Artificial intelligence-based three-dimensional model processing method and related device

Abstract

The invention is suitable for the technical field of data processing, provides a three-dimensional model processing method based on artificial intelligence and a related device, and aims to realize article splitting and identification of a scene three-dimensional model in the field of building decoration, obtain more and smaller article models and improve the utilization rate of the scene three-dimensional model in the field of building decoration. The method mainly comprises the steps of obtaining a target three-dimensional model, wherein the target three-dimensional model is a scene three-dimensional model in the field of building decoration; splitting each sub-model in the target three-dimensional model to obtain a sub-model set comprising a plurality of sub-models; removing the large hard decoration models in the sub-model set to obtain a residual sub-model set; performing multidirectional photographing on each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model; and inputting the photo set of each target sub-model into a preset artificial intelligence model for article identification to obtain a target article name of each target sub-model.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and in particular relates to a three-dimensional model processing method and related devices based on artificial intelligence. Background Art

[0002] In the field of building renovation, there are many large 3D scene models. These scene models are used to show users the effects of a renovated building, such as shopping malls, office buildings, and apartment designs of specific styles. These scene models contain many smaller object models, such as tables, chairs, bookcases, coffee tables, and screens. However, many large 3D scene models do not categorize the object models within them. Typically, all object models within a scene 3D model are simply referred to as "models," which leads to a failure to fully utilize the object models. Summary of the Invention

[0003] The purpose of this application is to provide an artificial intelligence-based three-dimensional model processing method and related devices to realize the object splitting and identification of scene three-dimensional models in the field of architectural decoration, obtain more and smaller object models, and improve the utilization rate of scene three-dimensional models in the field of architectural decoration.

[0004] In a first aspect, the present application provides a three-dimensional model processing method based on artificial intelligence, comprising:

[0005] Acquire a target three-dimensional model, where the target three-dimensional model is a scene three-dimensional model in the field of building decoration;

[0006] Splitting each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models;

[0007] Eliminating large hard decoration models from the sub-model set to obtain a remaining sub-model set;

[0008] Taking multi-directional photos of each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model;

[0009] The photo collection of each target sub-model is input into a preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to recognize the name of an object based on photos of multiple orientations of the object.

[0010] Optionally, the method further includes:

[0011] Use the object name to name the corresponding target sub-model to obtain the target naming sub-model;

[0012] All target named sub-models are classified and saved to obtain a target named sub-model set of the target three-dimensional model.

[0013] Optionally, excluding large hard decoration models from the sub-model set includes:

[0014] Defining the size standard of the large hard decoration model;

[0015] In the sub-model set, a sub-model that is larger than or equal to the size standard is identified as the large hard decoration model;

[0016] Eliminate the large hard decoration model from the sub-model set.

[0017] Optionally, after obtaining the remaining sub-model set and before taking multi-directional photos of each target sub-model in the remaining sub-model set, the method further includes:

[0018] Calculating the number of planes of each sub-model in the remaining sub-model set to obtain the number of planes of each sub-model in the remaining sub-model set;

[0019] Delete the sub-models whose number of planes is less than the preset number to obtain a new set of remaining sub-models;

[0020] The new remaining sub-model is regarded as the remaining sub-model.

[0021] Optionally, after obtaining the remaining sub-model set and before taking multi-directional photos of each target sub-model in the remaining sub-model set, the method further includes:

[0022] generating an enclosing polyhedron for each target sub-model in the set of remaining sub-models, wherein the enclosing polyhedron is a minimum polyhedron enclosing the target sub-model;

[0023] Splicing two or more target sub-models having an intersection of the bounding polyhedrons in the remaining sub-model set respectively to obtain a new target sub-model;

[0024] The new target sub-model is regarded as the target sub-model in the set of remaining sub-models.

[0025] Optionally, the enclosing polyhedron is specifically a minimum cube enclosing the target sub-model.

[0026] Optionally, taking multi-directional photos of each target sub-model in the remaining sub-model set includes:

[0027] Each target sub-model in the remaining sub-model set is photographed at least X times in a circle, where X is an integer greater than or equal to 9.

