Anchor point generation method and device for three-dimensional scene

Abstract

The invention discloses an anchor point generation method and device for a three-dimensional scene. The method comprises the following steps: determining a target object in the three-dimensional scene; determining a target observation distance of the target object; and determining a target observation visual angle of the target object based on the target observation distance, wherein the observation position coordinates corresponding to the target observation visual angle and the observation angle form anchor point information.

Description

Technical Field

[0001] The present application relates to the technical field of three-dimensional modeling, and more specifically, to a method and device for generating anchor points of a three-dimensional scene. Background Art

[0002] Anchors in a 3D scene are used to mark specific locations or objects in the scene, further providing users with an interactive navigation experience, allowing users to move and explore more conveniently in a virtual 3D scene with the help of anchors. The current creation of anchors mainly relies on developers to manually create anchors in development tools. However, this method of creating anchors requires developers to manually create them one by one using development tools, which consumes human resources and time. It also limits the number of available anchors in large 3D scenes, affecting the user experience in large scenes. Summary of the invention

[0003] In view of this, this application provides the following technical solutions:

[0004] The first aspect of the present application provides a method for generating anchor points of a three-dimensional scene, comprising:

[0005] Determine the target object in the three-dimensional scene;

[0006] Determining a target observation distance of the target object;

[0007] The target observation angle of the target object is determined based on the target observation distance, and the observation position coordinates and the observation angle corresponding to the target observation angle constitute anchor point information.

[0008] In one possible implementation, the method further includes:

[0009] Receive operation information;

[0010] In a case where the operation information corresponds to the anchor point information, rendering a target image according to the anchor point information and the three-dimensional scene;

[0011] The target image is output.

[0012] In a possible implementation, determining the target object in the three-dimensional scene includes:

[0013] Obtaining semantic information of a plurality of images corresponding to different perspectives in the three-dimensional scene;

[0014] A target object in the three-dimensional scene is determined based on the semantic information.

[0015] In a possible implementation, obtaining semantic information of multiple images corresponding to different perspectives in a three-dimensional scene includes:

[0016] Semantic segmentation is performed on multiple images corresponding to different perspectives in a three-dimensional scene to obtain semantic information including semantic segmentation images, semantic categories, and coordinates of semantic segmentation objects.

[0017] In a possible implementation, determining the target object in the three-dimensional scene based on the semantic information includes:

[0018] Determining a target area of ​​the three-dimensional scene based on boundary data of the three-dimensional scene;

[0019] At least one object whose coordinates are located in the target area, whose screen occupancy in the image of the three-dimensional scene reaches a set value, and whose semantic category meets preset conditions is determined as a target object.

[0020] In a possible implementation, determining the target observation distance of the target object includes at least one of the following:

[0021] Determine the observation distance corresponding to when the proportion of the target object in the screen image of the three-dimensional scene is within a set range as the target observation distance;

[0022] The target observation distance of the target object is determined based on the size information of the target object and the field of view of the image of the three-dimensional scene containing the target object.

[0023] In a possible implementation, determining the target observation viewing angle of the target object based on the target observation distance includes:

[0024] Based on the target observation distance, a plurality of observation images corresponding to observing the target object at different viewing angles are obtained;

[0025] Object recognition processing is performed on the multiple observation images, and a target observation angle of the target object is determined based on the recognition processing result.

[0026] In a possible implementation, performing object recognition processing on the plurality of observation images and determining a target observation angle of the target object based on the recognition processing result includes:

[0027] Performing object recognition processing on the plurality of observation images to obtain corresponding groups of recognition results, each group of the recognition results including a recognition category and a confidence level;

[0028] The viewing angle corresponding to the image with the highest confidence among the observation images whose recognition categories are the same as the semantic category of the target object is determined as the target observation viewing angle.

[0029] In a possible implementation, the anchor point information further includes anchor point identification data, and the anchor point identification data represents a relative position relationship between the anchor point identification and the target object.

