Method and system for high-speed visualization of large-volume 3-dimensional data, and method and system for learning a visibility decision model used therefor

Abstract

The present invention provides a method for training a visibility determination model, comprising: specifying any two points in a virtual reality space; generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space; generating a visibility map at each of the plurality of nodes corresponding to the path; specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the visibility map; generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes; and training a visibility determination model using the training data.

Description

Technology Field

[0001] The present invention relates to a high-speed visualization method and system for large-capacity 3D data, and a method and system for learning a visibility judgment model used therein. Background Technology

[0002] With recent advancements in graphics hardware and software algorithms, real-time visualization of 3D data has become possible. Virtual Reality (VR) or Augmented Reality (AR) systems utilizing such real-time visualization support user interaction and require real-time rendering; therefore, high-performance rendering technology and optimized data structures for large-scale 3D models are essential.

[0003] In this regard, a method for visualizing 3D data in real time includes setting a guard band area having a predetermined distance range from the position of a virtual camera in the line of sight outside a visualization area representing the area of ​​objects that can be displayed on a screen among a plurality of objects included in graphics data, obtaining position information of each of the plurality of objects, using the obtained position information to identify the area where at least one object is located, and performing a process of removing part or all of at least one object data according to the identified area.

[0004] At this time, clipping operations that remove part of object data or culling operations that remove the whole are performed through the CPU (Central Processing Unit), so GPU (Graphinc Processing Unit) resources are not used efficiently.

[0005] Accordingly, there is a need for a visualization method for 3D data that prevents delays caused by rapid changes in view in CPU-based culling and is easy to apply to streaming visualization. The problem to be solved

[0006] The present invention relates to a high-speed visualization method and system for large-capacity 3D data, capable of providing efficient real-time visualization by rapidly identifying an object to be rendered according to the position or orientation of a virtual camera in a virtual reality space containing large-capacity 3D data, and a method and system for learning a visibility determination model used therein.

[0007] In addition, the present invention relates to a high-speed visualization method and system for large-capacity 3D data capable of implementing high-speed visualization of 3D data, and a method and system for learning a visibility judgment model used therein. means of solving the problem

[0008] To solve the problem described above, the method for training a visibility determination model according to the present invention may include: a step of specifying any two points in a virtual reality space; a step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space; a step of generating a visibility map at each of the plurality of nodes corresponding to the path, and a step of specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the visibility map; and a step of generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and training a pre-prepared visibility determination model using the training data.

[0009] In addition, the visibility judgment model learning system according to the present invention includes a storage unit in which information about a virtual reality space is stored; and a control unit for learning a visibility judgment model prepared in advance based on the information about the virtual reality space. The control unit specifies any two points in the virtual reality space, generates a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space, generates a visibility map at each of the plurality of nodes corresponding to the path, specifies a visibility map generation direction corresponding to each of the plurality of nodes based on the visibility map, generates learning data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and can learn a visibility judgment model prepared in advance using the learning data.

[0010] Additionally, a program stored on a computer-readable recording medium according to the present invention is executed by one or more processes in an electronic device and is a program stored on a computer-readable recording medium, wherein the program may include instructions for performing the steps of: specifying any two points in a virtual reality space; generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space; generating a visibility map at each of the plurality of nodes corresponding to the path and specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the visibility map; and generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and training a pre-prepared visibility judgment model using the training data.

[0011] Additionally, the visualization method according to the present invention may include the steps of: loading a virtual reality space on a user terminal; identifying location data corresponding to a specific node in the virtual reality space based on user input; inputting the location data into a pre-trained visibility determination model to identify an object that can be identified at a point corresponding to the specific node in the virtual reality space; and rendering the identified object and displaying it on the user terminal.

[0012] In addition, the visualization system according to the present invention includes an input unit that receives user input through a user terminal; and a control unit that renders a virtual reality space on the user terminal based on the user input. The control unit loads the virtual reality space on the user terminal, identifies location data corresponding to a specific node in the virtual reality space based on the user input, inputs the location data into a pre-trained visibility determination model to identify an object that can be identified at a point corresponding to the specific node from the virtual reality space, and renders the identified object to display it on the user terminal.

[0013] Additionally, a program stored on a computer-readable recording medium according to the present invention is executed by one or more processes in an electronic device and is a program stored on a computer-readable recording medium, wherein the program may include instructions for performing the steps of: loading a virtual reality space on a user terminal; identifying location data corresponding to a specific node in the virtual reality space based on user input; inputting the location data into a pre-learned visibility determination model to identify an object that can be identified at a point corresponding to the specific node in the virtual reality space; and rendering the identified object and displaying it on the user terminal. Effects of the invention

[0014] According to various embodiments of the present invention, a high-speed visualization method and system for large-capacity 3D data, and a visibility determination model learning method and system used therein, learn a visibility map by position in a virtual reality space, estimate a visibility map corresponding to a specific location in the virtual reality space according to a user terminal, and render an object corresponding to the visibility map based thereon, thereby enabling efficient real-time visualization by rapidly identifying an object to be rendered according to the position or orientation of a virtual camera in a virtual reality space containing large-capacity 3D data.

[0015] In addition, according to various embodiments of the present invention, a high-speed visualization method and system for large-capacity 3D data and a visibility determination model learning method and system used therein can achieve high-speed visualization of 3D data by learning a visibility map for consecutive positions based on a path between any two points in a virtual reality space and performing rendering of an object according to the visibility map in a virtual reality space based thereon. Brief explanation of the drawing

[0016] FIG. 1 illustrates a visibility judgment model learning system according to the present invention. FIG. 2 illustrates a visualization system according to the present invention. FIG. 3 is a flowchart illustrating a visibility judgment model learning method according to the present invention. FIGS. 4 and FIGS. 5 illustrate an embodiment of specifying any two points in a virtual reality space. FIG. 6 illustrates an example of generating a path connecting any two points. FIG. 7 illustrates an example of generating a visibility map. FIG. 8 illustrates an embodiment that specifies the visibility map generation direction for each of a plurality of nodes. FIGS. 9 and FIGS. 10 illustrate an example of training a visibility determination model. FIG. 11 is a flowchart illustrating a visualization method according to the present invention. FIG. 12 illustrates an example of rendering an object according to a visibility map. FIG. 13 is a block diagram illustrating the structure of a computing device that performs the visibility determination model learning method and visualization method of the present invention. Specific details for implementing the invention

[0017] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components are assigned the same reference number regardless of the drawing symbols, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not have distinct meanings or roles in themselves. Furthermore, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0018] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0019] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0020] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0021] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0022] FIG. 1 illustrates a visibility judgment model learning system according to the present invention. FIG. 2 illustrates a visualization system according to the present invention.

