Spatial semantic occupation prediction method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a space semantic occupation prediction method and device, computer equipment, a readable storage medium and a program product. The method comprises the following steps: acquiring original point cloud data in a target space scene; converting the original point cloud data into a multi-view feature plane, and performing feature extraction on the multi-view feature plane to obtain multi-view point cloud features; inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and performing feature extraction on the two-dimensional semantic image to obtain multi-view image features; fusing the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; and for any query point in the target space scene, according to the multi-view fusion feature corresponding to the query point, outputting a space semantic occupation prediction result. By adopting the method, the efficiency of space semantic occupation prediction can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial technology, and in particular to a method, apparatus, computer device, computer-readable storage medium and computer program product for predicting spatial semantic occupancy. Background Art

[0002] The spatial semantic occupancy prediction task aims to build a more accurate model of the three-dimensional spatial scene. This construction process relies on assigning a clear semantic label to each voxel in the three-dimensional spatial scene.

[0003] However, point cloud voxel representation is highly dense in three-dimensional space scenes, which means that the spatial semantic occupancy prediction methods in related technologies often require a series of operations on a large amount of three-dimensional point cloud data, which faces a huge computational burden and is not conducive to improving the efficiency of spatial semantic occupancy prediction.

[0004] Therefore, the spatial semantic occupancy prediction in the related art has the problem of low efficiency. Summary of the invention

[0005] Based on this, it is necessary to provide a spatial semantic occupancy prediction method, apparatus, computer device, computer-readable storage medium and computer program product that can improve the efficiency of spatial semantic occupancy prediction in response to the above technical problems.

[0006] In a first aspect, the present application provides a spatial semantic occupancy prediction method, the method comprising:

[0007] Obtain the original point cloud data in the target space scene;

[0008] The original point cloud data is converted into a multi-view feature plane, and features are extracted from the multi-view feature plane to obtain multi-view point cloud features;

[0009] Inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and performing feature extraction on the two-dimensional semantic image to obtain multi-view image features;

[0010] Fusion of the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features;

[0011] For any query point located in the target space scene, a spatial semantic occupancy prediction result is output according to the multi-view fusion feature corresponding to the query point.

[0012] In one of the embodiments, outputting a spatial semantic occupancy prediction result according to the multi-view fusion feature corresponding to the query point includes:

[0013] Obtaining the fusion features corresponding to each perspective in the multi-perspective fusion features corresponding to the query point;

[0014] Fusing the corresponding fusion features under each viewing angle to obtain the three-dimensional feature corresponding to the query point;

[0015] The three-dimensional features of the query point are input into a pre-trained three-dimensional feature classification model to obtain the spatial semantic occupancy prediction result.

[0016] In one embodiment, converting the original point cloud data into a multi-view feature plane includes:

[0017] Dividing the original point cloud data into voxels in a cylindrical coordinate system to obtain voxelized point cloud data;

[0018] A pooling operation is performed on the voxelized point cloud data under three coordinate axes of the cylindrical coordinate system to obtain plane features under three viewing angles in the cylindrical coordinate system;

[0019] The plane features under the three viewing angles are obtained as the multi-view feature plane.

[0020] In one embodiment, extracting features from the multi-view feature plane to obtain multi-view point cloud features includes:

[0021] Encoding the multi-view feature plane through an encoder in a point cloud feature extraction model to obtain a feature plane encoding result;

[0022] The feature plane encoding result is decoded by a decoder in the point cloud feature extraction model to obtain the multi-view point cloud features at the target resolution.

[0023] In one embodiment, extracting features from the two-dimensional semantic image to obtain multi-view image features includes:

[0024] Encoding the two-dimensional semantic image through an encoder in an image feature extraction model to obtain a semantic image encoding result;

[0025] The semantic image encoding result is decoded by a decoder in the image feature extraction model to obtain the multi-view image features at the target resolution.

[0026] In one embodiment, the obtaining of original point cloud data in the target space scene includes:

[0027] According to preset sampling configuration information, sampling the point cloud data in the target space scene to obtain sampling data;

[0028] The sampling data is obtained as the original point cloud data.

