Uv cutting method and device of three-dimensional model, storage medium and electronic equipment

Abstract

The application provides a UV cutting method and device of a three-dimensional model, a storage medium and an electronic device, and the method comprises the following steps: determining a silhouette image set and a light shadow image set corresponding to a target three-dimensional model; inputting the silhouette image set into a first region identification model to obtain a first main body, a first accessory set and a first region set; inputting the light shadow image set into a second region identification model to obtain a second main body, a second accessory set and a second region set; determining a plurality of to-be-detected joint regions according to the first region set and the second region set; inputting each to-be-detected joint region into a joint line identification model respectively to obtain a first joint line corresponding to each to-be-detected joint region; and determining a plurality of UV cutting lines based on the first joint lines to perform UV cutting on the target three-dimensional model. By applying the method, the UV cutting lines of the three-dimensional model can be automatically arranged and optimized based on artificial intelligence technology, time and labor cost can be saved, and efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, and in particular to a UV cutting method and apparatus for 3D models, a storage medium, and an electronic device. Background Technology

[0002] In the process of 3D modeling, it is necessary to assign texture patterns to the 3D model. When determining the texture distribution on the 3D model, it is usually necessary to perform UV cutting on the 3D model to determine the location of the texture distribution on the 3D model. That is, the surface of the 3D model is split according to the UV cutting lines and mapped to a two-dimensional plane based on the UV coordinate system.

[0003] In existing 3D modeling processes, the generated 3D models come with a default UV cutting line layout. These UV cutting lines are generally generated according to uniform rules, which typically cannot accurately cut the 3D model, resulting in poor cutting effects. Current UV cutting methods usually require manual adjustment of the default UV cutting line layout of the 3D model before UV cutting is performed based on the manually edited layout.

[0004] Based on the existing UV cutting method, the arrangement of UV cutting lines on the 3D model needs to be re-edited by the staff during the UV cutting process, which consumes a lot of time and manpower, resulting in low efficiency of 3D modeling. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a UV cutting method for three-dimensional models to solve the problem that existing UV cutting methods require manual adjustment of the UV line arrangement of the three-dimensional model, which consumes a lot of time and manpower and is inefficient.

[0006] This invention also provides a UV cutting device for three-dimensional models to ensure the practical implementation and application of the above method.

[0007] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0008] A UV cutting method for a 3D model, comprising:

[0009] When UV cutting is required on a target 3D model, determine the silhouette image set and the light and shadow image set corresponding to the target 3D model;

[0010] The silhouette image set is input into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding 3D model, and the sample labels are the subject, appendages, and junction regions of the 3D model;

[0011] The light and shadow image set is input into the second region recognition model to obtain a second subject, a second accessory set, and a second region set; the second accessory set includes at least one accessory, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, accessory, and junction region of the three-dimensional model;

[0012] Based on the first set of regions and the second set of regions, multiple intersection regions to be detected are determined;

[0013] Each of the detected junction regions is input into the junction line recognition model to obtain the first junction line corresponding to each detected junction region; the junction line recognition model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to that junction region;

[0014] Based on each of the first intersection lines, multiple UV cutting lines are determined, and the target 3D model is UV-cut based on each of the UV cutting lines.

[0015] Optionally, the annotation process for the silhouette samples in the above method includes:

[0016] Based on the silhouette image set of the first 3D model, shape recognition is performed on the first 3D model to obtain multiple first shapes corresponding to the first 3D model; the first 3D model is the 3D model corresponding to the silhouette sample.

[0017] Determine the connection relationships between each of the first shapes, and based on the connection relationships between each of the first shapes, determine the intersection areas between each of the first shapes;

[0018] Determine the silhouette proportion corresponding to each of the first shapes;

[0019] Based on the silhouette proportions of each of the first shapes, a first main shape is determined among the first shapes, and each of the first shapes other than the first main shape is designated as a first subordinate shape.

[0020] The intersection areas between each of the first shape bodies are labeled as the intersection areas of the first three-dimensional model, the first main shape body is labeled as the main body of the first three-dimensional model, and each of the first subordinate shape bodies is labeled as the subordinate body of the first three-dimensional model.

[0021] Optionally, the annotation process for the light and shadow samples in the above method includes:

[0022] Based on the light and shadow image set of the second 3D model, the shadow regions of the second 3D model are determined; the second 3D model is the 3D model corresponding to the light and shadow sample.

[0023] Based on the light and shadow image set of each shadow region and the second three-dimensional model, shape recognition is performed on the second three-dimensional model to obtain multiple second shapes corresponding to the second three-dimensional model;

[0024] Determine the connection relationships between each of the second shapes, and based on the connection relationships between each of the second shapes, determine the intersection areas between each of the second shapes;

[0025] Determine the light and shadow ratio corresponding to each of the second shapes;

[0026] Based on the light and shadow ratio corresponding to each of the second shapes, a second main shape is determined in each of the second shapes, and each of the second shapes other than the second main shape is regarded as a second subordinate shape.

[0027] The intersection areas between the various second shape bodies are labeled as the intersection areas of the second three-dimensional model, the second main shape body is labeled as the main body of the second three-dimensional model, and each second subordinate shape body is labeled as a subordinate body of the second three-dimensional model.

[0028] Optionally, the labeling process for the samples in the handover area, as described above, includes:

[0029] Multiple candidate intersection lines and multiple contour lines are determined corresponding to the target intersection region; the target intersection region is the intersection region corresponding to the sample of the intersection region; each of the candidate intersection lines intersects with each of the contour lines respectively;

[0030] For each candidate intersection line, determine each target position corresponding to the candidate intersection line, and each target position corresponds one-to-one with each contour line. The target position corresponding to each contour line is the intersection point of the contour line and the candidate intersection line.

[0031] For each candidate intersection line, the priority of the candidate intersection line is determined based on the contour shape of the corresponding contour line at each target position.

[0032] Select the candidate intersection line with the highest priority from all the candidate intersection lines, and mark the candidate intersection line as the intersection line corresponding to the target intersection area.

[0033] Optionally, in the above method, determining the priority of the candidate intersection line based on the contour shape of the corresponding contour line at each target position includes:

[0034] For each target position corresponding to the candidate intersection line, determine the contour angle of the contour line corresponding to the target position at the target position, determine whether the contour angle is greater than a preset angle threshold, and determine the priority corresponding to the target position based on the determination result;

[0035] The priority of the candidate handover line is determined based on the priority of each target location corresponding to the candidate handover line.

[0036] Optionally, in the above method, determining multiple UV cutting lines based on each of the first intersection lines includes:

[0037] Based on the first subject and the second subject, the target subject corresponding to the target 3D model is determined;

[0038] Based on the first set of appendages and the second set of appendages, determine each target appendage corresponding to the target 3D model;

[0039] Determine the inner orientation and intersection point position of each target appendage;

[0040] In each of the first handover lines, the first handover line corresponding to each of the target appendages is determined;

[0041] For each target appendage, the first intersection line is adjusted according to the inner orientation and the intersection proximity point of the target appendage to obtain the second intersection line corresponding to the first intersection line.