[0028] In a second aspect, the present application provides a three-dimensional model processing device based on artificial intelligence, comprising:

[0029] An acquisition unit, configured to acquire a target three-dimensional model, wherein the target three-dimensional model is a three-dimensional model of a scene in the field of building decoration;

[0030] a splitting unit, configured to split each sub-model in the target three-dimensional model to obtain a sub-model set comprising a plurality of sub-models;

[0031] a culling unit, configured to cull large hard decoration models from the sub-model set to obtain a remaining sub-model set;

[0032] a photographing unit, configured to take multi-directional photographs of each target sub-model in the remaining sub-model set to obtain a set of photographs of each target sub-model;

[0033] The recognition unit is used to input the photo set of each target sub-model into a preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to recognize the name of the object based on multiple orientation photos of the object.

[0034] Optionally, the device further includes:

[0035] a naming unit, configured to use the object name to name the corresponding target sub-model, thereby obtaining a target named sub-model;

[0036] The storage unit is used to classify and store all target named sub-models to obtain a target named sub-model set of the target three-dimensional model.

[0037] Optionally, when the elimination unit eliminates the large hard decoration model in the sub-model set, it is specifically used to:

[0038] Defining the size standard of the large hard decoration model;

[0039] In the sub-model set, a sub-model that is larger than or equal to the size standard is identified as the large hard decoration model;

[0040] Eliminate the large hard decoration model from the sub-model set.

[0041] Optionally, the device further includes:

[0042] a calculating unit, configured to calculate the number of planes of each sub-model in the remaining sub-model set, and obtain the number of planes of each sub-model in the remaining sub-model set;

[0043] A deletion unit is used to delete sub-models whose number of planes is less than a preset number to obtain a new set of remaining sub-models;

[0044] The method is regarded as a unit, configured to regard the new remaining sub-model as the remaining sub-model.

[0045] Optionally, the device further includes:

[0046] a generating unit, configured to generate an enclosing polyhedron for each target sub-model in the set of remaining sub-models, wherein the enclosing polyhedron is a minimum polyhedron enclosing the target sub-model;

[0047] a splicing unit, configured to splice two or more target sub-models having an intersection of the bounding polyhedrons in the remaining sub-model set to obtain a new target sub-model;

[0048] The considering unit is further configured to consider the new target sub-model as the target sub-model in the set of remaining sub-models.

[0049] Optionally, the enclosing polyhedron is specifically a minimum cube enclosing the target sub-model.

[0050] Optionally, when the photographing unit takes multi-directional photographs of each target sub-model in the remaining sub-model set, it is specifically configured to:

[0051] Each target sub-model in the remaining sub-model set is photographed at least X times in a circle, where X is an integer greater than or equal to 9.

[0052] In a third aspect, the present application provides a computer device, comprising:

[0053] Processor, memory, bus, input and output interface, network interface;

[0054] The processor is connected to the memory, the input / output interface, and the network interface via a bus;

[0055] The memory stores a program;

[0056] When the processor executes the program stored in the memory, it implements the three-dimensional model processing method based on artificial intelligence described in any one of the first aspects above.

[0057] In a fourth aspect, the present application provides a computer storage medium, wherein the computer storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the artificial intelligence-based three-dimensional model processing method described in any one of the first aspects above.

[0058] In a fifth aspect, the present application provides a computer program product, which, when executed on a computer, enables the computer to execute any one of the artificial intelligence-based three-dimensional model processing methods described in the first aspect above.

[0059] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0060] The embodiment of the 3D model processing method based on artificial intelligence of the present application obtains a target 3D model to obtain a scene 3D model in the field of building decoration that needs to be processed; then, each sub-model in the target 3D model is split to obtain a sub-model set containing several sub-models; large hard decoration models in the sub-model set are eliminated, which are usually not universal and can be discarded for recognition to reduce computing power expenditure, thereby obtaining a remaining sub-model set; each target sub-model in the remaining sub-model set is photographed from multiple angles to understand the target sub-model from multiple angles, thereby obtaining a photo set of each target sub-model; each photo set of the target sub-model is input into a preset artificial intelligence model for object recognition, thereby obtaining the target object name of each sub-model, wherein the preset artificial intelligence model is a model that has been trained to recognize the object name based on multiple angle photos of the object. It can be seen that the above embodiment realizes the object splitting and recognition of the scene 3D model in the field of building decoration, and can obtain more and smaller object models. These object models with accurate names can be taken to other scenes in need for arrangement and reuse, thereby improving the utilization rate of the scene 3D model in the field of building decoration. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of an embodiment of the artificial intelligence-based three-dimensional model processing method of the present application;