[0029] In a second aspect, the present application further provides a spatial semantic occupancy prediction device, the device comprising:

[0030] An acquisition module is used to acquire the original point cloud data in the target space scene;

[0031] A conversion module, used for converting the original point cloud data into a multi-view feature plane, performing feature extraction on the multi-view feature plane, and obtaining multi-view point cloud features;

[0032] A generation module, used for inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and performing feature extraction on the two-dimensional semantic image to obtain multi-view image features;

[0033] A fusion module, used for fusing the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features;

[0034] The output module is used to output a spatial semantic occupancy prediction result for any query point located in the target space scene according to the multi-view fusion feature corresponding to the query point.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0037] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of the above method when executed by a processor.

[0038] The above-mentioned spatial semantic occupancy prediction method, device, computer equipment, computer-readable storage medium and computer program product obtain original point cloud data in the target spatial scene, convert the original point cloud data into a multi-view feature plane, extract features from the multi-view feature plane, and obtain multi-view point cloud features; obtain a two-dimensional semantic image by inputting the multi-view point cloud features into a pre-trained visual generative model, extract features from the two-dimensional semantic image, and obtain multi-view image features; then fuse the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; for any query point located in the target spatial scene, output the spatial semantic occupancy prediction result according to the multi-view fusion features corresponding to the query point. Result; In this way, it is possible to use the two-dimensional point cloud features of multiple perspectives and the two-dimensional image features obtained by encoding and decoding the two-dimensional semantic image output by the visual generative model to reduce the information loss in the projection process, which can further improve the model's ability to model complex three-dimensional scenes while maintaining high efficiency. At the same time, multi-perspective point cloud features and multi-perspective image features are fused in parallel, and the features of the query points on the three-perspective planes are aggregated by addition, which can effectively solve the problem that the two-dimensional feature method is difficult to depict the fine structure in complex three-dimensional scenes, effectively realize the use of two-dimensional features to model complex three-dimensional scenes, reduce the calculation and storage burden of the model, and thus improve the efficiency of spatial semantic occupancy prediction based on the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A diagram of an application environment of a spatial semantic occupancy prediction method in one embodiment;

[0041] Figure 2 A schematic diagram of a flow chart of a spatial semantic occupancy prediction method in one embodiment;

[0042] Figure 3 A network structure diagram of a spatial semantic occupancy prediction method in one embodiment;

[0043] Figure 4 A schematic diagram of three-view feature encoding in one embodiment;

[0044] Figure 5 A spatial semantic occupancy prediction system in one embodiment;

[0045] Figure 6is a flow chart of a spatial semantic occupancy prediction method in another embodiment;

[0046] Figure 7 is a structural block diagram of a spatial semantic occupancy prediction device in one embodiment;

[0047] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0049] The spatial semantic occupancy prediction method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the point cloud data acquisition device 102 communicates with the computer device 104. In practical applications, the point cloud data acquisition device 102 can capture the three-dimensional geometric information of various objects or environmental surfaces in the target space scene to obtain point cloud data in the target space scene. The point cloud data acquisition device 102 can transmit the point cloud data in the target space scene to the computer device 104. The computer device 104 obtains the original point cloud data in the target space scene; the computer device 104 converts the original point cloud data into a multi-view feature plane, extracts features from the multi-view feature plane, and obtains multi-view point cloud features; the computer device 104 inputs the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, extracts features from the two-dimensional semantic image, and obtains multi-view image features; the computer device 104 fuses the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; the computer device 104 outputs the spatial semantic occupancy prediction result for any query point located in the target space scene according to the multi-view fusion features corresponding to the query point. The point cloud acquisition device 102 may be, but is not limited to, various laser scanners, photogrammetry equipment, 3D scanners, drones equipped with laser radar or high-definition cameras, mobile devices equipped with depth sensors, and structured light scanners, etc. The computer device 104 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and servers. The server may be an independent physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0050] In an exemplary embodiment, Figure 2 As shown, a spatial semantic occupancy prediction method is provided, which is applied to Figure 1The computer device 104 in the example is used as an example to illustrate the method, which includes the following steps S202 to S208. Among them:

[0051] Step S202, obtaining original point cloud data in the target space scene.

[0052] The target space scene may refer to a three-dimensional space scene for which spatial semantic occupancy prediction is to be performed.