[0042] Each of the second intersection lines is taken as the UV cutting line.

[0043] Optionally, in the above method, determining multiple UV cutting lines based on each of the first intersection lines includes:

[0044] Each of the first intersection lines is taken as the UV cutting line.

[0045] A UV cutting device for a three-dimensional model, comprising:

[0046] The first determining unit is used to determine the silhouette image set and the light and shadow image set corresponding to the target three-dimensional model when UV cutting is required;

[0047] The first recognition unit is used to input the silhouette image set into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model;

[0048] The second recognition unit is used to input the light and shadow image set into the second region recognition model to obtain a second subject, a second set of appendages, and a second region set; the second set of appendages includes at least one appendage, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model;

[0049] The second determining unit is used to determine multiple intersection regions to be detected based on the first region set and the second region set;

[0050] The third identification unit is used to input each of the detected junction regions into the junction line identification model to obtain the first junction line corresponding to each of the detected junction regions; the junction line identification model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to the junction region;

[0051] The third determining unit is used to determine multiple UV cutting lines based on each of the first intersection lines, and to perform UV cutting on the target three-dimensional model based on each of the UV cutting lines.

[0052] A storage medium comprising stored instructions, wherein, when the instructions are executed, the device in which the storage medium resides executes a UV cutting method for a three-dimensional model as described above.

[0053] An electronic device includes a memory and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors using the UV cutting method for a three-dimensional model as described above.

[0054] Based on the above embodiments of the present invention, a UV cutting method for a three-dimensional model is provided. When UV cutting of a target three-dimensional model is required, a silhouette image set and a light and shadow image set corresponding to the target three-dimensional model are determined; the silhouette image set is input into a first region recognition model to obtain a first subject, a first accessory set, and a first region set; the first accessory set includes at least one accessory, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding three-dimensional model, and the sample labels are the subject, accessory, and junction region of the three-dimensional model; the light and shadow image set is input into a second region recognition model to obtain a second subject, a second accessory set, and a second region set; the second accessory set includes at least one accessory, and the first... The two region sets include at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding 3D model, and the sample labels are the main body, appendages, and junction regions of the 3D model; multiple junction regions to be detected are determined according to the first region set and the second region set; each junction region to be detected is input into the junction line recognition model to obtain the first junction line corresponding to each junction region to be detected; the junction line recognition model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to the junction region; multiple UV cutting lines are determined based on each first junction line, and UV cutting is performed on the target 3D model based on each UV cutting line. By applying the method provided in this embodiment of the invention, different region recognition models can be used to identify the main body, appendages, and junction regions of a 3D model, the junction regions of the 3D model can be determined according to the model's recognition results, the junction lines can be identified in the junction regions of the 3D model using the junction line recognition model, and the arrangement of UV cutting lines can be determined based on each junction line identified by the model for UV cutting. During the UV cutting process, artificial intelligence technology is used to optimize the arrangement of UV cutting lines on the 3D model. This can automatically generate a UV cutting line arrangement that matches the actual shape of the 3D model without the need for manual adjustment of the UV cutting lines. This can save a lot of time and manpower costs and improve the efficiency of 3D modeling. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0056] Figure 1 A flowchart of a UV cutting method for a three-dimensional model provided in an embodiment of the present invention;

[0057] Figure 2 An example image of a silhouette sample provided in an embodiment of the present invention;

[0058] Figure 3 An example diagram of a light and shadow sample provided in an embodiment of the present invention;

[0059] Figure 4 An example diagram of a junction area sample provided in an embodiment of the present invention;

[0060] Figure 5 Another example diagram of a junction area sample provided in an embodiment of the present invention;

[0061] Figure 6 An example diagram showing the interior and exterior orientation of a three-dimensional model provided in an embodiment of the present invention;

[0062] Figure 7 An example diagram of the near and far points of a three-dimensional model provided in an embodiment of the present invention;

[0063] Figure 8 An example diagram illustrating the UV cutting process of a three-dimensional model provided in an embodiment of the present invention;

[0064] Figure 9 A schematic diagram of the structure of a three-dimensional model UV cutting device provided in an embodiment of the present invention;

[0065] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] This invention provides a UV cutting method for a 3D model. The method can be applied to a 3D model texture processing platform, and its execution entity can be the platform's server. The method flowchart is shown below. Figure 1 As shown, it includes:

[0069] S101: When it is necessary to perform UV cutting on the target 3D model, determine the silhouette image set and the light and shadow image set corresponding to the target 3D model;

[0070] In the method provided by this invention, when a user needs to perform UV cutting on a 3D model, the 3D model to be UV-cut can be imported into the platform, and a UV cutting instruction can be triggered through the platform front-end. When the server receives the instruction to perform UV cutting on the specified 3D model, it can use that 3D model as the target 3D model. The server can generate a silhouette image set and a lighting image set of the target 3D model through a predetermined model display method. Specifically, the target 3D model can be projected from different angles to obtain silhouette images of the target 3D model from multiple angles, and these silhouette images from multiple angles can be combined into a corresponding silhouette image set. The target 3D model can be illuminated from different angles with different intensities to obtain lighting images of the target 3D model from different angles under illumination, and these lighting images of the target 3D model from different angles can be combined into a corresponding lighting image set.

[0071] S102: Input the silhouette image set into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model;

[0072] In the method provided by this invention, a region recognition model can be pre-constructed based on artificial intelligence technology. This model can be specifically obtained through training pre-labeled silhouette samples and machine learning algorithms; this model is the first region recognition model. The sample features of the silhouette samples are the silhouette image set of their corresponding 3D models, and the labels for the silhouette samples are the main body, appendages, and junction areas of the 3D model. Junction areas refer to the areas in the model where the main body and appendages, or appendages themselves, intersect. It can be understood that the trained first region recognition model can identify the main body, appendages, and junction areas of the corresponding 3D model based on the input silhouette image set.

[0073] In the method provided by this invention, when the server performs UV cutting on the target 3D model, it can load the silhouette image set corresponding to the target 3D model into the input layer of the first region recognition model. After processing by the first region recognition model, the recognition result output by the first region recognition model can be obtained, that is, the main body, appendages, and junction regions of the target 3D model recognized by the first region recognition model. The main body output by the first region recognition model is taken as the first main body, the various appendages output by it are combined into a first appendage set, and the various junction regions output by it are combined into a first region set.

[0074] It should be noted that in specific applications, there is only one main body of the 3D model, but the number of appendages and intersecting areas of the 3D model is determined by the actual model structure. There may be only one appendage or multiple intersecting areas, which does not affect the functionality of the method provided in the embodiments of the present invention.