[0062] Figure 2 This is a flow chart of another embodiment of the artificial intelligence-based three-dimensional model processing method of the present application;

[0063] Figure 3 This is a flow chart of another embodiment of the artificial intelligence-based three-dimensional model processing method of the present application;

[0064] Figure 4 This is a flow chart of another embodiment of the artificial intelligence-based three-dimensional model processing method of the present application;

[0065] Figure 5 This is a schematic structural diagram of an embodiment of an artificial intelligence-based three-dimensional model processing device of the present application;

[0066] Figure 6 This is a schematic structural diagram of another embodiment of the artificial intelligence-based three-dimensional model processing device of the present application;

[0067] Figure 7 This is a schematic diagram of the structure of an embodiment of the computer device of the present application;

[0068] Figure 8 This is a schematic diagram of a partial effect of an embodiment of a three-dimensional scene model in the field of building decoration of this application;

[0069] Figure 9 for Figure 8 Schematic diagram of an embodiment of a multi-directional photograph of a chair in a 3D model of a scene. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0071] With the rapid development and maturity of artificial intelligence (AI), AI technology has been introduced into the processing of 3D scene models in the field of architectural decoration. This embodiment provides a technical solution for splitting, identifying, and classifying object models within large 3D scene models. This allows for the extraction of more and smaller object models from the large 3D scene model. These object models, with accurate names, can be re-used in other scenes where needed, thereby improving the utilization rate of 3D scene models in the field of architectural decoration.

[0072] The three-dimensional model processing method based on artificial intelligence in the present application runs in a three-dimensional model processing system, which mainly includes: a central processing unit module, a memory module, an artificial intelligence chip module, a communication module, a power supply module, etc. The above-mentioned modules are connected through a printed circuit board (PCB) to obtain a so-called three-dimensional model processing system; wherein, the central processing unit is mainly used to comprehensively dispatch the various component modules connected to it to implement the execution of the three-dimensional model processing method based on artificial intelligence mentioned in this embodiment; the power supply module is mainly used to provide a suitable working voltage to the central processing unit module, the memory module, the artificial intelligence chip module, the communication module, etc. The power supply module can be a self-contained battery-powered mode or a mains power supply mode, and the power source of the power supply module is not limited here; the memory module mainly stores the program for implementing the relevant steps of this embodiment, and the memory also stores a preset trained artificial intelligence model; the artificial intelligence chip module is used to quickly load and run the artificial intelligence model stored in the memory module; the communication module is mainly used to communicate with the outside world (mobile phones, tablets, computers and other electronic devices) to obtain the target three-dimensional model, etc.

[0073] Based on the understanding of the above hardware system, please refer to Figure 1 An embodiment of the present application's artificial intelligence-based three-dimensional model processing method includes:

[0074] 101. Obtain a target three-dimensional model, where the target three-dimensional model is a three-dimensional model of a scene in the field of building decoration.

[0075] It should be noted that the scene 3D model in the field of building decoration may be generated by different 3D software models, and the file format of the scene 3D model may also be varied, which is not specifically limited here. For example, the target 3D model is created by 3D modeling software such as 3D Studio Max and AutoCAD (Autodesk Computer Aided Design) developed by Autodesk, and the saved file format is max, DWG, etc. A schematic diagram of a partial effect of the target 3D model is implemented as follows Figure 8 shown.

[0076] 102. Split each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models.

[0077] It is understandable that since the target 3D model in step 101 is a scene 3D model in the field of building decoration, the scene must contain many sub-models. Figure 8 The schematic diagram of a partial rendering of the target 3D model shows sub-models such as a chair, dining table, plate, door, potted plant, sofa, coffee table, carpet, and chandelier. However, in the field of architectural decoration, these sub-models are often not given corresponding names in 3D scene models; they are often simply referred to as "model" or "sub-model." This step decomposes each sub-model within the target 3D model, creating a sub-model set containing several sub-models.

[0078] 103. Eliminate large hard decoration models from the sub-model set to obtain the remaining sub-model set.