[0030] A second aspect of the present application provides a device for generating anchor points of a three-dimensional scene, comprising:

[0031] An object determination module, used to determine a target object in a three-dimensional scene;

[0032] A distance determination module, used to determine the target observation distance of the target object;

[0033] The anchor point determination module is used to determine the target observation angle of the target object based on the target observation distance, and the observation position coordinates and observation angle corresponding to the target observation angle constitute the anchor point information. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0035] Figure 1 A flowchart of a method for generating anchor points of a three-dimensional scene disclosed in an embodiment of the present application;

[0036] Figure 2 A flowchart of determining a target object disclosed in an embodiment of the present application;

[0037] Figure 3 The images of a target object disclosed in this embodiment at different viewing angles;

[0038] Figure 4 A flow chart of determining a target observation angle of a target object disclosed in an embodiment of the present application;

[0039] Figure 5 This is an example diagram of observation images of the same object from different perspectives disclosed in the embodiments of the present application;

[0040] Figure 6 This is a schematic diagram of the implementation process of the solution for automatically generating anchor points disclosed in the embodiment of the present application;

[0041] Figure 7 An anchor point generating device for a three-dimensional scene disclosed in an embodiment of the present application;

[0042] Figure 8 A schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0044] The embodiments of the present application can be applied to electronic devices. The present application does not limit the product form of the electronic device, which can include but is not limited to smart phones, tablet computers, wearable devices, personal computers (PCs), netbooks, etc., and can be selected according to application requirements.

[0045] Figure 1 This is a flow chart of a method for generating anchor points of a three-dimensional scene disclosed in an embodiment of the present application. Figure 1 As shown, the method for generating anchor points of a three-dimensional scene may include:

[0046] Step 101: Determine a target object in a three-dimensional scene.

[0047] The three-dimensional scene may be a three-dimensional scene that needs to be displayed in different applications and different devices, such as a scene in a virtual house viewing application, a scene in a game application, a scene displayed by a smart glasses device, etc. The target object is a virtual object existing in the three-dimensional scene. The target object may be an object identified by matching the relevant information input by the user, such as a user inputting a voice "animal" or inputting the text "animal" in an input box indicating the target object, then the virtual animal in the three-dimensional scene may be determined as the target object; the target object may also be an object directly determined based on a user's selection operation, such as a user directly clicking on an object in the three-dimensional scene with a mouse, finger, etc., then the object may be determined as the target object.

[0048] In implementation, the object categories that need to be paid attention to in the 3D scene can be determined according to the application purpose and / or user expectations of the 3D scene, and recorded as required data. For example, in a 3D scene of an art exhibition, the main object categories of interest are exhibits; or in a 3D scene of a scenic spot, the main object categories of interest may include the main entrances of buildings, cultural relics such as tables and chairs, etc.; or in a 3D scene of a high-tech park, the main objects of interest are displays, offices, promotional posters, etc.

[0049] There may be one or more target objects in a three-dimensional scene, and there is no fixed limit to the number. For each determined target object, the anchor point generation method disclosed in the embodiment of the present application may be used to obtain its anchor point information.

[0050] Step 102: Determine the target observation distance of the target object.

[0051] After determining the target object, the target observation distance of the target object may be determined. The target observation distance may be determined based on a preset standard, such as the proportion of the target object in the three-dimensional scene image.

[0052] In different application scenarios, the standard for determining the target observation distance can be different to meet the needs of different theme three-dimensional scenes. Of course, due to the different preferences and habits of different users, the standard for determining the target observation distance can also be set as a configurable parameter with a default value, so that the user can configure it based on their needs.

[0053] Step 103: Determine the target observation angle of the target object based on the target observation distance, and the observation position coordinates and observation angle corresponding to the target observation angle constitute anchor point information.

[0054] After determining the target observation distance, the target observation angle of the target object can be further determined based on the target distance. The target observation angle can be understood as the best observation angle of the target object, which should enable the user to clearly see the shape of the target object and enable the user to more easily identify the category of the target object.

