图像的相机位置，并将该相机位置作为目标顶点，图像和图像之间的二维匹配关系为目标边，构建具有有效匹配关系的多个目标待重建图像间的graph结构。本公开提供的方法，通过多次过滤存在误匹配关系的子匹配结果，得到第二匹配结果，提高了匹配率，同时完成多个目标待重建图像间的graph结构的构建，便于后续基于graph结构高效准确的构建目标对象的三维数据，完成二维数据和三维数据间的转换，为简化三维重建的操作流程提供技术支持。
在上述实施例的基础上，图4为本公开实施例提供的一种三维重建方法的流程示意图，可选的，所述基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，具体包括如图4所示的如下步骤S410至S420：
S410、基于所述图型结构进行稀疏重建，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数。
可理解的，基于图型结构进行目标对象的三维数据的稀疏重建，得到目标对象的第二重建结果，第二重建结果包括大量三维点，稀疏重建是基于二维特征点确定三维点的过程，在进行稀疏重建的过程中，还可以估计出多个目标待重建图像中每个目标待重建图像的相机参数，相机参数包括相机内参、相机外参和相机位姿，也就是拍摄的每张待重建图像都存在对应的相机参数，基于相机参数可以估计出相机的运行轨迹。
可选的，上述S410得到第二重建结果以及估计出多个目标带重建图像的相机参数具体可以通过如下步骤实现：
在所述图型结构对应的多个目标待重建图像中确定初始待重建图像，并对所述初始待重建图像进行双目重建，得到初始三维点。
基于所述初始三维点，优化所述图型结构中的相邻边以增加新的三维点，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数。
可理解的，通过图型结构可以获取多个含有有效二维匹配点的图像对，每个图像对是由两个目标待重建图像组成的，多个图像对即为图型结构对应的多个目标待重建图像组成的。随后在多个图像对中确定匹配质量最高的目标图像对，并将目标图像对包括的两个待重建图像作为初始待重建图像，匹配质量最高是指图像对之间匹配点的数量和分布最高。随后对初始待重建图像进行双目重建，得到牙齿三维模型的初始三维点，该初始三维点可以理解为起始三维点。确定初始三维点后，基于初始三维点，通过光束法平差优化算法(Bundle Adjustment，BA)增量式优化图型结构中的相邻边以增加新的三维点，也就是从初始三维点开始，基于图型结构中的二维特征点不断扩充牙齿三维模型的三维点，直至完成目标对象的稀疏重建，得到第二重建结果，同时在稀疏重建的过程中完成多个目标待重建图像中每个待重建图像的相机参数的估计。
S420、根据所述多个目标待重建图像的相机参数，对所述第二重 建结果进行优化，得到所述目标对象的第一重建结果。
可理解的，在上述S410的基础上，根据多个目标待重建图像中每个目标待重建图像的相机参数，基于第二重建结果中的三维点进行稠密重建，得到目标对象的第一重建结果，第一重建结果即为目标对象完整的三维网格数据。
可选的，上述S420得到第一重建结果具体通过如下步骤实现：
在所述第二重建结果的基础上，根据所述多个目标待重建图像的相机参数，构建所述目标对象的目标三维场景。
根据所述目标三维目标场景得到关于所述目标对象的神经辐射场景描述。
基于所述神经辐射场景描述生成所述目标对象的第一重建结果。
可理解的，得到第二重建结果后，根据估计的多个目标待重建图像的相机参数，通过神经辐射场(Neural Radiance Field，Nerf)非显式地对用户口内的场景进行建模，得到包括目标对象三维数据的目标三维场景，目标三维场景可以理解为牙齿三维场景，在Nerf训练完成后，可以将口内牙齿三维场景表示为一个输入为5D向量的函数，进而得到关于口内三维牙齿数据的神经辐射场描述，随后结合MachineCube算法在神经辐射场描述中抽取牙齿的三维数据，抽取的牙齿的三维数据即为第一重建结果，也就是对稀疏重建的第二重建结果进行稠密重建，将多个三维点进行补充描述，得到具有最佳重建效果的第一重建结果。
可理解的，得到第二重建结果后，通过SGM(Semi-Global Matching)或者传统的经典的PatchMatch算法，根据估计的多个目标待重建图像的相机参数在第二重建结果的基础上进行稠密重建，得到第一重建结果，对稀疏重建的第二重建结果的基础上进行稠密重建得到第一重建结果的实现方式不做限定，可根据用户需求自行确定。
本公开实施例提供了一种三维重建方法，基于图型结构进行稀疏重建，得到目标对象的第二重建结果，并估计出多个目标待重建图像的相机参数，其中，稀疏重建主要是重建出二维匹配点(二维特征点) 的三维点，确定一系列和目标对象相关的二维匹配点对应的三维点；随后根据多个目标待重建图像的相机参数，通过Nerf算法对第二重建结果进行稠密重建，以得到具有最佳三维重建效果的第一重建结果。
图5为本公开实施例提供的一种三维重建装置的结构示意图。本公开实施例提供的三维重建装置可以执行上述三维重建方法实施例提供的处理流程，如图5所示，三维重建装置500包括获取模块510、匹配模块520、构建模块530以及重建模块540，其中：