[0079] It is understandable that the target 3D model is a large scene 3D model in the field of building decoration, such as Figure 8As shown, large-scale hard decoration models such as partition walls, suspended ceilings, floors, and plaster lines are often customized to suit the current apartment structure and are not universal. This step removes these large-scale hard decoration models. For example, when the file format of the target three-dimensional model is max, this step can remove useless large-scale hard decorations through the maxscript script of 3D Studio Max. It can be understood that since this step does not know the name of the sub-model in the sub-model set, in order to accurately and quickly remove large-scale hard decoration models from the sub-model set, this step usually adopts the following strategy: define the size standard of large-scale hard decoration models, for example, define the size standard of large-scale hard decoration models as any one of the length, width, and height exceeding 3 meters; in the sub-model set, identify sub-models that are greater than or equal to the size standard as large-scale hard decoration models, and then remove large-scale hard decoration models that meet the size standard from the sub-model set to obtain the remaining sub-model set.

[0080] 104. Take multi-directional photos of each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model.

[0081] In order to accurately identify each target sub-model in the remaining sub-model set, this embodiment uses a trained artificial intelligence model. The artificial intelligence model of this embodiment mainly uses multi-angle photos of the object to accurately identify the name corresponding to the object; for this reason, this step requires taking multi-directional photos of each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model. Specifically, each target sub-model in the remaining sub-model set is photographed at least X times in a circle, where X is an integer greater than or equal to 9. The more photos and the more angles, the better. Figure 9 As shown, Figure 9 Yes Figure 8 A multi-directional photograph of chair A in a 3D scene model is specifically as follows: chair A is the subject of the photograph, the other models in the 3D scene model are hidden, chair A is photographed from a 45-degree angle, and photographs are taken around the center of chair A at a preset camera distance. A photograph is taken at 45-degree intervals around the center of chair A until the photographed photo overlaps the first one, resulting in a collection of nine multi-directional photographs of chair A. It is understood that the method for taking multi-directional photographs of each target sub-model can be customized based on actual needs and is not specifically limited here. The method is based on facilitating accurate name recognition by the AI model.

[0082] 105. Input the photo collection of each target sub-model into the preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model.

[0083] Each target sub-model photo set in the remaining sub-model set in step 104 is input into a preset artificial intelligence model for object recognition, thereby obtaining the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to recognize object names based on photos of objects from multiple orientations. The training of the preset artificial intelligence model in this embodiment can collect photos of various real-world objects and / or object models from different angles as training samples. If necessary, the names of the objects or object models in the different training samples can be manually labeled and classified to facilitate training of the artificial intelligence model, thereby obtaining a trained model that can recognize object names based on photos of objects from multiple orientations. It is understood that the preset artificial intelligence model in this embodiment includes a deep learning model, a CNN model, etc., such as the YOLO (You Only Look Once) deep learning-based object detection algorithm. As can be seen, the above embodiment implements object segmentation and recognition of three-dimensional scene models in the field of architectural decoration, and can obtain more and smaller object models with accurate names. These object models with accurate names can be taken to other scenes in need and reused, thereby improving the utilization rate of three-dimensional scene models in the field of architectural decoration.

[0084] See also Figure 2 Another embodiment of the artificial intelligence-based three-dimensional model processing method of the present application includes:

[0085] 201. Obtain a target three-dimensional model, where the target three-dimensional model is a scene three-dimensional model in the field of building decoration.

[0086] The execution of this step is the same as the above Figure 1 The execution of step 101 in the embodiment is similar and will not be repeated here.

[0087] 202. Split each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models.

[0088] The execution of this step is the same as the above Figure 1 The execution of step 102 in the embodiment is similar and will not be repeated here.

[0089] In another possible embodiment, this step may also utilize a pre-customized artificial intelligence model to perform decomposition processing on the sub-models within the target three-dimensional model. The pre-customized artificial intelligence model is a model that has been trained to decompose a large three-dimensional model into a collection of several small models. The training of the pre-customized artificial intelligence model in this embodiment may collect large three-dimensional models of different scenes as training samples, and combine this with manual labeling and classification of the different small models within the large three-dimensional model to train the pre-customized artificial intelligence model, resulting in a trained model that can decompose a large three-dimensional model into a collection of several small models. The training process of the artificial intelligence model is a relatively mature technology, and the specific training process of the artificial intelligence model will not be described in detail here.

[0090] 203. Eliminate the large hard decoration model from the sub-model set to obtain the remaining sub-model set.