[0023] Referring to FIG. 1, the visibility determination model learning system (100) according to the present invention can specify any two points in a virtual reality space (121), generate a path connecting the two points specified above based on a plurality of nodes in the virtual reality space (121), and train a visibility determination model (123) using a visibility map corresponding to a plurality of nodes belonging to the path as learning data.

[0024] A virtual reality space (121) may be a space rendered through a user terminal, implemented based on 3D data that designs a virtual space such as CAD (Computer-Aided Design) and BIM (Building Information Modeling), and a virtual object implemented based on a predetermined physical law may be placed in such a virtual reality space (121).

[0025] For example, the virtual reality space (121) can be implemented by being driven through a user terminal so that a portion of the virtual reality space (121) is displayed through the display of the user terminal. At this time, the portion of the virtual reality space displayed through the display of the user terminal may be referred to as a viewport, and such a viewport may include an area captured through a virtual camera placed at any point in the virtual reality space (121).

[0026] In this regard, in one embodiment, the virtual reality space (121) may have a predetermined area where a virtual camera can be placed. In this case, the user terminal adjusts the position of the virtual camera based on user input, and the screen captured at the position where the virtual camera is placed can be displayed as a viewport.

[0027] For example, the virtual reality space (121) may be a space in which an avatar operated by a user terminal is implemented. In this case, a virtual camera corresponding to the user terminal may be positioned on one side of the avatar (for example, the head area or a point on a sphere having the avatar's location as the origin) to display the virtual reality space (121) to the user terminal.

[0028] This virtual reality space (121) may be replaced with a space based on Augmented Reality (AR), Mixed Reality (MR), and Extended Reality (XR), and thus the virtual reality space (121) can be understood to include an Augmented Reality space, a Mixed Reality space, and an Extended Reality space.

[0029] A node may represent any point (or location) in the virtual reality space (121). Multiple nodes may be set based on a predetermined distance interval or an interval based on travel time in the virtual reality space (121). For example, a node may be created according to a predetermined interval based on a point specified in the virtual reality space (121) according to user input. Alternatively, a node may represent multiple predetermined points (or locations) in the virtual reality space (121). Accordingly, location data in the virtual reality space (121) may be assigned to each node, and the location data may include location coordinates in the virtual reality space (121).

[0030] These nodes can be set in a space where a camera can be placed, and if the camera is dependent on a specific object (e.g., an avatar), they can be set in a space where that object can be placed (e.g., a movable area).

[0031] Meanwhile, the path may include multiple nodes that lead from one point to another among the multiple nodes set in the virtual reality space (121). That is, the path may include multiple nodes that the camera passes through during the process of moving from one point to another.

[0032] To this end, the visibility determination model learning system (100) can specify two points among a plurality of nodes and generate a path according to a predetermined path planning algorithm for the plurality of nodes set in the virtual reality space (121) and the two points specified above.

[0033] Here, the path planning algorithm can be implemented to generate and explore a path from a specific point to another point based on a starting point and a target point or multiple nodes, and for example, the path planning algorithm may utilize the Rapidly-exploring Random Tree (RRT) algorithm, Dijkstra algorithm, A*(A Star) algorithm, Reeds-Sheep Path algorithm, etc.

[0034] A visibility map may be provided to indicate objects rendered on a user terminal at a specific point in a virtual reality space (121). That is, the visibility map may indicate whether objects existing in a specific direction at a specific point are visible. For example, the visibility map may include an image taken from a specific point toward a specific direction using a virtual camera. In this case, the visibility map can determine visibility by identifying the objects appearing in the image.

[0035] Meanwhile, the visibility determination model (123) is a model trained using training data, and can be trained to output a visibility map corresponding to a specific node when a specific node is input. At this time, depending on the embodiment, the visibility determination model (123) may be trained to output an object to be rendered corresponding to the node, and in this case, the object to be rendered may be determined based on whether it is visible from the visibility map.

[0036] To this end, the visibility judgment model (123) can be trained based on training data consisting of training input data and correct answer data, wherein the training input data may include location data corresponding to a specific node (or, specific point) in the virtual reality space (121).

[0037] Additionally, the ground truth data serves as label data for the training input data and can represent a visibility map at a location corresponding to the training input data. In this case, the ground truth data may include a visibility map generated based on the visibility map generation direction for each of the multiple nodes belonging to the training input data.

[0038] Accordingly, the visibility determination model (123) can be trained to output a visibility map corresponding to the location data when a predetermined location data in the virtual reality space (121) is input.

[0039] Meanwhile, the visibility judgment model learning system (100) may include an input unit (110), a storage unit (120), a control unit (130), and an output unit (140).

[0040] The input unit (110) may receive information necessary for the operation of the visibility judgment model learning system (100) according to the present invention. To this end, the input unit (110) may be connected to a separate input device, server, or external storage device via a wireless or wired network.

[0041] Accordingly, the input unit (110) can receive information about the virtual reality space (121) from a separate input device, server, or external storage device, etc. For example, the input unit (110) can receive information about multiple nodes in the virtual reality space (121), and can also receive a visibility map corresponding to the multiple nodes.

[0042] Additionally, the input unit (110) may receive user input specifying a predetermined point in the virtual reality space (121), and may also receive a predetermined user input required in the process of training a visibility judgment model according to a plurality of nodes and a visibility map.

[0043] Additionally, the storage unit (120) may store commands and information necessary for the operation of the visibility determination model learning system (100) according to the present invention. For example, the storage unit (120) may store information about a virtual reality space, and may also store a plurality of nodes determined in the virtual reality space and a visibility map corresponding to each of the plurality of nodes.