获取模块510，用于获取目标对象的多个待重建图像；
匹配模块520，用于基于提取的所述多个待重建图像的特征数据进行特征匹配，得到所述多个待重建图像的第一匹配结果；
构建模块530，用于根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构；
重建模块540，用于基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，其中，所述第一重建结果包括所述目标对象的三维数据。
可选的，重建模块540具体用于：
基于所述图型结构进行稀疏重建，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数；
根据所述多个目标待重建图像的相机参数，对所述第二重建结果进行优化，得到所述目标对象的第一重建结果。
可选的，匹配模块520具体用于：
提取所述多个待重建图像中每个待重建图像中包括的目标对象的特征，得到所述每个待重建图像的特征数据；
将所述多个待重建图像中每两个待重建图像的特征数据进行特征匹配，得到第一匹配结果。
可选的，构建模块530具体用于：
对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第二 匹配结果；
基于所述第二匹配结果构建多个目标待重建图像间的图型结构；
其中，所述多个目标待重建图像为所述第二匹配结果包括的多个子匹配结果对应的待重建图像。
可选的，构建模块530具体用于：
统计所述第一匹配结果包括的每个子匹配结果中的匹配点信息，以对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第三匹配结果；
确定所述第三匹配结果包括的每个子匹配结果对应的待重建图像间的对应关系，以对所述第三匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果。
可选的，构建模块530具体用于：
确定所述第二匹配结果中每个子匹配结果对应的待重建图像的相机位置，并将所述相机位置作为目标顶点；
将所述每个子匹配结果中待重建图像之间的二维匹配关系作为目标边；
基于所述目标顶点和所述目标边，构建所述多个目标待重建图像间的图型结构。
可选的，重建模块540具体用于：
在所述图型结构对应的多个目标待重建图像中确定初始待重建图像，并对所述初始待重建图像进行双目重建，得到初始三维点；
基于所述初始三维点，优化所述图型结构中的相邻边以增加新的三维点，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数。
可选的，重建模块540具体用于：
在所述第二重建结果的基础上，根据所述多个目标待重建图像的相机参数，构建所述目标对象的目标三维场景；
根据所述目标三维目标场景得到关于所述目标对象的神经辐射场 景描述；
基于所述神经辐射场景描述生成所述目标对象的第一重建结果。
图5所示实施例的三维重建装置可用于执行上述方法实施例的技术方案，其实现原理和技术效果类似，此处不再赘述。
图6为本公开实施例提供的一种电子设备的结构示意图。下面具体参考图6，其示出了适于用来实现本公开实施例中的电子设备600的结构示意图。本公开实施例中的电子设备600可以包括但不限于诸如移动电话、笔记本电脑、数字广播接收器、PDA(个人数字助理)、PAD(平板电脑)、PMP(便携式多媒体播放器)、车载终端(例如车载导航终端)、可穿戴电子设备等等的移动终端以及诸如数字TV、台式计算机、智能家居设备等等的固定终端。图6示出的电子设备仅仅是一个示例，不应对本公开实施例的功能和使用范围带来任何限制。
如图6所示，电子设备600可以包括处理装置(例如中央处理器、图形处理器等)601，其可以根据存储在只读存储器(ROM)602中的程序或者从存储装置608加载到随机访问存储器(RAM)603中的程序而执行各种适当的动作和处理以实现如本公开所述的实施例的三维重建方法。在RAM 603中，还存储有电子设备600操作所需的各种程序和数据。处理装置601、ROM 602以及RAM 603通过总线604彼此相连。输入/输出(I/O)接口605也连接至总线604。
通常，以下装置可以连接至I/O接口605：包括例如触摸屏、触摸板、键盘、鼠标、摄像头、麦克风、加速度计、陀螺仪等的输入装置606；包括例如液晶显示器(LCD)、扬声器、振动器等的输出装置607；包括例如磁带、硬盘等的存储装置608；以及通信装置609。通信装置609可以允许电子设备600与其他设备进行无线或有线通信以交换数据。虽然图6示出了具有各种装置的电子设备600，但是应理解的是，并不要求实施或具备所有示出的装置。可以替代地实施或具备更多或更少的装置。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以 被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括承载在非暂态计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码，从而实现如上所述的三维重建方法。在这样的实施例中，该计算机程序可以通过通信装置609从网络上被下载和安装，或者从存储装置608被安装，或者从ROM 602被安装。在该计算机程序被处理装置601执行时，执行本公开实施例的方法中限定的上述功能。