[0055] The anchor point information may include the observation position coordinates and the observation angle corresponding to the target observation angle, and these two information may be directly determined based on the target observation angle. In other implementations, in addition to the observation position coordinates and the observation angle corresponding to the target observation angle, the anchor point information may also include other information, such as the anchor point display position, the anchor point display method, the anchor point appearance, etc. This application does not impose fixed restrictions on the content of the anchor point information.

[0056] The anchor point generation method for a three-dimensional scene described in this embodiment can automatically determine the target observation distance and target observation viewing angle of the target object after determining the target object in the three-dimensional scene, and can obtain the anchor point information of the target object by combining the known coordinate information of the target object; the scheme for generating anchor point information can automatically generate anchor point information in batches, and compared with the traditional scheme of manually creating anchor point information, it is not only efficient, but also can meet the anchor point generation needs of three-dimensional scenes of different scales.

[0057] In other implementations, in addition to the steps disclosed in the aforementioned embodiments, the method for generating an anchor point for a three-dimensional scene may also include: receiving operation information; when the operation information corresponds to the anchor point information, rendering a target image according to the anchor point information and the three-dimensional scene; and outputting the target image.

[0058] The operation information is information triggered by the user. For example, the user directly "clicks" the anchor point mark corresponding to a certain exhibit in the virtual three-dimensional scene of the museum with his finger, or the user controls the cursor to click the anchor point mark of the exhibit in the three-dimensional scene of the museum with the mouse. The background (three-dimensional scene server) will render the display image of the above exhibit according to the anchor point information corresponding to the anchor point mark and the three-dimensional scene data, and then transmit the rendered image to the terminal device for display, so that the user can view the exhibit at the best observation angle.

[0059] In the aforementioned embodiment, determining the target object in the three-dimensional scene may include: obtaining semantic information of multiple images corresponding to different perspectives in the three-dimensional scene; and determining the target object in the three-dimensional scene based on the semantic information.

[0060] The images of different perspectives in the three-dimensional scene may be rendered images of the three-dimensional scene, or may be original photographed images used to generate the three-dimensional scene.

[0061] Determine the target object of the 3D scene, that is, determine the core semantic point in the 3D scene. The target object can be understood as the subject to which the anchor point is to be attached. Specifically, semantic segmentation can be performed on multiple images corresponding to different perspectives in the 3D scene to obtain semantic information including semantic segmentation images, semantic categories, and coordinates of semantic segmentation objects.

[0062] In the implementation, the image of the three-dimensional scene can be processed based on the semantic segmentation model to obtain a picture containing detailed semantic information and text information. The picture containing semantic information is also a semantic segmentation image, which has multiple semantic segmentation objects. Different categories of semantic segmentation objects are marked with different colors. The text information records the semantic categories of the objects with different color marks; in addition, the text information can also include the coordinates of the semantic segmentation objects.

[0063] Figure 2 This is a flow chart of determining a target object disclosed in an embodiment of the present application. Figure 2 As shown, determining the target object in the three-dimensional scene based on the semantic information may include:

[0064] Step 201: Determine a target area of ​​the three-dimensional scene based on boundary data of the three-dimensional scene.

[0065] Specifically, the middle area of ​​the three-dimensional scene whose area accounts for a set ratio value can be determined as the target area of ​​the three-dimensional scene, such as the set ratio value can be 70%; the shape of the target area is the same as or different from the boundary shape of the three-dimensional scene. If the boundary shape of the three-dimensional scene is a rectangle, the shape of the target area can be a rectangle, or other shapes such as a circle or a square. The boundary data of the three-dimensional scene is known when it is constructed, so the overall area of ​​the three-dimensional scene is also known or can be determined, so that the middle area of ​​the three-dimensional scene whose area accounts for a set ratio value can be determined as the target area of ​​the three-dimensional scene.

[0066] Step 202: Determine at least one object whose coordinates are located in the target area, whose screen occupancy in the image of the three-dimensional scene reaches a set value, and whose semantic category meets preset conditions as a target object.