[0044] Additionally, the storage unit (120) can store a visibility judgment model (123), and various information generated during the process of training the visibility judgment model (123) can be stored.

[0045] The control unit (130) can control the overall operation of the visibility judgment model learning system (100) according to the present invention. That is, the control unit (130) can train a visibility judgment model prepared in advance based on information about the virtual reality space.

[0046] To this end, the control unit (130) can specify any two points in the virtual reality space (121), generate a path connecting the two points specified above based on multiple nodes in the virtual reality space (121), and train a visibility judgment model (123) using a visibility map corresponding to multiple nodes in the path as training data.

[0047] Specifically, the control unit (130) can specify any two points in the virtual reality space (121). That is, the control unit (130) can specify any two points that satisfy a predetermined distance condition among the areas set as empty space in the virtual reality space (121).

[0048] Accordingly, the control unit (130) can generate a path along a plurality of nodes leading from one of the two previously specified points to another point based on a node representing a location in the virtual reality space (121).

[0049] To this end, the control unit (130) can set one of the two previously specified points as the starting point and the other point as the destination point, and can generate a path by specifying multiple nodes connecting from the starting point to the destination point based on a predetermined path planning algorithm.

[0050] Additionally, the control unit (130) can generate a visibility map at each of the multiple nodes corresponding to the previously generated path, and determine the direction for generating a visibility map corresponding to each of the multiple nodes based on the generated visibility map.

[0051] At this time, the control unit (130) can generate a visibility map according to a predetermined time at each of the plurality of nodes and specify the direction of generating the visibility map corresponding to each of the plurality of nodes based on the surface shape of one or more objects appearing on the visibility map.

[0052] Through this, the control unit (130) can generate training data using a plurality of nodes and a visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and train a pre-prepared visibility judgment model (123) using the training data.

[0053] At this time, the control unit (130) can perform interpolation for the multiple nodes based on the visibility map generation direction corresponding to each of the multiple nodes, and generate a visibility map corresponding to the interpolated node.

[0054] Furthermore, the control unit (130) can specify a plurality of previously generated nodes and a previously interpolated node as learning input data, and specify a visibility map corresponding to each of the plurality of previously generated nodes and a previously interpolated node as correct answer data, which is label data for the learning input data, thereby generating learning data consisting of the learning input data and the correct answer data.

[0055] Accordingly, the control unit (130) inputs learning input data into a pre-prepared visibility judgment model (123) to generate learning output data, and compares the learning output data with the correct answer data to train the visibility judgment model (123) based on the loss according to the comparison result.

[0056] At this time, the control unit (130) may generate training data different from the previously generated training data by specifying each of the two points specified in the virtual reality space (121) and arbitrary two points different from each other. Accordingly, the visibility judgment model learning system (100) may re-train the visibility judgment model (123) using multiple training data generated based on multiple different points.

[0057] The output unit (140) may output information generated by the operation of the visibility judgment model learning system (100) according to the present invention. To this end, the output unit (140) may be connected to a separate visual output device, server, or external storage device via a wireless or wired network.

[0058] Accordingly, the output unit (140) can output a virtual reality space (121), multiple nodes, and a visibility map, etc., so that the user can visually check them through a separate output device, server, or external storage device, and can also output various information generated during the process of training the visibility judgment model (123).

[0059] Additionally, depending on the embodiment, the output unit (140) may transmit various information generated during the process of training a virtual reality space (121), a plurality of nodes, a visibility map, and a visibility determination model (123) to another device.

[0060] Meanwhile, referring to FIG. 2, the visualization system (200) according to the present invention can identify location data in a virtual reality space (221) based on a user terminal (1) and input the identified location data into a pre-learned visibility determination model (223) to obtain a visibility map corresponding to a specific location. Accordingly, the visualization system (200) can render one or more objects based on the previously obtained visibility map and output them to the viewport of the user terminal (1).

[0061] Here, the user terminal (1) may be implemented to be able to access the virtual reality space (221), that is, the user terminal (1) may be implemented to operate the virtual reality space (221) based on user input.

[0062] In this regard, the user terminal (1) may be referred to as a tablet, personal computer, smartphone, and wearable device, and may render a virtual reality space (221) based on a pre-implemented process. At this time, the user terminal (1) may provide the virtual reality space (221) to the user through a display, or may be connected to a device provided to output the virtual reality space (221) and provide the virtual reality space (221) rendered through the user terminal (1) to the user.

[0063] Meanwhile, the viewport may represent a scene (or area) that is output to the user terminal (1) in a virtual reality space (221), and this may mean a scene that is output to the display of the user terminal (1).

[0064] Location data may indicate the location of a virtual camera that specifies a viewport according to the user terminal (1), and according to an embodiment, the location data may include attitude data indicating the direction of the virtual camera.

[0065] Additionally, the visibility determination model (223) is trained to output a visibility map corresponding to a location when predetermined location data is input. In one embodiment, the visibility determination model (223) may be trained by the visibility determination model learning system (100) according to the present invention.

[0066] Meanwhile, the visualization system (200) may include an input unit (210), a storage unit (220), a control unit (230), and an output unit (240).

[0067] The input unit (210) may receive information necessary for the operation of the visualization system (200) according to the present invention. To this end, the input unit (210) may be connected to a user terminal (1), a separate input device, a server, or an external storage device via a wireless or wired network.

[0068] Accordingly, the input unit (210) can receive user input for the virtual reality space (221) from a user terminal (1), a separate input device, a server, or an external storage device, etc. At this time, the user input may be input to manipulate the position of a virtual camera or the position of an avatar in the virtual reality space (221).

[0069] Additionally, the storage unit (220) may store commands and information necessary for the operation of the visualization system (200) according to the present invention. For example, the storage unit (220) may store information related to a virtual reality space (221). Additionally, the storage unit (220) may store a pre-learned visibility judgment model (223).