[0091] The execution of this step is the same as the above Figure 1 The execution of step 103 in the embodiment is similar and will not be repeated here.

[0092] It should be noted that simply defining the size standard of large hard decoration models to eliminate large hard decoration models from the model set is not enough. In smaller rooms (bathrooms, utility rooms, etc.), there are still many small hard decoration models such as partition walls, ceilings, floors, plaster lines, etc. These small hard decoration models are often customized to suit the current apartment structure and are not universal. This step also eliminates these small hard decoration models. This step further adopts the following strategy, please refer to Figure 3 :301. Calculate the number of planes for each sub-model in the remaining sub-model set to obtain the number of planes for each sub-model in the remaining sub-model set; that is, small item models usually have more planes, usually more than 20 planes, while hard decoration models such as partition walls, ceilings, and floors have fewer planes. For example, partition walls are usually simple cuboids (6 planes); 302. Delete sub-models with fewer planes than the preset number to obtain a new set of remaining sub-models; for example, the preset number is 20; 303. Treat the new remaining sub-models as remaining sub-models; it can be seen that Figure 3 The embodiment scheme can further effectively remove the smaller hard decoration models in the remaining sub-model set, streamline the number of models in the remaining sub-model set, reduce the computing power expenditure of subsequent steps, and improve the computing speed.

[0093] It should be noted that no matter how the sub-models in the target 3D model are split, the resulting sub-models may not necessarily accurately represent the shape of the object. For example, there may be problems with the split transition: a table may be missing a leg, a chair may be missing a leg, etc. Usually, there is no situation where the object model has other parts that are not fully split. In view of this, please refer to Figure 4 , this step can adopt the following strategy: 401. Generate an enclosing polyhedron for each target sub-model in the remaining sub-model set, the enclosing polyhedron is the minimum polyhedron enclosing the target sub-model, and the enclosing polyhedron can be specifically the minimum cube enclosing the target sub-model; 402. Separately splice two or more target sub-models that have intersections generated by the enclosing polyhedrons in the remaining sub-model set to obtain a new target sub-model, and try to restore the original appearance of the target sub-model; 403. Treat the new target sub-model as the target sub-model in the remaining sub-model set, thereby increasing the number of target sub-models in the remaining sub-model set, and the newly added target sub-model is likely to be the original appearance of the object model.

[0094] 204. Take multi-directional photos of each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model.

[0095] The execution of this step is the same as the above Figure 1 The execution of step 104 in the embodiment is similar and will not be repeated here.

[0096] 205. Input the photo collection of each target sub-model into a preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model.

[0097] The execution of this step is the same as the above Figure 1 The execution of step 105 in the embodiment is similar and will not be repeated here.

[0098] 206. Use the name of the object to name the corresponding target sub-model to obtain the target naming sub-model.

[0099] After knowing the item name of each target sub-model in the remaining sub-model set in step 205, this step can directly name the corresponding target sub-model using the item name to obtain the target named sub-model.

[0100] 207. Classify and save all target named sub-models to obtain a target named sub-model set of the target three-dimensional model.

[0101] This step can classify and save the corresponding names of the target named sub-models into the database, realize the object splitting and identification of the scene three-dimensional models in the field of architectural decoration, and obtain more and smaller object models. These object models with accurate names can be taken to other scenes in need for arrangement and reuse, thereby improving the utilization rate of scene three-dimensional models in the field of architectural decoration.

[0102] In another embodiment, it is also possible to further rely on manual or other trained artificial intelligence models to select target naming sub-models that do not match the shape of the object represented by the name based on the names corresponding to the target naming sub-models, thereby reducing the number of sub-models with incomplete shapes in the target naming sub-model set and improving the model quality of the target naming sub-models in the target naming sub-model set. Of course, the training process of the other trained artificial intelligence models in this embodiment is a relatively mature technology, and the specific artificial intelligence model training process will not be described in detail here.