[0067] Among them, the requirement that the screen ratio in the image of the three-dimensional scene reaches a set value can prevent some too small objects in the three-dimensional scene from being identified as target objects. These relatively small objects are usually not the subject of user attention. The preset conditions can be the objects that the user needs to pay attention to in the three-dimensional scene mentioned above. For example, there are objects such as houses, roads, and flowers in the target area, and the objects that the user determines to be concerned about are buildings. In this case, only the houses in the target area will be determined as target objects, and the roads and flowers will not be determined as target objects. In implementation, different preset conditions can also be set for the three-dimensional scene, so that different categories of anchor points can also be generated.

[0068] In other implementations, the process of determining the target object may further include: determining the target main viewing angle of the three-dimensional scene based on the three-dimensional scene image triggered and selected by the user. Then, determining as the target object one or more objects whose coordinates are located in the target area, whose screen proportion in the image of the three-dimensional scene reaches a set value, and whose semantic category meets the preset conditions may include: determining as the target object one or more objects whose coordinates are located in the target area, whose screen proportion in the image corresponding to the target main viewing angle reaches a set value, and whose semantic category meets the preset conditions.

[0069] The user can select several recommended main viewing angles from the screen images of the 3D scene. The recommended main viewing angles include the objects that the user is interested in (target objects) and can also represent the best viewing angle of the 3D scene. Under the best viewing angle, the target object in the 3D scene can be clearly displayed. In the implementation, an optimal image can be evaluated from the 3D scene images selected by the user, and the viewing angle corresponding to the image is determined as the target main viewing angle. Figure 3 For images of a target object at different viewing angles, the viewing angle corresponding to the image on the right that can better display the animal's morphology can be determined as the target main viewing angle according to the user's selection.

[0070] The above content introduces in detail various implementations of determining the target object, which is convenient for technical personnel in the field to better understand and implement the technical solution of the present application.

[0071] In the aforementioned embodiments, the determination of the target observation distance of the target object can also be implemented in different ways, two of which will be described below.

[0072] In one implementation, the observation distance corresponding to when the proportion of the target object in the screen image of the three-dimensional scene is within a set range can be determined as the target observation distance. For example, for the target object, the proportion of the target object in the screen image is about 60% as a standard, and the optimal observation distance of the observation point relative to the target object, that is, the target observation distance, is determined.

[0073] In another implementation, considering that the same target object may present different shapes and sizes under different field of view angles, in order to better determine the target observation distance, in this implementation, the target observation distance of the target object can be comprehensively determined based on the size information of the target object and the field of view angle of the three-dimensional scene image containing the target object. The size information can characterize the size of the target object, and the field of view angle can match the presentation shape of the target object. The target observation position determined by combining the two can ensure that the shape recognition of the target object is high (enabling the user to more easily determine the category of the target object), and the size of the observed target object in the picture is appropriate. Specifically, the initial distance can be calculated first based on the size information of the target object, and then the field of view angle that satisfies the initial distance and can clearly identify the shape of the target object can be determined based on the field of view angle of the three-dimensional scene image, and finally the target observation distance containing the field of view angle information is determined.

[0074] It is understandable that due to the different shapes and sizes of different target objects, the target observation distances of different target objects will also be different. In the application, it is only necessary to determine the target observation distance based on a pre-set standard, such as the proportion of the aforementioned area in the three-dimensional scene image, or a pre-set distance determination logic (related to the aforementioned size information and field of view).

[0075] After the target observation distance is determined, the target observation viewing angle of the target object may be determined based on the target observation distance. Figure 4 This is a flow chart of determining the target observation angle of the target object disclosed in the embodiment of the present application. Figure 4 As shown, determining the target observation angle may include:

[0076] Step 401: Based on the target observation distance, a plurality of observation images corresponding to observing the target object at different viewing angles are obtained.