[0053] In a specific implementation, after the computer device acquires the point cloud data transmitted by the point cloud data acquisition device, the computer device can sample the point cloud data in the target space scene to obtain the above-mentioned original point cloud data.

[0054] Step S204: converting the original point cloud data into a multi-view feature plane, performing feature extraction on the multi-view feature plane, and obtaining multi-view point cloud features.

[0055] In a specific implementation, after the computer device acquires the original point cloud data in the target space scene, the computer device can convert the original point cloud data into a multi-view feature plane. Specifically, the computer device can convert the original point cloud data into two-dimensional features at three different viewpoints in a cylindrical coordinate system to obtain feature planes at three different viewpoints. The computer device uses the feature planes at three different viewpoints as the three-view feature plane, i.e., the multi-view feature plane mentioned above.

[0056] After the computer device obtains the multi-view feature plane, it can extract features from the multi-view feature plane to obtain the multi-view point cloud features of the multi-view feature plane. In practical applications, the computer device can use a two-dimensional neural network to extract features from the multi-view feature plane. To encode the three-view feature plane and decode the high-resolution three-view point cloud features; the three-view point cloud features can be expressed as ; Among them, hw, wd, dh represent the three coordinate axes of the cylindrical coordinate system respectively.

[0057] Step S206, inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, performing feature extraction on the two-dimensional semantic image to obtain multi-view image features.

[0058] Among them, the visual generative model can be a large model used to generate new visual content.

[0059] In a specific implementation, after the computer device obtains the multi-view point cloud features, the computer device can input the multi-view point cloud features into a pre-trained visual generative model, and process the multi-view point cloud features through the pre-trained visual generative model to output a two-dimensional semantic image.

[0060] Then, the computer device can extract features from the two-dimensional semantic image to obtain multi-view image features. In practical applications, the computer device can use a two-dimensional neural network to extract features from the two-dimensional semantic image. To encode the two-dimensional semantic image and decode the high-resolution multi-view image features, such as three-view image features. The three-view image features can be expressed as ; Among them, hw, wd, dh represent the three coordinate axes of the cylindrical coordinate system respectively.

[0061] Step S208, fusing the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features.

[0062] In a specific implementation, the computer device can fuse the multi-view point cloud features and the multi-view image features to obtain the multi-view fusion features; specifically, the computer device performs a splicing operation on the multi-view point cloud features and the multi-view image features to obtain the multi-view fusion features, such as the three-view fusion features. Continuing with the above example, the three-view fusion features can be expressed as ; Among them, the fusion features corresponding to the three-view fusion features under three different perspectives can be expressed as:

[0063]

[0064]

[0065]

[0066] in, Represents a concatenation operation.

[0067] Step S210: for any query point located in the target space scene, output a spatial semantic occupancy prediction result according to the multi-view fusion features corresponding to the query point.

[0068] It can be understood that the computer device can achieve modeling of the target space scene using two-dimensional features through the above steps S202, S204 to S208. Then, the computer device can perform spatial semantic occupancy prediction. When the computer device performs spatial semantic occupancy prediction, for any query point located in the target space scene, the computer device can obtain the multi-view fusion feature corresponding to the query point and output the spatial semantic occupancy prediction result.

[0069] Specifically, the computer device can determine the three-dimensional features of the query point based on the multi-view fusion features corresponding to the query point; then, the computer device can output the spatial semantic occupancy prediction result of the query point based on the three-dimensional features of the query point. For example, the semantic label associated with the query point in the target space acquisition. It should be noted that the following will further explain how the computer device outputs the spatial semantic occupancy prediction result based on the multi-view fusion features corresponding to the query point.

[0070] In the above-mentioned spatial semantic occupancy prediction method, by obtaining the original point cloud data in the target spatial scene, and converting the original point cloud data into a multi-view feature plane, the multi-view feature plane is subjected to feature extraction to obtain the multi-view point cloud feature; by inputting the multi-view point cloud feature into a pre-trained visual generative model, a two-dimensional semantic image is obtained, and feature extraction is performed on the two-dimensional semantic image to obtain the multi-view image feature; the multi-view point cloud feature and the multi-view image feature are then fused to obtain the multi-view fusion feature; for any query point located in the target spatial scene, the spatial semantic occupancy prediction result is output according to the multi-view fusion feature corresponding to the query point; in this way, the two-dimensional point cloud features of multiple views and the two-dimensional image features obtained by encoding and decoding the two-dimensional semantic image output by the visual generative model can be used to reduce the information loss in the projection process, and the ability of the model to model complex three-dimensional scenes can be further improved while maintaining high efficiency. At the same time, the multi-view point cloud features and the multi-view image features are fused in parallel to effectively realize the use of two-dimensional features to model complex three-dimensional scenes, reduce the calculation and storage burden of the model, and realize comprehensive two-dimensional scene modeling and efficient spatial semantic occupancy prediction.