[0070] The control unit (230) can control the overall operation of the visualization system (200) according to the present invention. That is, the control unit (230) can render a virtual reality space (221) on the user terminal (1) based on user input through the user terminal (1), check location data in the virtual reality space (221) based on the user terminal (1), and input the previously checked location data into a pre-learned visibility judgment model (223) to obtain a visibility map corresponding to a specific location. Accordingly, the control unit (230) can render one or more objects based on the previously obtained visibility map and output them to the viewport of the user terminal (1).

[0071] Specifically, the control unit (230) loads the virtual reality space (221) from the user terminal (1) and can check location data corresponding to a specific node in the virtual reality space (221) based on user input. That is, the control unit (230) can check the location data of a virtual camera corresponding to the viewport of the user terminal (1) in the virtual reality space (221) implemented through the user terminal (1).

[0072] Accordingly, the control unit (230) inputs location data into a pre-learned visibility determination model (223) to identify an object that can be verified at a point corresponding to a specific node from the virtual reality space (221), and renders the previously identified object to display it on the user terminal (1).

[0073] To this end, the control unit (230) inputs previously identified location data into the visibility determination model (223) to obtain a visibility map corresponding to the input location data, and can identify one or more objects identifiable through the viewport of the user terminal (1) based on the obtained visibility map. Through this, the control unit (230) can render the previously identified one or more objects and display them on the user terminal (1).

[0074] The output unit (240) can output information generated by the operation of the visualization system (200) according to the present invention. To this end, the output unit (240) may be connected to the display of the user terminal (1), a separate visual output device, a server, or an external storage device via a wireless or wired network.

[0075] Accordingly, the output unit (240) can output a virtual reality space (221) so that the user can visually check it through the display of the user terminal (1), a separate output device, a server, or an external storage device, and, depending on the embodiment, may also transmit a certain scene of the virtual reality space (221) to another device.

[0076] Based on the configuration of the visibility judgment model learning system (100) and visualization system (200) examined above, the visibility judgment model learning method and visualization method will be explained in more detail below.

[0077] FIG. 3 is a flowchart illustrating a method for training a visibility determination model according to the present invention. FIG. 4 and FIG. 5 illustrate an embodiment for specifying any two points in a virtual reality space. FIG. 6 illustrates an embodiment for generating a path connecting any two points. FIG. 7 illustrates an embodiment for generating a visibility map. FIG. 8 illustrates an embodiment for specifying the visibility map generation direction of each of a plurality of nodes. FIG. 9 and FIG. 10 illustrate an embodiment for training a visibility determination model. FIG. 11 is a flowchart illustrating a visualization method according to the present invention. FIG. 12 illustrates an embodiment for rendering an object according to a visibility map.

[0078] Referring to FIG. 3, the visibility determination model learning system (100) according to the present invention can specify any two points in a virtual reality space (S100).

[0079] Specifically, the visibility judgment model learning system (100) can identify any two points that satisfy predetermined distance conditions among the areas set as empty spaces in the virtual reality space.

[0080] For example, with reference to FIG. 4, the visibility determination model learning system (100) can calculate a threshold distance (13) based on the size of the area set as empty space in the virtual reality space (10). That is, the visibility determination model learning system (100) can designate a distance corresponding to a predetermined ratio as the threshold distance (13) according to the size of the empty space excluding the area occupied by a predetermined object (14) in the virtual reality space (10). According to an embodiment, the visibility determination model learning system (100) may also calculate the threshold distance (13) based on the size of the virtual reality space (10).

[0081] Accordingly, the visibility determination model learning system (100) can specify an arbitrary point (11) in a virtual reality space (10) and specify another arbitrary point (12) in an area that is outside the previously specified threshold distance (13) from the specified arbitrary point (11).

[0082] Referring to FIG. 5, as another example, a visibility determination model learning system (100) can identify any two points (11, 12) in a predetermined movable area (15) in a virtual reality space (10). At this time, the visibility determination model learning system (100) can set a threshold distance by considering the size of the movable area (15) and identify any two points (11, 12) based on the set threshold distance.

[0083] As another example, the visibility determination model learning system (100) can identify any two nodes among a plurality of predetermined nodes in a virtual reality space. In this case, the visibility determination model learning system (100) can set a threshold distance based on the size of the virtual reality space or the number of a plurality of nodes determined in the virtual reality space, and identify any two nodes based on the set threshold distance.

[0084] As another example, the visibility determination model learning system (100) can specify any single point in a predetermined first area in a virtual reality space and specify any single point in a second area that is different from the first area in the virtual reality space, thereby specifying any two points in the virtual reality space.

[0085] In this case, each of the first and second regions may be a region containing a specific object that is determined to be switched from a virtual reality space to another virtual reality space when a virtual reality space and another virtual reality space are hierarchically connected based on a specific object.

[0086] In one embodiment, a specific object may be an object that connects a specific space with another space, such as a door, and thus, the virtual reality space and the other virtual reality space may represent separate spaces. Additionally, these virtual reality space and the other virtual reality space may refer to different spaces predetermined within a single virtual reality space.

[0087] Referring again to FIG. 3, the visibility determination model learning system (100) according to the present invention can generate a path according to a plurality of nodes leading from one of the two previously specified points to another point based on a node representing a location in a virtual reality space (S200).

[0088] Specifically, the visibility determination model learning system (100) can set one of the two previously specified points as a starting point and the other point as a destination point, and generate a path by specifying multiple nodes connecting from the starting point to the destination point based on a predetermined path planning algorithm.

[0089] For example, with reference to FIG. 6, the visibility determination model learning system (100) can generate a plurality of nodes (16) based on a predetermined distance interval based on two previously specified points (11, 12). At this time, the visibility determination model learning system (100) can generate a plurality of nodes (16) in an empty space or a movable area in the virtual reality space (10).

[0090] Accordingly, the visibility determination model learning system (100) can generate a path (18) from one of the two previously specified points (11, 12) to another destination point (12) based on a plurality of nodes (16) using a path planning algorithm. Thus, the path (18) may include a plurality of nodes (17) connecting from the starting point (11) to the destination point (12).