[0103] The above embodiment describes the three-dimensional model processing method based on artificial intelligence of the present application. The following describes the three-dimensional model processing device based on artificial intelligence of the present application. Figure 5 An embodiment of the present application of a three-dimensional model processing device based on artificial intelligence includes:

[0104] An acquisition unit 501 is configured to acquire a target three-dimensional model, wherein the target three-dimensional model is a three-dimensional model of a scene in the field of building decoration;

[0105] a splitting unit 502, configured to split each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models;

[0106] A removal unit 503 is configured to remove large hard decoration models from the sub-model set to obtain a remaining sub-model set;

[0107] a photographing unit 504, configured to take multi-directional photographs of each target sub-model in the remaining sub-model set to obtain a set of photographs of each target sub-model;

[0108] The identification unit 505 is used to input the photo set of each target sub-model into a preset artificial intelligence model for object identification to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to identify the name of the object based on multiple orientation photos of the object.

[0109] The operations performed by the 3D model processing device based on artificial intelligence in this embodiment are similar to those in the aforementioned Figure 1 The operations performed in the embodiment are similar and will not be described again here.

[0110] See also Figure 6 Another embodiment of the three-dimensional model processing device based on artificial intelligence of the present application includes:

[0111] An acquisition unit 601 is configured to acquire a target three-dimensional model, wherein the target three-dimensional model is a three-dimensional model of a scene in the field of building decoration;

[0112] A splitting unit 602 is configured to split each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models;

[0113] A removal unit 603 is used to remove large hard decoration models from the sub-model set to obtain a remaining sub-model set;

[0114] A photographing unit 604 is configured to take multi-directional photographs of each target sub-model in the remaining sub-model set to obtain a set of photographs of each target sub-model;

[0115] The identification unit 605 is used to input the photo set of each target sub-model into a preset artificial intelligence model for object identification to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to identify the name of the object based on multiple orientation photos of the object.

[0116] Optionally, the device further includes:

[0117] A naming unit 606 is configured to use the object name to name the corresponding target sub-model to obtain a target named sub-model;

[0118] The saving unit 607 is configured to classify and save all target named sub-models to obtain a target named sub-model set of the target three-dimensional model.

[0119] Optionally, when the elimination unit 603 eliminates the large hard decoration model in the sub-model set, it is specifically used to:

[0120] Defining the size standard of the large hard decoration model;

[0121] In the sub-model set, a sub-model that is larger than or equal to the size standard is identified as the large hard decoration model;

[0122] Eliminate the large hard decoration model from the sub-model set.

[0123] Optionally, the device further includes:

[0124] a calculating unit 608, configured to calculate the number of planes of each sub-model in the remaining sub-model set, and obtain the number of planes of each sub-model in the remaining sub-model set;

[0125] A deleting unit 609 is configured to delete sub-models with fewer planes than a preset number to obtain a new set of remaining sub-models;

[0126] The treating unit 610 is configured to treat the new remaining sub-model as the remaining sub-model.

[0127] Optionally, the device further includes:

[0128] A generating unit 611 is configured to generate an enclosing polyhedron for each target sub-model in the set of remaining sub-models, wherein the enclosing polyhedron is a minimum polyhedron that encloses the target sub-model;

[0129] a splicing unit 612 for splicing two or more target sub-models having an intersection of the bounding polyhedrons in the remaining sub-model set to obtain a new target sub-model;

[0130] The considering unit 610 is further configured to consider the new target sub-model as the target sub-model in the set of remaining sub-models.

[0131] Optionally, the enclosing polyhedron is specifically a minimum cube enclosing the target sub-model.

[0132] Optionally, when the photographing unit 604 takes multi-directional photographs of each target sub-model in the remaining sub-model set, it is specifically configured to:

[0133] Each target sub-model in the remaining sub-model set is photographed at least X times in a circle, where X is an integer greater than or equal to 9.

[0134] The operations performed by the 3D model processing device based on artificial intelligence in this embodiment are similar to those in the aforementioned Figure 1 、 Figure 2 、 Figure 3 as well as Figure 4 The operations performed in the embodiment are similar and will not be described again here.

[0135] The computer device in the embodiment of the present application is described below. Figure 7 , an embodiment of the computer device in the embodiment of the present application includes:

[0136] The computer device 700 may include one or more processors (central processing units, CPU) 701 and a memory 702, in which one or more applications or data are stored. Among them, the memory 702 is volatile storage or persistent storage. The program stored in the memory 702 may include one or more modules, and each module may include a series of instruction operations in the computer device. Furthermore, the processor 701 may be configured to communicate with the memory 702 and execute a series of instruction operations in the memory 702 on the computer device 700. The computer device 700 may also include: one or more network interfaces 703, one or more input and output interfaces 704, and / or one or more operating systems, such as HarmonyOS, Windows Server, Mac OS, Unix, Linux, FreeBSD, etc. The processor 701 may execute the aforementioned Figures 1 to 4 The operations performed in any of the illustrated embodiments will not be described in detail here.