[0077] After determining the target observation distance, the target observation distance can be used as a reference to circle around the target object in a virtual three-dimensional scene. During the process, multiple observation images are randomly captured from observation angles, such as 10 observation images. The viewing angles of these observation images are all different. It can be understood that a person holds a camera and circles around an object, taking multiple photos of the object from different angles. Figure 5 This is an example of an observation image of the same object from different perspectives disclosed in the embodiment of the present application, which can be combined with Figure 5 Understand the implementation of step 401.

[0078] Step 402: performing object recognition processing on the plurality of observation images, and determining a target observation angle of the target object based on the recognition processing result.

[0079] After obtaining multiple observation images, each observation image can be respectively recognized and processed based on the object recognition model to obtain a recognition result, and the recognition result can include but is not limited to including a recognition category and a confidence level. Based on this, object recognition processing is performed on the multiple observation images, and the target observation angle of the target object is determined based on the recognition processing results, which can include: performing object recognition processing on the multiple observation images to obtain corresponding multiple groups of recognition results, each group of the recognition results includes a recognition category and a confidence level; and determining the angle corresponding to the image with the highest confidence level among the observation images whose recognition category is the same as the semantic category of the target object as the target observation angle.

[0080] The object recognition model can be any object recognition model that already exists in the field or may appear in the future, which can be a convolutional neural network, a recurrent neural network, a generative adversarial network, etc. Specifically, it can be YOLOv8. In an example, the obtained multiple observation images are sequentially provided as input to the object recognition model YOLOv8, and the known semantic category of the target object is used as a reference to perform target recognition on these observation images. The result of YOLO target recognition will include the category and confidence score of the object; the confidence score is a value of 0-1, indicating the degree of confidence of the YOLOv8 model for the detected target; based on the confidence score, the observation images with the same semantic category as the target object are sorted based on the confidence score, and the perspective of the picture with the highest confidence is the best observation perspective for observing the target object, that is, the target observation perspective.

[0081] Once the target observation angle is determined, the anchor point information including the observation position and observation angle of the target object can be determined. In other implementations, the anchor point information may also include anchor point identification data, which characterizes the relative position relationship between the anchor point identification and the target object. That is, it is set where the anchor point identification should be displayed on the target object, such as above, in front, upper right, etc.

[0082] Figure 6 This is a schematic diagram of the implementation process of the solution for automatically generating anchor points disclosed in the embodiment of this application. Figure 6 As shown, in a more detailed implementation, the three-dimensional scene image is first input into the semantic segmentation model for semantic segmentation processing to obtain the semantic segmentation image and text information, which can be combined with the three-dimensional scene data (including object coordinates) to obtain semantic information; then, the target object is determined based on the semantic information, the required data in the three-dimensional scene data (limiting the user's attention object category), the recommended main viewing angle of the three-dimensional image, and the target area obtained based on the three-dimensional scene boundary data; on the basis of the determined target object, the target observation viewing angle of the target object is determined by combining the three-dimensional scene image with the object recognition model, and after the target observation viewing angle is determined, the anchor point information can be determined.

[0083] The present application scheme realizes the automatic generation of anchor points in three-dimensional scenes based on semantic segmentation models and object recognition models. In the process, the existing data of the three-dimensional scene (such as coordinates, boundaries, images and other data) is fully utilized, and the anchor points are automatically generated, saving human resources. At the same time, since the anchor points can be generated in batches, the number is not limited, which can improve the user experience in large three-dimensional scenes.

[0084] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0085] The method is described in detail in the embodiments disclosed in the above-mentioned application. The method of the application can be implemented by various forms of devices. Therefore, the application also discloses a device, and a specific embodiment is given below for detailed description.

[0086] Figure 7 This is a device for generating anchor points of a three-dimensional scene disclosed in an embodiment of the present application. Figure 7 As shown, the anchor point generating device 70 may include:

[0087] The object determination module 701 is used to determine a target object in a three-dimensional scene.

[0088] The distance determination module 702 is used to determine the target observation distance of the target object.

[0089] The anchor point determination module 703 is used to determine the target observation angle of the target object based on the target observation distance. The observation position coordinates and the observation angle corresponding to the target observation angle constitute the anchor point information.