[0075] S103: Input the light and shadow image set into the second region recognition model to obtain a second subject, a second accessory set, and a second region set; the second accessory set includes at least one accessory, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, accessory, and junction region of the three-dimensional model;

[0076] In the method provided by this invention, a region recognition model can be pre-constructed based on artificial intelligence technology. This model can be specifically obtained through training pre-labeled light and shadow samples and machine learning algorithms; this model is the second region recognition model. The sample features of the light and shadow samples are the light and shadow image sets of their corresponding 3D models, and the labels for the light and shadow samples are the main body, appendages, and junction areas of the 3D model. It can be understood that the trained second region recognition model can identify the main body, appendages, and junction areas of the corresponding 3D model based on the input light and shadow image set.

[0077] In the method provided by this invention, when the server performs UV cutting on the target 3D model, it can load the light and shadow image set corresponding to the target 3D model into the input layer of the second region recognition model. After processing by the second region recognition model, the recognition result output by the second region recognition model can be obtained, namely, the main body, appendages, and junction regions of the target 3D model identified by it. The main body output by the second region recognition model is taken as the second main body, the various appendages output by it are combined into a second appendage set, and the various junction regions output by it are combined into a second region set.

[0078] S104: Based on the first region set and the second region set, determine multiple intersection regions to be detected;

[0079] In the method provided by this invention, the server uses two region recognition models to identify the main body, appendages, and junction regions of the target 3D model. Combining the recognition results of the two region recognition models, the junction regions of the target 3D model are determined, and these finally determined junction regions are used as junction regions to be detected. Specifically, the junction regions in the first and second region sets can be integrated, overlapping junction regions can be merged, and the integrated junction regions are used as junction regions to be detected.

[0080] S105: Input each of the detected junction regions into the junction line recognition model to obtain the first junction line corresponding to each of the detected junction regions; the junction line recognition model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to the junction region;

[0081] In the method provided by this invention, a boundary line recognition model is pre-constructed based on artificial intelligence technology. This model is trained using multiple pre-labeled boundary region samples and machine learning algorithms. The sample features of a boundary region sample are a boundary region of its corresponding 3D model, and the label for this boundary region sample is the boundary line corresponding to that boundary region. In this invention, the boundary region contains multiple contour lines of the 3D model, where a certain contour line is the boundary line at the model's boundary position. The trained boundary line recognition model can identify each contour line in the input boundary region and identify the contour line most likely to be the boundary line from among all the contour lines.

[0082] In the method provided by this invention, the server can load each intersection region to be detected into the input layer of the intersection line recognition model according to the input data format requirements of the intersection line recognition model. After processing the input intersection region to be detected, the intersection line recognition model can output the intersection line in the intersection region to be detected, and use the intersection line as the first intersection line corresponding to the intersection region to be detected. After recognizing each intersection region to be detected using the intersection line recognition model, the first intersection line corresponding to each intersection region to be detected can be obtained.

[0083] S106: Based on each of the first intersection lines, determine a plurality of UV cutting lines, and perform UV cutting on the target three-dimensional model based on each of the UV cutting lines.

[0084] In the method provided by this invention, the server can determine the UV cutting line arrangement of the target 3D model based on each first intersection line, that is, determine the position of each UV cutting line of the target 3D model, and perform UV cutting on the target 3D model based on the position of each UV cutting line. Specifically, the UV cutting lines can be arranged directly according to the arrangement of each first intersection line. Alternatively, an optimization strategy can be preset, and the position of each first intersection line can be optimized and adjusted according to the optimization strategy based on the arrangement of each first intersection line, and the UV cutting lines can be arranged accordingly according to the adjusted arrangement of each first intersection line.

[0085] Based on the method provided in this embodiment of the invention, when UV cutting is required on a target 3D model, a silhouette image set and a light and shadow image set corresponding to the target 3D model are determined; the silhouette image set is input into a first region recognition model to obtain a first main body, a first set of appendages, and a first region set; the light and shadow image set is input into a second region recognition model to obtain a second main body, a second set of appendages, and a second region set; multiple intersection regions to be detected are determined based on the first region set and the second region set; each intersection region to be detected is input into an intersection line recognition model to obtain a first intersection line corresponding to each intersection region to be detected; multiple UV cutting lines are determined based on each first intersection line, and UV cutting is performed on the target 3D model based on each UV cutting line. By applying the method provided in this embodiment of the invention, different region recognition models can be used to identify the main body, appendages, and intersection regions of a 3D model. The intersection regions of the 3D model are determined based on the model's recognition results. An intersection line recognition model is used to identify intersection lines in the intersection regions of the 3D model. Based on the intersection lines obtained from the model recognition, the arrangement of UV cutting lines is determined for UV cutting. During the UV cutting process, artificial intelligence technology is used to optimize the arrangement of UV cutting lines on the 3D model. This can automatically generate a UV cutting line arrangement that matches the actual shape of the 3D model without the need for manual adjustment of the UV cutting lines. This can save a lot of time and manpower costs and improve the efficiency of 3D modeling.

[0086] exist Figure 1 Based on the method shown, the annotation process of the silhouette sample mentioned in step S102 of the method provided in this embodiment of the invention includes:

[0087] Based on the silhouette image set of the first 3D model, shape recognition is performed on the first 3D model to obtain multiple first shapes corresponding to the first 3D model; the first 3D model is the 3D model corresponding to the silhouette sample.

[0088] In the method provided by this invention, a set of silhouette images of existing 3D models can be collected from materials accumulated in the production environment or publicly available materials. This set consists of silhouette images of the 3D model from different angles, and the entire set of silhouette images of a 3D model is used as a silhouette sample. When annotating a silhouette sample, image recognition can be performed based on the silhouette image set of its corresponding 3D model (i.e., the first 3D model). The silhouettes of various shapes constituting the silhouette are identified from each silhouette image. Based on the silhouettes of these shapes, the various shapes that make up the 3D model are identified, and the identified shapes are the corresponding first shapes of the first 3D model. For example... Figure 2The silhouette sample shown illustrates a set of silhouette images of a 3D model in the shape of a teapot. Through silhouette image recognition, silhouettes of shapes such as A-teapot handle, F-teapot body, D-teapot spout, and E-teapot lid can be identified. Based on the silhouettes of the shapes identified in each silhouette image, the 3D model of the teapot can be divided into the teapot body, teapot handle, teapot spout, and teapot lid.

[0089] Determine the connection relationships between each of the first shapes, and based on the connection relationships between each of the first shapes, determine the intersection areas between each of the first shapes;

[0090] In the method provided by this invention, the connection relationships between the first shapes can be determined based on their positions in the first three-dimensional model; that is, for each shape, other shapes connected to it are identified. Based on these connection relationships, the area where two connected shapes meet is designated as the intersection area between the two shapes. This allows the intersection areas between the first shapes to be obtained. For example... Figure 2 The silhouette sample shown can be identified as follows: in its corresponding 3D model, the shape of the teapot handle is connected to the shape of the teapot body. Based on the interconnected teapot handle and teapot body, the area where the connection point is located can be identified, such as... Figure 2 The regions indicated by j1 and j2 shown in the diagram.