[0137] In the several embodiments provided in the embodiments of the present application, those skilled in the art should understand that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the unit is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0138] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0139] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements or improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A three-dimensional model processing method based on artificial intelligence, characterized in that: include: Acquire a target three-dimensional model, where the target three-dimensional model is a scene three-dimensional model in the field of building decoration; Splitting each sub-model in the target three-dimensional model to obtain a sub-model set including a plurality of sub-models; Eliminating large hard decoration models from the sub-model set to obtain a remaining sub-model set; Taking multi-directional photos of each target sub-model in the remaining sub-model set to obtain a photo set of each target sub-model; The photo collection of each target sub-model is input into a preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to recognize the name of an object based on photos of multiple orientations of the object.

2. The three-dimensional model processing method according to claim 1, characterized in that: The method further comprises: Use the object name to name the corresponding target sub-model to obtain the target naming sub-model; All target named sub-models are classified and saved to obtain a target named sub-model set of the target three-dimensional model.

3. The three-dimensional model processing method according to claim 1, characterized in that: The large hard decoration models excluded from the sub-model set include: Defining the size standard of the large hard decoration model; In the sub-model set, a sub-model that is larger than or equal to the size standard is identified as the large hard decoration model; Eliminate the large hard decoration model from the sub-model set.

4. The three-dimensional model processing method according to claim 3, characterized in that: After obtaining the remaining sub-model set and before taking multi-directional photos of each target sub-model in the remaining sub-model set, the method further includes: Calculating the number of planes of each sub-model in the remaining sub-model set to obtain the number of planes of each sub-model in the remaining sub-model set; Delete the sub-models whose number of planes is less than the preset number to obtain a new set of remaining sub-models; The new remaining sub-model is regarded as the remaining sub-model.

5. The three-dimensional model processing method according to claim 4, characterized in that: After obtaining the remaining sub-model set, and before taking multi-directional photos of each target sub-model in the remaining sub-model set, the method further includes: generating an enclosing polyhedron for each target sub-model in the set of remaining sub-models, wherein the enclosing polyhedron is a minimum polyhedron enclosing the target sub-model; Splicing two or more target sub-models having an intersection of the bounding polyhedrons in the remaining sub-model set respectively to obtain a new target sub-model; The new target sub-model is regarded as the target sub-model in the set of remaining sub-models.

6. The three-dimensional model processing method according to claim 5, characterized in that: The enclosing polyhedron is specifically a minimum cube enclosing the target sub-model.

7. The three-dimensional model processing method according to claim 1, characterized in that: Taking multi-directional photos of each target sub-model in the remaining sub-model set includes: Each target sub-model in the remaining sub-model set is photographed at least X times in a circle, where X is an integer greater than or equal to 9.

8. A three-dimensional model processing device based on artificial intelligence, characterized in that: include: An acquisition unit, configured to acquire a target three-dimensional model, wherein the target three-dimensional model is a three-dimensional model of a scene in the field of building decoration; a splitting unit, configured to split each sub-model in the target three-dimensional model to obtain a sub-model set comprising a plurality of sub-models; a culling unit, configured to cull large hard decoration models from the sub-model set to obtain a remaining sub-model set; a photographing unit, configured to take multi-directional photographs of each target sub-model in the remaining sub-model set to obtain a set of photographs of each target sub-model; The recognition unit is used to input the photo set of each target sub-model into a preset artificial intelligence model for object recognition to obtain the target object name of each target sub-model. The preset artificial intelligence model is a model that has been trained to recognize the name of the object based on multiple orientation photos of the object.

9. A computer device, characterized in that: include: Processor, memory, bus, input and output interface, network interface; The processor is connected to the memory, the input / output interface, and the network interface via a bus; The memory stores a program; When the processor executes the program stored in the memory, it implements the artificial intelligence-based three-dimensional model processing method described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that The computer storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the artificial intelligence-based three-dimensional model processing method described in any one of claims 1 to 7.

Images