需要说明的是，本公开上述的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本公开中，计算机可读信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读信号介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：电线、光缆、RF(射频)等等，或者上述的任意合适的组合。
在一些实施方式中，客户端、服务器可以利用诸如HTTP(HyperText Transfer Protocol，超文本传输协议)之类的任何当前已知 或未来研发的网络协议进行通信，并且可以与任意形式或介质的数字数据通信(例如，通信网络)互连。通信网络的示例包括局域网(“LAN”)，广域网(“WAN”)，网际网(例如，互联网)以及端对端网络(例如，ad hoc端对端网络)，以及任何当前已知或未来研发的网络。
上述计算机可读介质可以是上述电子设备中所包含的；也可以是单独存在，而未装配入该电子设备中。
可选的，当上述一个或者多个程序被该电子设备执行时，该电子设备还可以执行上述实施例所述的其他步骤。
可以以一种或多种程序设计语言或其组合来编写用于执行本公开的操作的计算机程序代码，上述程序设计语言包括但不限于面向对象的程序设计语言—诸如Java、Smalltalk、C++，还包括常规的过程式程序设计语言—诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络——包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。
附图中的流程图和框图，图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，该模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组 合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本公开实施例中所涉及到的单元可以通过软件的方式实现，也可以通过硬件的方式来实现。其中，单元的名称在某种情况下并不构成对该单元本身的限定。
本文中以上描述的功能可以至少部分地由一个或多个硬件逻辑部件来执行。例如，非限制性地，可以使用的示范类型的硬件逻辑部件包括：现场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、片上系统(SOC)、复杂可编程逻辑设备(CPLD)等等。
在本公开的上下文中，机器可读介质可以是有形的介质，其可以包含或存储以供指令执行系统、装置或设备使用或与指令执行系统、装置或设备结合地使用的程序。机器可读介质可以是机器可读信号介质或机器可读储存介质。机器可读介质可以包括但不限于电子的、磁性的、光学的、电磁的、红外的、或半导体系统、装置或设备，或者上述内容的任何合适组合。机器可读存储介质的更具体示例会包括基于一个或多个线的电气连接、便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦除可编程只读存储器(EPROM或快闪存储器)、光纤、便捷式紧凑盘只读存储器(CD-ROM)、光学储存设备、磁储存设备、或上述内容的任何合适组合。
另外，本公开实施例还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行以实现上述实施例所述的三维重建方法。
此外，本公开实施例还提供了一种计算机程序产品，该计算机程序产品包括计算机程序或指令，该计算机程序或指令被处理器执行时实现如上所述的三维重建方法。
需要说明的是，在本文中，诸如“第一”和“第二”等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来， 而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上所述仅是本公开的具体实施方式，使本领域技术人员能够理解或实现本公开。对这些实施例的多种修改对本领域的技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本公开的精神或范围的情况下，在其它实施例中实现。因此，本公开将不会被限制于本文所述的这些实施例，而是要符合与本文所公开的原理和新颖特点相一致的最宽的范围。
工业实用性
本公开提供的三维重建方法，可使用具有拍摄功能的可移动终端采集牙颌数据，设备硬件成本低，使用成本和持有成本也比较低，基于采集的牙颌数据能够很好的构建牙齿三维模型，且三维重建的精度和效率也比较高，具有很强的工业实用性。