[0091] Determine the silhouette proportion corresponding to each of the first shapes;

[0092] In the method provided by this embodiment of the invention, the total silhouette area of ​​the first three-dimensional model and the silhouette area corresponding to each first shape can be calculated based on the silhouette image set of the first three-dimensional model. For each first shape of the first three-dimensional model, the silhouette area corresponding to the shape is divided by the total silhouette area of ​​the three-dimensional model, and the quotient is taken as the silhouette proportion corresponding to the first shape, thereby obtaining the silhouette proportion corresponding to each first shape.

[0093] Based on the silhouette proportions of each of the first shapes, a first main shape is determined among the first shapes, and each of the first shapes other than the first main shape is designated as a first subordinate shape.

[0094] In the method provided by this invention, subject recognition rules can be pre-defined. According to these rules, the main shape (i.e., the model's main body) and subordinate shape (i.e., the model's appendages) of the first three-dimensional model are identified among the various first shapes. In this invention, the main shape and subordinate shape are determined based on the silhouette proportions of each first shape. Specifically, a model library can be pre-built, in which various main models, typically of the main body shape, can be deployed. For each first shape, the silhouette proportion corresponding to that shape can be compared with a preset silhouette threshold. If the silhouette proportion corresponding to that shape is greater than the silhouette threshold, then that shape is considered a candidate shape. Each candidate shape is matched with each main model in the preset model library, and the candidate shape that matches the main model in the preset model library is designated as the main shape. Each first shape other than the main shape is designated as a subordinate shape.

[0095] The intersection areas between each of the first shape bodies are labeled as the intersection areas of the first three-dimensional model, the first main shape body is labeled as the main body of the first three-dimensional model, and each of the first subordinate shape bodies is labeled as the subordinate body of the first three-dimensional model.

[0096] In the method provided by the embodiments of the present invention, for a silhouette sample, the intersection area between each first shape is marked as the intersection area of ​​the three-dimensional model corresponding to the silhouette sample, the main shape in each first shape is marked as the main body of the three-dimensional model, and the subordinate shape in each first shape is marked as the subordinate body of the three-dimensional model, thereby obtaining the sample label of the silhouette sample.

[0097] Based on the method provided in this embodiment of the invention, the silhouette recognition algorithm can identify the main body, appendages and junction areas of a 3D model based on the silhouette image set of the silhouette sample, thereby annotating the silhouette sample. This can realize automated annotation of the silhouette sample, reduce the workload of manual annotation, and improve the efficiency of sample annotation.

[0098] exist Figure 1 Based on the method shown, the annotation process of the light and shadow samples mentioned in step S103 of the method provided in this embodiment of the invention includes:

[0099] Based on the light and shadow image set of the second 3D model, the shadow regions of the second 3D model are determined; the second 3D model is the 3D model corresponding to the light and shadow sample.

[0100] The method provided in this invention can acquire a set of lighting and shadow images of an existing 3D model, using one such set as a lighting and shadow sample. When labeling the lighting and shadow sample, the various shadow regions of the corresponding 3D model (i.e., the second 3D model) are identified based on its lighting and shadow image. Specifically, the pixel values ​​of each pixel in the lighting and shadow image are obtained and compared with a preset black and white threshold to label the shadow regions in the lighting and shadow image. By combining the shadow regions in each lighting and shadow image, the corresponding shadow regions on the 3D model are labeled. For example... Figure 3 The silhouette sample shown illustrates a set of light and shadow images of a 3D model in the shape of a teapot. By comparing the light and shadow images in the set of light and shadow images, shadow areas such as B1, B2, and B3 can be identified.

[0101] Based on the light and shadow image set of each shadow region and the second three-dimensional model, shape recognition is performed on the second three-dimensional model to obtain multiple second shapes corresponding to the second three-dimensional model;

[0102] In the method provided by the embodiments of the present invention, image recognition is performed on the light and shadow image set of the light and shadow sample in combination with each shadow region of the second three-dimensional model, and the shape images constituting the model are identified from each light and shadow image. Based on the shape images identified from each light and shadow image, the shape objects constituting the three-dimensional model (i.e., the second shape objects) are identified.

[0103] Determine the connection relationships between each of the second shapes, and based on the connection relationships between each of the second shapes, determine the intersection areas between each of the second shapes;

[0104] In the method provided by the embodiments of the present invention, the connection relationship between each second shape body can be determined according to the position of each second shape body in the second three-dimensional model, and the area where the connection position of two connected shapes bodies is located is taken as the intersection area between the corresponding two shapes bodies.

[0105] Determine the light and shadow ratio corresponding to each of the second shapes;

[0106] In the method provided by this embodiment of the invention, the total area of ​​light and shadow of the three-dimensional model and the area of ​​light and shadow corresponding to each second shape can be calculated based on the light and shadow image set of the second three-dimensional model. For each second shape of the three-dimensional model, the area of ​​light and shadow corresponding to the shape is divided by the total area of ​​light and shadow of the three-dimensional model, and the quotient is taken as the proportion of light and shadow corresponding to the shape, thereby obtaining the proportion of light and shadow corresponding to each second shape.

[0107] Based on the light and shadow ratio corresponding to each of the second shapes, a second main shape is determined in each of the second shapes, and each of the second shapes other than the second main shape is regarded as a second subordinate shape.

[0108] In the method provided by the embodiments of the present invention, a main shape and an auxiliary shape are determined in each second shape according to the light and shadow ratio corresponding to each second shape. Specifically, for each second shape, the light and shadow ratio corresponding to the shape can be compared with a preset image area threshold. If the light and shadow ratio corresponding to the shape is greater than the image area threshold, the shape is regarded as a candidate shape. Each candidate shape is matched with each main model in a preset model library. The candidate shape that matches the main model in the preset model library is regarded as the main shape, and each second shape other than the main shape is regarded as an auxiliary shape.

[0109] The intersection areas between the various second shape bodies are labeled as the intersection areas of the second three-dimensional model, the second main shape body is labeled as the main body of the second three-dimensional model, and each second subordinate shape body is labeled as a subordinate body of the second three-dimensional model.

[0110] In the method provided by the embodiments of the present invention, for a light and shadow sample, the intersection area between each second shape is marked as the intersection area of ​​the three-dimensional model corresponding to the light and shadow sample, the main shape in each second shape is marked as the main body of the three-dimensional model, and the subordinate shape in each second shape is marked as the subordinate body of the three-dimensional model, thereby obtaining the sample label of the light and shadow sample.