[0091] At this time, the visibility determination model learning system (100), according to an embodiment, may generate a path (18) from one of the two previously specified starting points (11) to another destination point (12), and may generate a path different from the previously generated path (18) by changing the starting point (11) and the destination point (12) to each other. In this case, each of the two previously generated paths may be used to generate different training data.

[0092] As another example, the visibility determination model learning system (100) can set one of the two previously specified points as the starting point and the other point as the destination point, and can generate a path between the starting point and the destination point using a path planning algorithm based on multiple predetermined nodes in a virtual reality space.

[0093] Referring again to FIG. 3, the visibility determination model learning system (100) according to the present invention generates a visibility map at each of the plurality of nodes corresponding to the previously generated path, and can specify the direction for generating a visibility map corresponding to each of the plurality of nodes based on the generated visibility map (S300).

[0094] Specifically, the visibility determination model learning system (100) can generate a visibility map according to a predetermined time at each of a plurality of nodes and determine the direction of visibility map generation corresponding to each of the plurality of nodes based on the surface shape of one or more objects appearing on the visibility map.

[0095] Referring again to FIG. 6, for example, the visibility determination model learning system (100) can generate a visibility map at a point corresponding to the direction toward the next node during the process of exploring a path (18) from a specific node to the next node. That is, the visibility determination model learning system (100) can determine the location where a virtual camera is placed based on location data corresponding to a specific node, and determine the direction of the virtual camera based on a direction vector according to the location data of the specific node and the location data of the next node.

[0096] Accordingly, as shown in FIG. 7, the visibility determination model learning system (100) can generate a visibility map (33) to indicate whether there is visibility for a plurality of objects (35) existing in a part area (30) of a virtual reality space captured by a virtual camera (31) at a specific node. At this time, the visibility determination model learning system (100) can generate a visibility map by specifying the objects appearing in the image captured by the virtual camera (31) for the area (30) as objects for which visibility is secured.

[0097] Alternatively, the visibility determination model learning system (100) can generate a visibility map (33) by using a ray function according to a virtual camera (31) to identify an object (35) that has visibility secured by the ray function among the objects (35) existing in a previously specified direction at a previously specified location.

[0098] Additionally, the visibility determination model learning system (100) can analyze the surface shape of one or more objects appearing on an image (or visibility map) captured by a virtual camera (31). To this end, the visibility determination model learning system (100) can analyze the surface shape of the objects based on a Gauss Map.

[0099] That is, as shown in FIG. 8, the visibility determination model learning system (100) can estimate a normal vector corresponding to the surface shape (21) of an object on a unit sphere according to a Gaussian map (23), and determine the visibility map generation direction (25) based on the shape of the estimated normal vector.

[0100] In one embodiment, the visibility determination model learning system (100) may have a normal vector appear as a single specific point on a unit sphere according to a Gaussian map (23) when the surface shape (21) of an object at a specific node is planar, and in this case, the visibility map generation direction (25) for the node may be specified to be in the opposite direction of the normal vector.

[0101] In another embodiment, the visibility determination model learning system (100) may, when the surface shape (21) of an object at a specific node is a shape corresponding to the side surface of a cylinder, have a normal vector on a unit sphere according to a Gaussian map (23) appear along a perimeter having a radius within a predetermined threshold range for the radius of the unit sphere, and in this case, the visibility map generation direction (25) for the node may be specified to be opposite to the direction of summing the normal vectors.

[0102] In another embodiment, the visibility determination model learning system (100) may, when the surface shape (21) of an object at a specific node is a shape corresponding to the side surface of a cone, have a normal vector on a unit sphere according to a Gaussian map (23) appear along a perimeter having a radius smaller than a predetermined threshold range for the radius of the unit sphere, and in this case, the visibility map generation direction (25) for the node may be specified to be directed in the direction of summing the normal vectors.

[0103] In another embodiment, the visibility determination model learning system (100) may have normal vectors along the hemispherical surface on a unit sphere according to a Gaussian map (23) when the surface shape (21) of an object at a specific node appears in a hemispherical shape, and in this case, the visibility map generation direction (25) for the node may be specified to be opposite to the direction of summing the normal vectors.

[0104] In another embodiment, the visibility determination model learning system (100) may, when the surface shape (21) of an object at a specific node appears as at least a portion of the surface having a predetermined curvature in the unit sphere, a normal vector on the unit sphere according to the Gaussian map (23) may appear along the surface, and in this case, the visibility map generation direction (25) for the node may be specified to be opposite to the direction of summing the normal vectors.

[0105] Referring again to FIG. 3, the visibility determination model learning system (100) according to the present invention can generate learning data using a plurality of nodes and a visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and can train a pre-prepared visibility determination model using the learning data (S400).

[0106] Specifically, the visibility determination model learning system (100) can perform interpolation on a plurality of nodes based on the visibility map generation direction corresponding to each of the plurality of nodes, and generate a visibility map corresponding to the interpolated node.

[0107] For example, the visibility determination model learning system (100) may, at any one of the multiple nodes belonging to the previously generated path, generate one or more nodes along a visibility map generation direction specific to the node, and generate a visibility map for the generated one or more nodes. At this time, the one or more nodes generated along the visibility map generation direction may be generated between the specific node and a node adjacent to the specific node.

[0108] Furthermore, the visibility judgment model learning system (100) can specify a plurality of previously generated nodes and a previously interpolated node as learning input data, and specify a visibility map corresponding to each of the plurality of previously generated nodes and the previously interpolated node as correct data, which is label data for the learning input data, thereby generating learning data consisting of the learning input data and the correct data.

[0109] For example, with reference to FIG. 9, a visibility determination model learning system (100) may specify a plurality of nodes corresponding to a path according to any two points previously specified as learning input data (41), and specify a visibility map generated at each of the plurality of nodes as correct data (42). According to an embodiment, the visibility determination model learning system (100) may specify, together with the plurality of nodes, virtual camera pose data corresponding to the time when a visibility map was generated at each of the plurality of nodes as learning input data (41). Here, the pose data may include information indicating the direction of the virtual camera.

[0110] At this time, the visibility judgment model learning system (100) may generate training data different from the previously generated training data (40) by specifying each of the previously specified arbitrary two points again.