[0090] The anchor point generation device for the three-dimensional scene described in this embodiment can automatically determine the target observation distance and target observation viewing angle of the target object after determining the target object in the three-dimensional scene, and can obtain the anchor point information of the target object in combination with the known coordinate information of the target object; the scheme for generating anchor point information can automatically generate anchor point information in batches, and compared with the traditional scheme of manually creating anchor point information, it is not only efficient, but also can meet the anchor point generation needs of three-dimensional scenes of different scales.

[0091] In one implementation, the object determination module may include: a semantic processing module for obtaining semantic information of multiple images corresponding to different perspectives in the three-dimensional scene; and an object determination submodule for determining a target object in the three-dimensional scene based on the semantic information.

[0092] In one implementation, the semantic processing module can be specifically used to perform semantic segmentation processing on multiple images corresponding to different perspectives in a three-dimensional scene to obtain semantic information including semantic segmentation images, semantic categories, and coordinates of semantic segmentation objects.

[0093] In one implementation, the object determination submodule may include: a target area determination module, used to determine the target area of ​​the three-dimensional scene based on the boundary data of the three-dimensional scene; an object selection module, used to determine as a target object at least one object whose coordinates are located in the target area, whose screen occupancy in the image of the three-dimensional scene reaches a set value and whose semantic category meets preset conditions.

[0094] In one implementation, the target area determination module can be specifically used to: determine the middle area of ​​the three-dimensional scene whose area accounts for a set proportion value as the target area of ​​the three-dimensional scene, and the shape of the target area is the same as or different from the boundary shape of the three-dimensional scene.

[0095] In one implementation, the device may also include: a main viewing angle determination module, used to determine the target main viewing angle of the three-dimensional scene based on the three-dimensional scene image triggered and selected by the user; then the object selection module can be specifically used to: determine one or more objects whose coordinates are located in the target area, whose screen occupancy in the image corresponding to the target main viewing angle reaches a set value and whose semantic category meets preset conditions as target objects.

[0096] In one implementation, the distance determination module is specifically used to: determine the observation distance corresponding to when the proportion of the target object in the screen image of the three-dimensional scene is within a set range as the target observation distance; or, determine the target observation distance of the target object based on the size information of the target object and the field of view angle of the image of the three-dimensional scene containing the target object.

[0097] In one implementation, the anchor point determination module may include: an observation image acquisition module, which is used to obtain multiple observation images corresponding to observing the target object at different perspectives based on the target observation distance; a perspective determination module, which is used to perform object recognition processing on the multiple observation images and determine the target observation perspective of the target object based on the recognition processing results.

[0098] In one implementation, the perspective determination module can be specifically used to: perform object recognition processing on the multiple observation images to obtain corresponding multiple groups of recognition results, each group of recognition results including a recognition category and a confidence level; and determine the perspective corresponding to the image with the highest confidence level among the observation images whose recognition category is the same as the semantic category of the target object as the target observation perspective.

[0099] In one implementation, the anchor point information further includes anchor point identification data, and the anchor point identification data represents a relative position relationship between the anchor point identification and the target object.

[0100] The anchor point generation device for any one of the three-dimensional scenes described in the above embodiments includes a processor and a memory. The object determination module, distance determination module, anchor point determination module, viewing angle determination module, observation image acquisition module, main viewing angle determination module, etc. in the above embodiments are all stored in the memory as program modules, and the processor executes the above program modules stored in the memory to implement corresponding functions.

[0101] The processor includes a kernel, which retrieves the corresponding program module from the memory. One or more kernels can be set, and the processing of the access data can be realized by adjusting the kernel parameters.

[0102] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0103] In an exemplary embodiment, a computer-readable storage medium is also provided, which can be directly loaded into the internal memory of a computer and contains software code. After being loaded and executed by a computer, the computer program can implement the steps shown in any embodiment of the above-mentioned method for generating anchor points of a three-dimensional scene.