[0111] Based on the method provided in this invention, the shadow area of ​​a 3D model can be calibrated using a black-and-white threshold comparison algorithm based on a set of light and shadow images of light and shadow samples. On the basis of the light and shadow image set, the main body, appendages and junction areas of the 3D model can be identified by combining the shadow area of ​​the 3D model. This allows for the annotation of light and shadow samples, enabling automated annotation of light and shadow samples, reducing the workload of manual annotation, and improving the efficiency of sample annotation.

[0112] exist Figure 1 Based on the method shown, the annotation process of the junction area samples mentioned in step S105 of the method provided in this embodiment of the invention includes:

[0113] Multiple candidate intersection lines and multiple contour lines are determined corresponding to the target intersection region; the target intersection region is the intersection region corresponding to the sample of the intersection region; each of the candidate intersection lines intersects with each of the contour lines respectively;

[0114] In the method provided by this invention, the model shapes of the intersection regions can be collected from existing 3D model materials, and each collected model shape of the intersection region is taken as an intersection region sample. When annotating the intersection region sample, the intersection region corresponding to the sample can be taken as the target intersection region. Contour line recognition is performed on the target intersection region to obtain the various model contour lines on the target intersection region. These contour lines are typically presented in a line mesh style, meaning each contour line consists of lines in two directions. Model contour lines whose line direction is parallel to the connection point of the shape are taken as candidate intersection lines, and model contour lines in the other direction are taken as individual contour lines. This yields the candidate intersection lines and individual contour lines in the target intersection region. It can be understood that each candidate intersection line intersects with each individual contour line. For example… Figure 4 and Figure 5 The sample of the junction area shown, wherein, Figure 4 Two junction area samples are shown; the left grid represents one junction area sample, and the right grid represents another junction area sample. Figure 4 In the two intersection region samples shown, the intersection region is formed by the intersection of two left and right shapes. Each intersection region contains the contour lines on the axes represented by xa, xb, xc, xd, xe and xf, and the candidate intersection lines on the axes represented by 1, 2, 3, 4, 5 and 6. Figure 5 The sample of the intersection region shown includes the contour lines on the axes represented by ya, yb, yc, yd, ye and yf, and the candidate intersection lines on the axes represented by 1, 2, 3, 4, 5 and 6.

[0115] For each candidate intersection line, determine each target position corresponding to the candidate intersection line, and each target position corresponds one-to-one with each contour line. The target position corresponding to each contour line is the intersection point of the contour line and the candidate intersection line.

[0116] In the method provided by the embodiments of the present invention, for each candidate intersection line, the intersection point of the candidate intersection line and each contour line is taken as a target position of the candidate intersection line.

[0117] For each candidate intersection line, the priority of the candidate intersection line is determined based on the contour shape of the corresponding contour line at each target position.

[0118] In the method provided by this embodiment of the invention, for each candidate intersection line, the probability that the position is a shape intersection position is determined based on the contour shape of the contour line intersecting with the candidate intersection line at each target position, thereby determining the priority corresponding to the candidate intersection line. The higher the probability that the position of the candidate intersection line is a shape intersection position, the higher its priority level.

[0119] Select the candidate intersection line with the highest priority from all the candidate intersection lines, and mark the candidate intersection line as the intersection line corresponding to the target intersection area.

[0120] In the method provided by the embodiments of the present invention, the candidate intersection line with the highest priority is selected and marked as the intersection line of the target intersection area according to the priority of each candidate intersection line, thereby obtaining the sample label of the current intersection area sample.

[0121] Based on the method provided in this embodiment of the invention, the junction line of the junction area can be identified by the model outline lines in the junction area, thereby annotating the junction area sample. This can realize the automated annotation of the junction area sample, reduce the workload of manual annotation, and improve the efficiency of sample annotation.

[0122] Based on the method provided in the above embodiments, the method provided in this embodiment of the invention, wherein the process of determining the priority of the candidate intersection line based on the contour shape of the corresponding contour line at each target position corresponding to the candidate intersection line includes:

[0123] For each target position corresponding to the candidate intersection line, determine the contour angle of the contour line corresponding to the target position at the target position, determine whether the contour angle is greater than a preset angle threshold, and determine the priority corresponding to the target position based on the determination result;

[0124] In the method provided by this invention, at each target position of the current candidate intersection line, the contour angle of the contour line intersecting the candidate intersection line at that target position is determined, and the contour angle is compared with a preset angle threshold. The priority corresponding to the target position is determined based on the comparison result. It can be understood that if the direction of a line in a model contour line is bent, it indicates that the bend may be the intersection of different shapes. Correspondingly, if the contour angle of the contour line is less than the preset angle threshold, it indicates that the contour line has a significant bend at the target position, and the priority corresponding to that target position can be marked as high priority. If the contour angle of the contour line is greater than the preset angle threshold, it indicates that the direction of the contour line at the target position has no significant bend, and the priority corresponding to that target position can be marked as low priority. The contour angle of the contour line refers to the angle formed by the contour line at the corresponding target position, for example... Figure 4 The sample of the intersection area shown on the left side of the image, taking the candidate intersection line pointed to by number 3 as an example, the intersection point of the candidate intersection line and the contour line on the xe axis is the target position on the candidate intersection line corresponding to the contour line. On the contour line, with the target position as the dividing point, the angle formed by the left and right line segments at the dividing point is regarded as the contour angle of the contour line at the target position.

[0125] The priority of the candidate handover line is determined based on the priority of each target location corresponding to the candidate handover line.

[0126] In the method provided by this invention, the priority of the candidate handover line can be evaluated by combining the priorities of each target position corresponding to the candidate handover line. The specific evaluation method can be set according to actual needs. For example, the more high-priority targets there are, the higher the priority level of the candidate handover line.

[0127] Based on the method provided in the embodiments of the present invention, the possibility that the location of the candidate intersection line is at the intersection of shapes can be identified based on the contour angle of the contour line. The process of contour angle identification and calculation is relatively convenient and fast, which helps to improve efficiency.

[0128] exist Figure 1 Based on the method shown, the method provided in this embodiment of the invention, the process of determining multiple UV cutting lines based on each of the first intersection lines mentioned in step S106, includes:

[0129] Based on the first subject and the second subject, the target subject corresponding to the target 3D model is determined;

[0130] In the method provided by this invention, the server can determine the main body in the target 3D model by combining the recognition results of the first region recognition model and the second region recognition model according to a pre-set subject selection strategy, and then use that main body as the target subject. Specifically, if the first subject and the second subject represent the same shape, then that shape can be directly used as the main body of the target 3D model. If the first subject and the second subject represent different shapes, then one of the first and second subjects is selected as the target subject according to the subject selection strategy. The subject selection strategy can be set according to actual needs; for example, it can be set to prioritize the subject identified by a certain region recognition model.