[0111] That is, the visibility judgment model learning system (100) can specify multiple nodes corresponding to a path different from the previously generated path as learning input data, and specify visibility maps generated from each of the multiple nodes belonging to the different path as correct data. At this time, the learning input data (41) and correct data (42) specified based on the previously generated path, and the learning input data and correct data specified based on the different path may be different learning data (40).

[0112] As another example, the visibility determination model learning system (100) may identify one or more objects identifiable from a visibility map corresponding to each of the previously generated multiple nodes and the previously interpolated nodes, and identify one or more identified objects as correct data (42), which is label data for the learning input data (41).

[0113] To this end, the visibility determination model learning system (100) can identify one or more objects identifiable in the visibility map based on a ray function in the direction of generating the visibility map corresponding to the node, based on the position of the node corresponding to the visibility map.

[0114] Furthermore, the visibility judgment model learning system (100) can generate learning output data (46) by inputting learning input data (41) into a pre-prepared visibility judgment model (45), and can train the visibility judgment model (45) based on the loss according to the comparison result by comparing the learning output data (46) with the correct answer data (42).

[0115] At this time, the visibility judgment model learning system (100) may generate learning data different from the previously generated learning data (40) by specifying each of the two points specified in the virtual reality space and two arbitrary points different from each other. Accordingly, the visibility judgment model learning system (100) may re-train the visibility judgment model (45) using multiple learning data generated based on multiple different points.

[0116] Additionally, as shown in FIG. 10, the visibility judgment model learning system (100) may sequentially input the multiple nodes (43, 44) belonging to the learning input data (41) into the visibility judgment model (47) according to the previously specified visibility map generation direction for the multiple nodes (43, 44) belonging to the learning input data (41) to generate learning output data (48), and compare the learning output data (48) with the correct data (42) to train the visibility judgment model (47) based on the loss according to the comparison result. In this case, the visibility judgment model (47) may be implemented to consider the previous node (43) according to the visibility map generation direction together during the process of inferring the learning output data (48) corresponding to the specific node (44).

[0117] Through this, the visibility determination model learning system (100) can train a visibility determination model (47) so that when location data corresponding to a predetermined node in a virtual reality space is input, a visibility map corresponding to the input location data or one or more objects identifiable according to the visibility map are output.

[0118] Referring to FIG. 11, the visualization system (200) according to the present invention loads a virtual reality space on a user terminal (S500) and can check location data corresponding to a specific node in the virtual reality space based on user input (S600).

[0119] Specifically, the visualization system (200) can check the position data of a virtual camera corresponding to the viewport of the user terminal in a virtual reality space implemented through the user terminal.

[0120] For example, the visualization system (200) can check the position data of a virtual camera corresponding to the viewport of a user terminal in a virtual reality space at predetermined time intervals. At this time, depending on the embodiment, the visualization system (200) may also check the previous position data confirmed in the previous time interval.

[0121] Additionally, the visualization system (200) can check the virtual camera's orientation data together during the process of checking the virtual camera's position data. In this case, the orientation data can indicate the direction of the virtual camera.

[0122] As another example, the visualization system (200) can check at least one of the virtual camera's position data and attitude data when movement of a predetermined distance interval is detected according to the movement of the virtual camera. That is, the visualization system (200) can check at least one of the virtual camera's position data and attitude data whenever the virtual camera moves by a predetermined distance interval.

[0123] The visualization system (200) according to the present invention inputs location data into a pre-learned visibility determination model to identify an object that can be identified at a point corresponding to a specific node in a virtual reality space (S700), and can render the previously identified object and display it on a user terminal (S800).

[0124] Specifically, the visualization system (200) inputs location data confirmed prior to the visibility determination model to obtain a visibility map corresponding to the input location data, and can identify one or more objects identifiable through the viewport of the user terminal based on the obtained visibility map.

[0125] For example, with reference to FIG. 12, when the visualization system (200) confirms the position data (53) of a virtual camera in a virtual reality space (51), it inputs the previously confirmed position data (53) into a pre-trained visibility judgment model (60) to obtain a visibility map (61) corresponding to the input position data (53). At this time, according to an embodiment, the visualization system (200) may input the pose data of the virtual camera along with the position data (53) into the visibility judgment model (60).

[0126] Accordingly, the visualization system (200) can identify one or more objects corresponding to the previously acquired visibility map (61) among the objects placed in the virtual reality space (51), and render the identified one or more objects to display on the user terminal (50).

[0127] As another example, when multiple location data are sequentially verified in a virtual reality space according to a predetermined time period (or distance interval), the visualization system (200) may input the previous location data along with the current location data into a visibility determination model to obtain a visibility map that takes into account the previous location data and the current location data.

[0128] Accordingly, the visualization system (200) can specify one or more objects to be rendered through a user terminal from the previously obtained visibility map, and furthermore, the visualization system (200) can render one or more specified objects based on the visibility map obtained from the visibility determination model in a virtual reality space.

[0129] Through the above configurations, the visibility determination model learning system (100) and visualization system (200) according to the present invention learn a visibility map by position in a virtual reality space, estimate a visibility map corresponding to a specific position in the virtual reality space according to a user terminal, and render an object corresponding to the visibility map based on this, thereby quickly identifying an object to be rendered according to the position or orientation of a virtual camera in a virtual reality space containing a large amount of 3D data, and providing efficient real-time visualization.

[0130] In addition, the visibility determination model learning system (100) and visualization system (200) according to the present invention can learn a visibility map for consecutive locations based on a path between any two points in a virtual reality space, and by performing rendering of an object based on the visibility map in a virtual reality space based thereon, high-speed visualization of 3D data can be achieved.

[0131] Furthermore, the visibility judgment model learning system (100) and the visualization system (200) according to the present invention are composed of a computing device and can perform at least one function related to the aforementioned visibility judgment model learning method and visualization method.

[0132] FIG. 13 is a block diagram illustrating the structure of a computing device that performs the visibility determination model learning method and visualization method of the present invention.

[0133] The computing device (1000) may include a user interface module (1001), a network communication module (1002), one or more processors (1003), data storage (1004), one or more camera(s) (1018), one or more sensors (1020) and a power system (1022), all of which may be connected to each other via a system bus, a network or other connection mechanism (1005).