[0104] In an exemplary embodiment, a computer program product is also provided, which can be directly loaded into the internal memory of a computer and contains software code. After being loaded and executed by a computer, the computer program can implement the steps shown in any embodiment of the above-mentioned method for generating anchor points of a three-dimensional scene.

[0105] Furthermore, an embodiment of the present application provides an electronic device. Figure 8This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application. Figure 8 As shown, the electronic device 80 includes at least one processor 801, and at least one memory 802 and a bus 803 connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is used to call the program instructions in the memory to execute the above-mentioned anchor point generation method for the three-dimensional scene.

[0106] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0107] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0108] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0109] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be 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 present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating anchor points of a three-dimensional scene, comprising: Determine the target object in the three-dimensional scene; Determining a target observation distance of the target object; The target observation angle of the target object is determined based on the target observation distance, and the observation position coordinates and the observation angle corresponding to the target observation angle constitute anchor point information.

2. The method for generating anchor points of a three-dimensional scene according to claim 1, further comprising: Receive operation information; In a case where the operation information corresponds to the anchor point information, rendering a target image according to the anchor point information and the three-dimensional scene; The target image is output.

3. The method for generating anchor points of a three-dimensional scene according to claim 1, wherein determining the target object in the three-dimensional scene comprises: Obtaining semantic information of a plurality of images corresponding to different perspectives in the three-dimensional scene; A target object in the three-dimensional scene is determined based on the semantic information.

4. The method for generating anchor points of a three-dimensional scene according to claim 3, wherein obtaining semantic information of a plurality of images corresponding to different perspectives in the three-dimensional scene comprises: Semantic segmentation is performed on multiple images corresponding to different perspectives in a three-dimensional scene to obtain semantic information including semantic segmentation images, semantic categories, and coordinates of semantic segmentation objects.

5. The method for generating anchor points of a three-dimensional scene according to claim 4, wherein determining the target object in the three-dimensional scene based on the semantic information comprises: Determining a target area of ​​the three-dimensional scene based on boundary data of the three-dimensional scene; At least one object whose coordinates are located in the target area, whose screen occupancy in the image of the three-dimensional scene reaches a set value, and whose semantic category meets preset conditions is determined as a target object.

6. The method for generating anchor points of a three-dimensional scene according to claim 1, wherein determining the target observation distance of the target object comprises at least one of the following: Determine the observation distance corresponding to when the proportion of the target object in the screen image of the three-dimensional scene is within a set range as the target observation distance; The target observation distance of the target object is determined based on the size information of the target object and the field of view of the image of the three-dimensional scene containing the target object.

7. The method for generating anchor points of a three-dimensional scene according to claim 1, wherein determining the target observation angle of the target object based on the target observation distance comprises: Based on the target observation distance, a plurality of observation images corresponding to observing the target object at different viewing angles are obtained; Object recognition processing is performed on the multiple observation images, and a target observation angle of the target object is determined based on the recognition processing result.

8. The method for generating anchor points of a three-dimensional scene according to claim 7, wherein the performing object recognition processing on the plurality of observation images and determining the target observation angle of the target object based on the recognition processing result comprises: Performing object recognition processing on the plurality of observation images to obtain corresponding groups of recognition results, each group of the recognition results including a recognition category and a confidence level; The viewing angle corresponding to the image with the highest confidence among the observation images whose recognition categories are the same as the semantic category of the target object is determined as the target observation viewing angle.

9. The method for generating anchor points of a three-dimensional scene according to claim 1, wherein: The anchor point information also includes anchor point identification data, and the anchor point identification data represents the relative position relationship between the anchor point identification and the target object.

10. A device for generating anchor points of a three-dimensional scene, comprising: An object determination module, used to determine a target object in a three-dimensional scene; A distance determination module, used to determine the target observation distance of the target object; The anchor point determination module is used to determine the target observation angle of the target object based on the target observation distance, and the observation position coordinates and observation angle corresponding to the target observation angle constitute the anchor point information.

Images