Claims (11)

  1. 一种三维重建方法，其特征在于，包括：
    获取目标对象的多个待重建图像；
    基于提取的所述多个待重建图像的特征数据进行特征匹配，得到所述多个待重建图像的第一匹配结果；
    根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构；
    基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，其中，所述第一重建结果包括所述目标对象的三维数据。
  2. 根据权利要求1所述的方法，其特征在于，所述基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，包括：
    基于所述图型结构进行稀疏重建，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数；
    根据所述多个目标待重建图像的相机参数，对所述第二重建结果进行优化，得到所述目标对象的第一重建结果。
  3. 根据权利要求1所述的方法，其特征在于，所述基于提取的所述多个待重建图像的特征数据进行特征匹配，得到第一匹配结果，包括：
    提取所述多个待重建图像中每个待重建图像中包括的目标对象的特征，得到所述每个待重建图像的特征数据；
    将所述多个待重建图像中每两个待重建图像的特征数据进行特征匹配，得到第一匹配结果。
  4. 根据权利要求1所述的方法，其特征在于，所述根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构，包括：
    对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果；
    基于所述第二匹配结果构建多个目标待重建图像间的图型结构；
    其中，所述多个目标待重建图像为所述第二匹配结果包括的多个子匹配结果对应的待重建图像。
  5. 根据权利要求4所述的方法，其特征在于，所述对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果，包括：
    统计所述第一匹配结果包括的每个子匹配结果中的匹配点信息，以对所述第一匹配结果包括的多个子匹配结果进行过滤，得到第三匹配结果；
    确定所述第三匹配结果包括的每个子匹配结果对应的待重建图像间的对应关系，以对所述第三匹配结果包括的多个子匹配结果进行过滤，得到第二匹配结果。
  6. 根据权利要求4所述的方法，其特征在于，所述基于所述第二匹配结果构建多个目标待重建图像间的图型结构，包括：
    确定所述第二匹配结果中每个子匹配结果对应的待重建图像的相机位置，并将所述相机位置作为目标顶点；
    将所述每个子匹配结果中待重建图像之间的二维匹配关系作为目标边；
    基于所述目标顶点和所述目标边，构建所述多个目标待重建图像间的图型结构。
  7. 根据权利要求2所述的方法，其特征在于，所述基于所述图型结构进行稀疏重建，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数，包括：
    在所述图型结构对应的多个目标待重建图像中确定初始待重建图像，并对所述初始待重建图像进行双目重建，得到初始三维点；
    基于所述初始三维点，优化所述图型结构中的相邻边以增加新的三维点，得到所述目标对象的第二重建结果，并估计出所述多个目标待重建图像的相机参数。
  8. 根据权利要求2所述的方法，其特征在于，所述根据所述多个目标待重建图像的相机参数，对所述第二重建结果进行优化，得到所述 目标对象的第一重建结果，包括：
    在所述第二重建结果的基础上，根据所述多个目标待重建图像的相机参数，构建所述目标对象的目标三维场景；
    根据所述目标三维目标场景得到关于所述目标对象的神经辐射场景描述；
    基于所述神经辐射场景描述生成所述目标对象的第一重建结果。
  9. 一种三维重建装置，其特征在于，包括：
    获取模块，用于获取目标对象的多个待重建图像；
    匹配模块，用于基于提取的所述多个待重建图像的特征数据进行特征匹配，得到所述多个待重建图像的第一匹配结果；
    构建模块，用于根据所述第一匹配结果，构建所述多个待重建图像中多个目标待重建图像间的图型结构；
    重建模块，用于基于所述图型结构进行三维重建，得到所述目标对象的第一重建结果，其中，所述第一重建结果包括所述目标对象的三维数据。
  10. 一种电子设备，其特征在于，包括：
    存储器；
    处理器；以及
    计算机程序；
    其中，所述计算机程序存储在所述存储器中，并被配置为由所述处理器执行以实现如权利要求1至8中任一所述的三维重建方法。
  11. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至8中任一所述的三维重建方法的步骤。
PCT/CN2023/117506 2022-09-09 2023-09-07 一种三维重建方法、装置、设备和存储介质 WO2024051783A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211104909.2 2022-09-09
CN202211104909.2A CN117689804A (zh) 2022-09-09 2022-09-09 一种三维重建方法、装置、设备和存储介质