[0134] The user interface module (1001) may be operable to transmit data to an external user input / output device and / or receive data from an external user input / output device.

[0135] For example, in the present invention, the visibility determination model learning system (100) receiving a plurality of nodes in a virtual reality space and a visibility map corresponding to each of the plurality of nodes, or the visualization system (200) receiving user input in a virtual reality space, may be performed by external input using a user interface module. At this time, the user interface module (1001) may include a touch screen, a computer mouse, a keyboard, a keypad, a touchpad, a trackball, a joystick, a voice recognition module, or other similar devices.

[0136] Additionally, the user interface module (1001) may also be configured to provide output to one or more user display devices, such as a cathode ray tube (CRT), liquid crystal display, light-emitting diode (LED), display using digital light processing (DLP) technology, or printer.

[0137] The user interface module (1001) may also be configured to produce an audible output using a device such as a speaker, a speaker jack, an audio output port, an audio output device, an earphone and / or other similar device.

[0138] The user interface module (1001) may additionally be configured with one or more tactile devices capable of generating tactile outputs, such as vibrations and / or other outputs, detectable by touch and / or physical contact with the computing device (1000).

[0139] The network communication module (1002) may include one or more devices that provide one or more wireless interface(s) (1007) and / or one or more wired interface(s) (1608) that are configurable to communicate through a network.

[0140] Additionally, the network communication module (1002) may be configured to provide reliable security and / or authenticated communication.

[0141] One or more processors (1003) may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processor, tensor processing unit (TPU), graphics processing unit (GPU), neural network processing unit (NPU), application integrated circuit, application semiconductor (ASIC), etc.). One or more processors (1003) may be configured to execute computer-readable instructions (1006) contained in data storage (1004) and / or other instructions described herein.

[0142] As an example of this, the learning and inference described in this specification are performed in a neural network processing unit (NPU), and efficiency can be increased by processing data operations at high speed with low power.

[0143] The data storage (1004) may include one or more non-transient computer-readable storage media that can be read and / or accessed by at least one of one or more processors (1003).

[0144] One or more computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, data storage (1004) may be implemented using a single physical device (e.g., one optical, magnetic, organic, or other memory or disk storage device), whereas in other examples, data storage (1004) may be implemented using two or more physical devices.

[0145] The data storage (1004) may include computer-readable instructions (1006) and additional data. The data storage (1004) may include a storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the device and network.

[0146] In some examples, the data storage (1004) may include a storage for the learned neural network model (1010) (e.g., a visibility determination model) described in the present invention.

[0147] Meanwhile, the computing device (1000) may include one or more camera(s) (1018), one or more sensors (1020) and / or a power system (1022).

[0148] The camera(s) (1018) can capture light and / or electromagnetic radiation emitted as visible light, infrared radiation, ultraviolet radiation and / or light of one or more other frequencies. The sensor (1020) may be configured to measure conditions within the computing device (1000) and / or conditions of the computing device (1000) environment and to provide data regarding these conditions. The power system (1022) may include one or more batteries (1024) and / or one or more external power interfaces (1026) for providing power to the computing device (1000).

[0149] Meanwhile, although the visibility judgment model learning system (100) and visualization system (200) of the present invention have been described above as being implemented as a computing device, the present invention is not limited thereto. For example, the functions of the neural network and / or computing device may be distributed among a plurality of computing clusters.

[0150] Furthermore, the present invention described above can be implemented as a program that is executed by one or more processes in an electronic device and stored on a computer-readable recording medium.

[0151] Accordingly, the present invention can be implemented as computer-readable code or instructions on a medium on which a program is recorded. That is, various control methods according to the present invention can be provided in the form of integrated or individual programs.

[0152] Meanwhile, computer-readable media include all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0153] Furthermore, the computer-readable medium may be a server or cloud storage that includes a storage and is accessible to an electronic device via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage via wired or wireless communication.

[0154] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, namely a CPU (Central Processing Unit), and no special limitations are placed on its type.