[0131] Based on the first set of appendages and the second set of appendages, determine each target appendage corresponding to the target 3D model;

[0132] In the method provided by this invention, the server can determine each appendage in the target 3D model by combining a first appendage set and a second appendage set according to a pre-set appendage filtering strategy, and regard these appendages as target appendages. For example, based on the target subject as the subject, if each appendage in the first appendage set is the same as each appendage in the second appendage set, then this appendage can be regarded as a target appendage. If the appendages in the two appendage sets are different, then each target appendage of the target 3D model can be filtered out according to the appendage filtering strategy. The appendage filtering strategy can be set according to actual needs. For example, it can be set to prioritize the appendages identified by the model in a certain area, or the two appendage sets can be integrated, and the appendages obtained after integration can be regarded as target appendages, etc.

[0133] Determine the inner orientation and intersection point position of each target appendage;

[0134] In the method provided by this embodiment of the invention, the server can mark the inner orientation and intersection proximity point of each target appendage on the target 3D model according to a preset object orientation setting method. Specifically, for each target appendage, with the main body of the target 3D model as the center, the inward orientation of the target appendage closest to the main body is marked as the inner orientation of the target appendage, and the position on the target appendage closest to the intersection area corresponding to the target appendage is marked as the intersection proximity point of the target appendage. For example... Figure 6 The 3D model shown shows the orientation indicated by p2 as the inner orientation of each appendage, as... Figure 7 The three-dimensional model shown is indicated by p5, which is the intersection point of the three-dimensional model.

[0135] In each of the first handover lines, the first handover line corresponding to each of the target appendages is determined;

[0136] In the method provided by the embodiments of the present invention, the server can determine the first intersection line corresponding to each target appendage in each first intersection line, that is, the intersection line corresponding to the intersection area corresponding to the target appendage.

[0137] For each target appendage, the first intersection line is adjusted according to the inner orientation and the intersection proximity point of the target appendage to obtain the second intersection line corresponding to the first intersection line.

[0138] In the method provided by this embodiment of the invention, the server can adjust the position of the first intersection line corresponding to each target appendage according to the inner orientation and the intersection proximity point position of the target appendage, so as to avoid the UV cutting line being in a conspicuous area. Specifically, the position of the first intersection line can be adjusted inward and closer to the intersection proximity point position according to a preset adjustment threshold, and the adjusted first intersection line is used as the corresponding second intersection line.

[0139] Each of the second intersection lines is taken as the UV cutting line.

[0140] In the method provided by the embodiments of the present invention, the server arranges the UV cutting lines of the target three-dimensional model according to the arrangement of each second intersection line, that is, the second intersection line is used as the UV cutting lines of the target three-dimensional model.

[0141] Based on the method provided in the embodiments of the present invention, the position of the intersection line can be further adjusted based on the inner orientation of the appendage, which helps to avoid the UV cutting line being in a conspicuous position of the model and improves the aesthetics of UV cutting.

[0142] exist Figure 1 Based on the method shown, the method provided in this embodiment of the invention, the process of determining multiple UV cutting lines based on each of the first intersection lines mentioned in step S106, includes:

[0143] Each of the first intersection lines is taken as the UV cutting line.

[0144] In the method provided by the embodiments of the present invention, the server arranges the UV cutting lines of the target three-dimensional model according to the arrangement of each first intersection line, that is, each first intersection line is used as each UV cutting line of the target three-dimensional model.

[0145] To better illustrate the method provided in the embodiments of the present invention, and in conjunction with practical application scenarios, the embodiments of the present invention provide another UV cutting method for three-dimensional models. The method provided in the embodiments of the present invention can be applied to a digital texture distribution system, which can be used to perform UV cutting on three-dimensional models, assign textures to the three-dimensional models, and achieve three-dimensional modeling. The digital texture distribution system in the embodiments of the present invention deploys three AI models, namely the first region recognition model, the second region recognition model, and the intersection line recognition model mentioned in the previous embodiments.

[0146] The following is a brief explanation of the construction of the three AI models.

[0147] First, a brief explanation of the construction of the two region recognition models used for region recognition in 3D models will be given.

[0148] The first region recognition model is obtained through model training based on silhouette samples. The annotation of silhouette samples can be achieved using a silhouette recognition algorithm. Specifically, the silhouette recognition algorithm can identify the main body, appendages, and junction areas of the corresponding 3D model based on a silhouette image set of silhouette samples. In the method provided by this embodiment, the main body is defined as a continuous shape without small protrusions, whose silhouette proportion exceeds a preset threshold percentage of the total silhouette area, and whose shape matches one of the shapes in a predetermined model library. All other shapes in the 3D model besides the main body are appendages. In particular, when a part of the initially identified main body has an outward, intersecting, or non-intersecting relationship with the main body, if the volume of that part exceeds a predetermined threshold, that part is considered an appendage. By recognizing each silhouette image in the silhouette image set, the main body, appendages, and junction areas of the 3D model of the silhouette sample can be comprehensively evaluated, thereby annotating the silhouette sample. For example... Figure 2 The silhouette samples shown can identify the main silhouettes such as F, P, Z, V, N, etc., the accessory silhouettes such as A, E, D, Q, G, T, S, G, T, W, U, J, I, L, M, N, O, etc., and the junction areas such as j1, j2, j3, ..., j16.

[0149] The second region recognition model is trained based on lighting and shadow samples. The annotation process for these samples can be implemented using a black-and-white threshold recognition algorithm. Specifically, this algorithm identifies various shadow regions of the 3D model based on a set of lighting and shadow images. Then, based on this image set and the shadow regions of the 3D model, it identifies the main body, appendages, and transitional areas of the 3D model. Recognition based on shadow regions is beneficial for identifying shapes and transitional areas that are difficult to discern from silhouettes. For example... Figure 3The light and shadow samples shown can identify shadow regions such as B1, B2, B3, ..., B22 in various light and shadow images by comparing the pixel values ​​of each pixel with a predetermined black and white threshold. Combining these shadow regions, the transition regions such as j1, j2, j3, ..., j15 can be identified.

[0150] The following is a brief explanation of the construction of the intersection line recognition model. The intersection line recognition model is trained based on intersection region samples. The annotation process of the intersection region samples can be achieved based on the angle recognition of the intersection line. Specifically, for example... Figure 4 As shown, the model outline mesh on the left is the line distribution presented by the expansion of an intersection region. This region includes model outline lines in the horizontal directions xa, xb, xc, xd, xe, and xf, and model outline lines in the vertical directions 1, 2, 3, 4, 5, and 6. Model outline lines in the horizontal directions are considered outline lines, and model outline lines in the vertical directions are considered candidate intersection lines. For each candidate intersection line, find its intersection points with each contour line. For example, for the candidate intersection line corresponding to the vertical axis 3, find the intersection points (xa,3), (xb,3), (xc,3), (xd,3), (xe,3), and (xf,3). Then, calculate the contour angle of the contour line intersecting the candidate intersection line at each intersection point. For example, at the intersection point (xa,3), calculate the angle formed by the contour line corresponding to the horizontal axis x3 at that position, which is the angle formed by the line segment from point (xa,4) to point (xa,3) and the line segment from point (xa,3) to point (xa,2) on that contour line. Compare this contour angle with a preset angle threshold. If the contour angle is greater than the angle threshold, mark the intersection point of the candidate intersection line as low priority; if the contour angle is less than the angle threshold, mark the intersection point of the candidate intersection line as high priority. Based on the priority levels of all intersection point markings on the candidate intersection line, determine the priority of the candidate intersection line. Similarly, the priority of all candidate intersection lines can be assessed, and then the optimal candidate intersection line can be selected as the intersection line of the boundary area according to the priority level, so as to label the sample of the boundary area.