Publications (1)

Publication Number Publication Date
WO2024051783A1 true WO2024051783A1 (zh) 2024-03-14

Family

ID=90127185

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/117506 WO2024051783A1 (zh) 2022-09-09 2023-09-07 一种三维重建方法、装置、设备和存储介质

Country Status (2)

Country Link
CN (1) CN117689804A (zh)
WO (1) WO2024051783A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111833447A (zh) * 2020-07-13 2020-10-27 Oppo广东移动通信有限公司 三维地图构建方法、三维地图构建装置及终端设备
CN112102475A (zh) * 2020-09-04 2020-12-18 西北工业大学 一种基于图像序列轨迹跟踪的空间目标三维稀疏重构方法
US20210358206A1 (en) * 2020-05-14 2021-11-18 Star Institute Of Intelligent Systems Unmanned aerial vehicle navigation map construction system and method based on three-dimensional image reconstruction technology
US20220051481A1 (en) * 2020-08-17 2022-02-17 Nvidia Corporation Three-dimensional model recovery from two-dimensional images
CN114067051A (zh) * 2021-11-16 2022-02-18 百度在线网络技术（北京）有限公司 三维重建处理方法、装置、电子设备以及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210358206A1 (en) * 2020-05-14 2021-11-18 Star Institute Of Intelligent Systems Unmanned aerial vehicle navigation map construction system and method based on three-dimensional image reconstruction technology
CN111833447A (zh) * 2020-07-13 2020-10-27 Oppo广东移动通信有限公司 三维地图构建方法、三维地图构建装置及终端设备
US20220051481A1 (en) * 2020-08-17 2022-02-17 Nvidia Corporation Three-dimensional model recovery from two-dimensional images
CN112102475A (zh) * 2020-09-04 2020-12-18 西北工业大学 一种基于图像序列轨迹跟踪的空间目标三维稀疏重构方法
CN114067051A (zh) * 2021-11-16 2022-02-18 百度在线网络技术（北京）有限公司 三维重建处理方法、装置、电子设备以及存储介质

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"Master Theses ", 15 January 2020, MONGOLIA INDUSTRIAL UNIVERSITY, CN, article HAN, CHAO: "SFM Algorithm of 3D Reconstruction from UAV Aerial Images", pages: 1 - 64, XP009553261 *
HAO, HAOJIE, ; LI CHAOKUI; FANG JUN; GAO MINGXIA;: "Improved AKAZE Feature Extraction Global SFM 3D Reconstruction Algorithm", JOURNAL OF GEOMATICS SCIENCE AND TECHNOLOGY, CN, vol. 38, no. 05, 15 October 2021 (2021-10-15), CN, pages 492 - 498, XP009553267, ISSN: 1673-6338 *

Also Published As

Publication number Publication date
CN117689804A (zh) 2024-03-12

Similar Documents

Publication Publication Date Title
CN110427917B (zh) 用于检测关键点的方法和装置
US10699431B2 (en) Method and apparatus for generating image generative model
CN107018293B (zh) 生成表示光场的数据的方法和装置
WO2022012231A1 (zh) 视频生成方法、装置、可读介质及电子设备
WO2023213254A1 (zh) 一种口内扫描处理方法、系统、电子设备及介质
WO2023030370A1 (zh) 内窥镜图像检测方法、装置、存储介质及电子设备
CN109754464B (zh) 用于生成信息的方法和装置
CN112927362A (zh) 地图重建方法及装置、计算机可读介质和电子设备
JP2021140780A (ja) 地図作成のためのコンピュータ実施方法及び装置、電子機器、記憶媒体並びにコンピュータプログラム
CN113327318B (zh) 图像显示方法、装置、电子设备和计算机可读介质
WO2023030427A1 (zh) 生成模型的训练方法、息肉识别方法、装置、介质及设备
WO2023241704A1 (zh) 牙齿模型的获取方法、装置、设备及介质
WO2023143129A1 (zh) 图像处理方法、装置、电子设备及存储介质
WO2023213253A1 (zh) 一种扫描数据处理方法、装置、电子设备及介质
US20140198177A1 (en) Realtime photo retouching of live video
WO2023213252A1 (zh) 扫描数据处理方法、装置、设备及介质
WO2024051783A1 (zh) 一种三维重建方法、装置、设备和存储介质
TW202221646A (zh) 深度檢測方法、電子設備及電腦可讀儲存媒體
WO2020155908A1 (zh) 用于生成信息的方法和装置
WO2023198101A1 (zh) 基于人工智能的口腔检测方法、装置、电子设备及介质
CN112800276A (zh) 视频封面确定方法、装置、介质及设备
WO2023098649A1 (zh) 视频生成方法、装置、设备及存储介质
CN115082636B (zh) 基于混合高斯网络的单图像三维重建方法及设备
WO2022083213A1 (zh) 图像生成方法、装置、设备和计算机可读介质
CN115546413A (zh) 一种基于便携摄像头的牙齿矫正效果监控方法、设备及存储介质

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: 23862477

Country of ref document: EP

Kind code of ref document: A1