[0155] Meanwhile, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

Claims

Claim 1 A step of specifying any two points in a virtual reality space in a control unit; a step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space in the control unit; a step of generating a visibility map indicating the visibility status of a plurality of objects existing in a part of the virtual reality space captured by a virtual camera at each of the plurality of nodes corresponding to the path in the control unit; a step of specifying an object with secured visibility among objects existing in a specific direction based on the visibility map in the control unit, analyzing the surface shape of the object, and specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the surface shape of the object, wherein the normal vector or contour information is extracted from the surface shape, and the visibility map generation direction corresponding to each of the plurality of nodes is individually specified for each of the plurality of nodes based on the extracted normal vector or contour information. A method for training a visibility determination model, characterized by including the step of generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes in the control unit, and training a visibility determination model that is pre-prepared to receive location data corresponding to a specific node as input and output a visibility map representing an object that can be identified at that location using the training data. Claim 2 A method for training a visibility determination model according to claim 1, further comprising the step of re-specifying each of the two specified arbitrary points and two different arbitrary points in the control unit to generate training data different from the generated training data. Claim 3 A method for training a visibility determination model according to claim 1, wherein the step of training the visibility determination model comprises: a step of performing interpolation for the plurality of nodes based on a visibility map generation direction corresponding to each of the plurality of nodes in the control unit; and a step of generating a visibility map corresponding to the interpolated node in the control unit. Claim 4 A method for training a visibility determination model according to claim 3, wherein the step of training the visibility determination model further comprises the step of specifying a plurality of nodes corresponding to the path and the interpolated node as training input data in the control unit, and specifying a visibility map corresponding to each of the plurality of nodes corresponding to the path and the interpolated node as correct data which is label data for the training input data, thereby generating training data composed of the training input data and the correct data. Claim 5 A storage unit where information about the virtual reality space is stored; The control unit includes a control unit that trains a visibility determination model prepared in advance based on information regarding the virtual reality space, wherein the control unit specifies any two points in the virtual reality space and generates a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space, and generates a visibility map indicating the visibility status of a plurality of objects existing in a part of the virtual reality space captured by a virtual camera at each of the plurality of nodes corresponding to the path, and specifies an object with secured visibility among objects existing in a specific direction based on the visibility map, analyzes the surface shape of the object, and specifies a visibility map generation direction corresponding to each of the plurality of nodes based on the surface shape of the object, wherein normal vector or contour information is extracted from the surface shape, and the visibility map generation direction corresponding to each of the plurality of nodes is individually specified for each of the plurality of nodes based on the extracted normal vector or contour information, and generates training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and the training A visibility determination model training system that uses data to train a visibility determination model prepared in advance to receive location data corresponding to a specific node as input and output a visibility map representing objects that can be identified at that location. Claim 6 A program that is executed by one or more processes in an electronic device and stored on a computer-readable recording medium, wherein the program comprises: a step of specifying any two points in a virtual reality space; a step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space; a step of generating a visibility map indicating the visibility of a plurality of objects existing in a portion of the virtual reality space captured by a virtual camera at each of the plurality of nodes corresponding to the path; a step of specifying an object whose visibility is secured among objects existing in a specific direction based on the visibility map, analyzing the surface shape of the object, and specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the surface shape of the object, wherein the normal vector or contour information is extracted from the surface shape, and the visibility map generation direction corresponding to each of the plurality of nodes is individually specified for each of the plurality of nodes based on the extracted normal vector or contour information. A program stored on a computer-readable recording medium, characterized by including instructions for generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes, and training a visibility determination model prepared in advance to receive location data corresponding to a specific node as input and output a visibility map representing an object that can be identified at that location using the training data. Claim 7 A step of loading a virtual reality space from a user terminal in a control unit; a step of verifying location data corresponding to a specific node in the virtual reality space based on user input in the control unit; a step of inputting the location data into a pre-trained visibility determination model in the control unit to identify an object that can be verified at a point corresponding to the specific node in the virtual reality space; The method includes the step of rendering the specified object in the control unit and displaying it on the user terminal, wherein the pre-trained visibility determination model is trained according to a visibility determination model training method, and the visibility determination model training method comprises: the step of specifying any two points in a virtual reality space in the control unit; the step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space in the control unit; the step of generating a visibility map indicating the visibility status of a plurality of objects existing in a part of the virtual reality space captured based on a virtual camera at each of the plurality of nodes corresponding to the path in the control unit; and the step of specifying an object with secured visibility among objects existing in a specific direction based on the visibility map in the control unit, analyzing the surface shape of the object, and specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the surface shape of the object, wherein the method comprises extracting normal vector or contour information from the surface shape, and the method of specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the extracted normal vector or contour information. A step of individually specifying each node;A visualization method characterized by including the step of generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes in the control unit, and training a visibility judgment model prepared in advance to receive location data corresponding to a specific node as input and output a visibility map representing an object that can be identified at that location using the training data. Claim 8 delete Claim 9 An input unit that receives user input through a user terminal; The control unit includes a control unit that renders a virtual reality space on a user terminal based on the user input, wherein the control unit loads the virtual reality space on the user terminal, verifies location data corresponding to a specific node in the virtual reality space based on the user input, inputs the location data into a pre-trained visibility determination model to identify an object that can be identified at a point corresponding to the specific node in the virtual reality space, and renders the identified object to display it on the user terminal, wherein the pre-trained visibility determination model is trained according to a visibility determination model training method, and the visibility determination model training method comprises: a step of identifying any two points in the virtual reality space in the control unit; a step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space in the control unit; a step of generating a visibility map indicating the visibility status of a plurality of objects existing in a part of the virtual reality space captured by a virtual camera at each of the plurality of nodes corresponding to the path in the control unit; and a step of the control unit identifying an object among the objects existing in a specific direction for which visibility is secured based on the visibility map. A step of specifying, analyzing the surface shape of the object, and specifying a visibility map generation direction corresponding to each of the plurality of nodes based on the surface shape of the object, wherein the normal vector or contour information is extracted from the surface shape, and the visibility map generation direction corresponding to each of the plurality of nodes is individually specified for each of the plurality of nodes based on the extracted normal vector or contour information;A visualization system characterized by including the step of generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes in the control unit, and training a visibility judgment model prepared in advance to receive location data corresponding to a specific node as input and output a visibility map representing an object that can be identified at that location using the training data. Claim 10 A program that is executed by one or more processes in an electronic device and stored on a computer-readable recording medium, wherein the program comprises: a step of loading a virtual reality space from a user terminal in a control unit; a step of identifying location data corresponding to a specific node in the virtual reality space based on user input in the control unit; and a step of inputting the location data into a pre-learned visibility determination model in the control unit to identify an object that can be identified at a point corresponding to the specific node in the virtual reality space. The method includes instructions for performing the step of rendering the specified object and displaying it on the user terminal in the control unit, wherein the pre-trained visibility determination model is trained according to a visibility determination model training method, and the visibility determination model training method comprises: a step of specifying any two points in a virtual reality space in the control unit; a step of generating a path according to a plurality of nodes extending from one of the two points to another point based on a node representing a location in the virtual reality space in the control unit; a step of generating a visibility map indicating the visibility status of a plurality of objects existing in a part of the virtual reality space captured based on a virtual camera at each of the plurality of nodes corresponding to the path in the control unit; a step of specifying an object with secured visibility among objects existing in a specific direction based on the visibility map in the control unit, analyzing the surface shape of the object, and specifying a direction for generating a visibility map corresponding to each of the plurality of nodes based on the surface shape of the object, wherein normal vector or contour information is extracted from the surface shape, and a visibility map corresponding to each of the plurality of nodes is generated based on the extracted normal vector or contour information. A step of individually specifying the generation direction for each of the plurality of nodes;A program stored on a computer-readable recording medium, characterized by including the step of generating training data using the plurality of nodes and the visibility map corresponding to the visibility map generation direction for each of the plurality of nodes in the control unit, and training a visibility determination model prepared in advance to receive location data corresponding to a specific node as input and output a visibility map representing an object that can be identified at that location using the training data.

Images