[0151] The three AI models mentioned above can be further improved through continuous AI learning and data accumulation. Building upon these three AI models, as... Figure 8As shown, when a user needs to perform UV cutting on a 3D model using a digital texture distribution acquisition system, the 3D model can be imported into the system. On one hand, the user can use the system to manually determine the UV cutting line layout of the 3D model for UV cutting. On the other hand, the user can directly trigger an automated UV layout command, allowing the system to perform UV cutting based on three AI models. Specifically, in scenarios where three AI models are used for UV cutting, the UV cutting process of the 3D model provided in this embodiment mainly includes:

[0152] AI-based model and silhouette image recognition for 3D models: For 3D models that require UV cutting, the first region recognition model can be used to identify the main body, appendages and junction areas of the 3D model based on its silhouette image, and a set of recognition results for the main body, appendages and junction areas can be obtained.

[0153] AI-based model and light and shadow image recognition of 3D models: For 3D models that need to be UV-cut, a second region recognition model can be used to identify the main body, appendages and junction areas of the 3D model based on its light and shadow image, and obtain a set of recognition results for the main body, appendages and junction areas.

[0154] AI-based model for identifying intersection lines: Combining the identification results of the first region identification model and the second region identification model, the intersection regions of the 3D model are determined, and the intersection line identification model is used to identify the intersection lines of each intersection region, thus obtaining the intersection lines corresponding to each intersection region.

[0155] Determine the object's internal and external orientation, main body, and appendages: Combining the recognition results of the first and second region recognition models, determine the main body and appendages of the 3D model. Then, according to a predetermined orientation setting method, set the internal and external orientations of the 3D model. On one hand, it is necessary to identify the inner and outer orientations of the appendages on the 3D model, setting the main body as the center, and the appendages closer to the main body having their inward orientation defined as the inner side. For example... Figure 6 The orientation shown in p2 is set to the side furthest from the main body, with diffusion defined as the outer side, for example... Figure 6 The orientation is shown in p1. On the other hand, it is necessary to identify the proximity and distance points of the intersections of appendages on the 3D model, setting the proximity point as the closest point to the intersection area, for example... Figure 7 J30 is the junction area. Figure 7 The location shown on page 5 is the near point of intersection. The farthest point of intersection is defined as the location furthest from the intersection area. For example... Figure 7 The position shown on page 6.

[0156] UV cutting: The intersection lines of each intersection region identified by the intersection line recognition model can be directly used as UV cutting lines of the 3D model to achieve UV cutting. Alternatively, the UV arrangement can be further optimized based on the above intersection lines. Specifically, the position of the intersection lines can be adjusted according to a predetermined adjustment threshold, making them closer to the intersection point on the inner side of the appendage. UV cutting can then be performed based on the optimized intersection lines to avoid UV cutting in conspicuous areas.

[0157] The method provided in this invention utilizes artificial intelligence technology to automatically optimize the UV layout of a 3D model, providing preliminary work for the topology of the 3D model into a low-poly application-level model. This helps maintain the integrity and rationality of subsequent texture layout, reduces manual operations in the UV layout of the 3D model, and improves efficiency.

[0158] and Figure 1 Corresponding to the UV cutting method for a three-dimensional model shown, this embodiment of the invention also provides a UV cutting device for a three-dimensional model, used for UV cutting of a three-dimensional model. Figure 1 The specific implementation of the method shown is illustrated in the following diagram. Figure 9 As shown, it includes:

[0159] The first determining unit 201 is used to determine the silhouette image set and the light and shadow image set corresponding to the target three-dimensional model when UV cutting of the target three-dimensional model is required.

[0160] The first recognition unit 202 is used to input the silhouette image set into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model;

[0161] The second recognition unit 203 is used to input the light and shadow image set into the second region recognition model to obtain a second subject, a second set of appendages, and a second region set; the second set of appendages includes at least one appendage, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model;

[0162] The second determining unit 204 is used to determine multiple intersection regions to be detected based on the first region set and the second region set;

[0163] The third identification unit 205 is used to input each of the detected junction regions into the junction line identification model to obtain the first junction line corresponding to each of the detected junction regions; the junction line identification model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to the junction region;

[0164] The third determining unit 206 is used to determine multiple UV cutting lines based on each of the first intersection lines, and to perform UV cutting on the target three-dimensional model based on each of the UV cutting lines.

[0165] The apparatus provided in this invention can identify the main body, appendages, and junction areas of a 3D model using different region recognition models. Based on the model's recognition results, the junction areas of the 3D model are determined. Junction line recognition models are used to identify junction lines within these junction areas. Based on the junction lines identified by the model, the arrangement of UV cutting lines is determined for UV cutting. During UV cutting, artificial intelligence technology optimizes the arrangement of UV cutting lines on the 3D model, automatically generating a UV cutting line arrangement that matches the actual shape of the 3D model without requiring manual adjustment. This saves significant time and labor costs and improves the efficiency of 3D modeling.

[0166] exist Figure 9 Based on the device shown, the device provided in this embodiment of the invention can be further extended to include multiple units. The functions of each unit can be found in the descriptions of the various embodiments of the UV cutting method for three-dimensional models provided above, and will not be further illustrated here.

[0167] This invention also provides a storage medium that includes stored instructions, wherein when the instructions are executed, the device containing the storage medium is controlled to perform the UV cutting method for the three-dimensional model as described above.

[0168] This invention also provides an electronic device, the structural schematic of which is shown below. Figure 10 As shown, it specifically includes a memory 301 and one or more instructions 302, wherein one or more instructions 302 are stored in the memory 301 and configured to be executed by one or more processors 303 to perform the following operations:

[0169] When UV cutting is required on a target 3D model, determine the silhouette image set and the light and shadow image set corresponding to the target 3D model;

[0170] The silhouette image set is input into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding 3D model, and the sample labels are the subject, appendages, and junction regions of the 3D model;

[0171] The light and shadow image set is input into the second region recognition model to obtain a second subject, a second accessory set, and a second region set; the second accessory set includes at least one accessory, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, accessory, and junction region of the three-dimensional model;

[0172] Based on the first set of regions and the second set of regions, multiple intersection regions to be detected are determined;

[0173] Each of the detected junction regions is input into the junction line recognition model to obtain the first junction line corresponding to each detected junction region; the junction line recognition model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to that junction region;

[0174] Based on each of the first intersection lines, multiple UV cutting lines are determined, and the target 3D model is UV-cut based on each of the UV cutting lines.

[0175] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0176] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0177] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A UV cutting method for a three-dimensional model, characterized in that, include: When UV cutting is required on a target 3D model, determine the silhouette image set and the light and shadow image set corresponding to the target 3D model; The silhouette image set is input into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding 3D model, and the sample labels are the subject, appendages, and junction regions of the 3D model; The light and shadow image set is input into the second region recognition model to obtain a second subject, a second accessory set, and a second region set; the second accessory set includes at least one accessory, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, accessory, and junction region of the three-dimensional model; Based on the first set of regions and the second set of regions, multiple intersection regions to be detected are determined; Each of the detected junction regions is input into the junction line recognition model to obtain the first junction line corresponding to each detected junction region; the junction line recognition model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to that junction region; Based on each of the first intersection lines, multiple UV cutting lines are determined, and the target 3D model is UV-cut based on each of the UV cutting lines.

2. The method according to claim 1, characterized in that, The annotation process for the silhouette samples includes: Based on the silhouette image set of the first 3D model, shape recognition is performed on the first 3D model to obtain multiple first shapes corresponding to the first 3D model; the first 3D model is the 3D model corresponding to the silhouette sample. Determine the connection relationships between each of the first shapes, and based on the connection relationships between each of the first shapes, determine the intersection areas between each of the first shapes; Determine the silhouette proportion corresponding to each of the first shapes; Based on the silhouette proportions of each of the first shapes, a first main shape is determined among the first shapes, and each of the first shapes other than the first main shape is designated as a first subordinate shape. The intersection areas between each of the first shape bodies are labeled as the intersection areas of the first three-dimensional model, the first main shape body is labeled as the main body of the first three-dimensional model, and each of the first subordinate shape bodies is labeled as the subordinate body of the first three-dimensional model.

3. The method according to claim 1, characterized in that, The annotation process for the light and shadow samples includes: Based on the light and shadow image set of the second 3D model, the shadow regions of the second 3D model are determined; the second 3D model is the 3D model corresponding to the light and shadow sample. Based on the light and shadow image set of each shadow region and the second three-dimensional model, shape recognition is performed on the second three-dimensional model to obtain multiple second shapes corresponding to the second three-dimensional model; Determine the connection relationships between each of the second shapes, and based on the connection relationships between each of the second shapes, determine the intersection areas between each of the second shapes; Determine the light and shadow ratio corresponding to each of the second shapes; Based on the light and shadow ratio corresponding to each of the second shapes, a second main shape is determined in each of the second shapes, and each of the second shapes other than the second main shape is regarded as a second subordinate shape. The intersection areas between the various second shape bodies are labeled as the intersection areas of the second three-dimensional model, the second main shape body is labeled as the main body of the second three-dimensional model, and each second subordinate shape body is labeled as a subordinate body of the second three-dimensional model.

4. The method according to claim 1, characterized in that, The annotation process for the samples in the handover area includes: Multiple candidate intersection lines and multiple contour lines are determined corresponding to the target intersection region; the target intersection region is the intersection region corresponding to the sample of the intersection region; each of the candidate intersection lines intersects with each of the contour lines respectively; For each candidate intersection line, determine each target position corresponding to the candidate intersection line, and each target position corresponds one-to-one with each contour line. The target position corresponding to each contour line is the intersection point of the contour line and the candidate intersection line. For each candidate intersection line, the priority of the candidate intersection line is determined based on the contour shape of the corresponding contour line at each target position. Select the candidate intersection line with the highest priority from all the candidate intersection lines, and mark the candidate intersection line as the intersection line corresponding to the target intersection area.

5. The method according to claim 4, characterized in that, The step of determining the priority of a candidate intersection line based on the contour shape of the corresponding contour line at each target position includes: For each target position corresponding to the candidate intersection line, determine the contour angle of the contour line corresponding to the target position at the target position, determine whether the contour angle is greater than a preset angle threshold, and determine the priority corresponding to the target position based on the determination result; The priority of the candidate handover line is determined based on the priority of each target location corresponding to the candidate handover line.

6. The method according to claim 1, characterized in that, The determination of multiple UV cutting lines based on each of the first intersection lines includes: Based on the first subject and the second subject, the target subject corresponding to the target 3D model is determined; Based on the first set of appendages and the second set of appendages, determine each target appendage corresponding to the target 3D model; Determine the inner orientation and intersection point position of each target appendage; In each of the first handover lines, the first handover line corresponding to each of the target appendages is determined; For each target appendage, the first intersection line is adjusted according to the inner orientation and the intersection proximity point of the target appendage to obtain the second intersection line corresponding to the first intersection line. Each of the second intersection lines is taken as the UV cutting line.

7. The method according to claim 1, characterized in that, The determination of multiple UV cutting lines based on each of the first intersection lines includes: Each of the first intersection lines is taken as the UV cutting line.

8. A UV cutting device for a three-dimensional model, characterized in that, include: The first determining unit is used to determine the silhouette image set and the light and shadow image set corresponding to the target three-dimensional model when UV cutting is required; The first recognition unit is used to input the silhouette image set into the first region recognition model to obtain a first subject, a first set of appendages, and a first region set; the first set of appendages includes at least one appendage, and the first region set includes at least one junction region; the first region recognition model is a model trained based on multiple labeled silhouette samples; the sample features of each silhouette sample are the silhouette image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model; The second recognition unit is used to input the light and shadow image set into the second region recognition model to obtain a second subject, a second set of appendages, and a second region set; the second set of appendages includes at least one appendage, and the second region set includes at least one junction region; the second region recognition model is a model trained based on multiple labeled light and shadow samples; the sample features of each light and shadow sample are the light and shadow image set of its corresponding three-dimensional model, and the sample labels are the subject, appendages, and junction regions of the three-dimensional model; The second determining unit is used to determine multiple intersection regions to be detected based on the first region set and the second region set; The third identification unit is used to input each of the detected junction regions into the junction line identification model to obtain the first junction line corresponding to each of the detected junction regions; the junction line identification model is a model trained based on multiple labeled junction region samples; the sample features of each junction region sample are its corresponding junction region, and the sample label is the junction line corresponding to the junction region; The third determining unit is used to determine multiple UV cutting lines based on each of the first intersection lines, and to perform UV cutting on the target three-dimensional model based on each of the UV cutting lines.

9. A storage medium, characterized in that, The storage medium includes stored instructions, wherein, when the instructions are executed, the device containing the storage medium is controlled to perform the UV cutting method for the three-dimensional model as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes a memory, and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors using the UV cutting method for a three-dimensional model as described in any one of claims 1 to 7.

Images