[0071] In an exemplary embodiment, a spatial semantic occupancy prediction result is output based on the multi-view fusion features corresponding to the query point, including: obtaining the fusion features corresponding to each view in the multi-view fusion features corresponding to the query point; fusing the fusion features corresponding to each view to obtain the three-dimensional features corresponding to the query point; and inputting the three-dimensional features of the query point into a pre-trained three-dimensional feature classification model to obtain the spatial semantic occupancy prediction result.

[0072] In a specific implementation, when the computer device outputs the spatial semantic occupancy prediction result according to the multi-view fusion features corresponding to the query point, the computer device can obtain the corresponding fusion features under each view in the multi-view fusion features corresponding to the query point. For example, the computer device can obtain the fusion features corresponding to the query point under three perspectives. , and Then, the computer device can aggregate the corresponding fusion features under the three perspectives in an additive manner. , and , get the three-dimensional features corresponding to the query point , expressed as:

[0073]

[0074] Then, the computer device can input the three-dimensional features corresponding to the query point into a pre-trained three-dimensional feature classification model, such as a lightweight classification network; the three-dimensional features corresponding to the query point are processed by the pre-trained three-dimensional feature classification model to obtain a classification result of the query point; the classification result of the query point can represent the semantic label associated with the query point in the target space acquisition, thereby realizing spatial semantic occupancy prediction.

[0075] In order to facilitate the understanding of those skilled in the art, Figure 3 A network structure diagram of a spatial semantic occupancy prediction method is provided as an example; wherein, the point cloud data in the target space scene is input into a computer device, and the computer device samples the point cloud data in the target space scene to obtain the original point cloud data in the target space scene; then, the computer device can perform three-view transformation on the original point cloud data in the target space scene to obtain a three-view feature plane; then, the computer device can perform three-view feature encoding; specifically, the computer device can perform feature extraction on the three-view feature plane to obtain a three-view point cloud feature; the three-view point cloud feature is input into a pre-trained visual generative model to obtain a two-dimensional semantic image, and feature extraction is performed on the two-dimensional semantic image to obtain a three-view image feature; the three-view point cloud feature and the three-view image feature are fused to obtain a three-view fusion feature. Then, the computer device can perform query point feature encoding; specifically, the computer device can obtain the corresponding fusion feature under each view in the three-view fusion feature corresponding to the query point; the corresponding fusion feature under each view is fused to obtain the three-dimensional feature corresponding to the query point. Then, the computer device can input the three-dimensional feature of the query point into a pre-trained three-dimensional feature classification model to output a prediction result.

[0076] The technical solution of this embodiment obtains the three-dimensional features corresponding to the query point by fusing the corresponding fusion features under each perspective in the multi-perspective fusion features corresponding to the query point, and inputs the three-dimensional features of the query point into a pre-trained three-dimensional feature classification model to obtain the spatial semantic occupancy prediction result, thereby realizing the use of an additive method to aggregate the features of the query point on the multi-perspective plane, effectively solving the problem that the two-dimensional feature method is difficult to depict the fine structure in a complex three-dimensional scene, and realizing efficient spatial semantic occupancy prediction.

[0077] In an exemplary embodiment, the original point cloud data is converted into a multi-view feature plane, including: performing voxel division on the original point cloud data in a cylindrical coordinate system to obtain voxelized point cloud data; performing a pooling operation on the voxelized point cloud data under three coordinate axes of the cylindrical coordinate system to obtain plane features under three view angles in the cylindrical coordinate system; and obtaining the plane features under three view angles as a multi-view feature plane.

[0078] In a specific implementation, when a computer device converts the original point cloud data into a multi-view feature plane, the computer device can perform voxel division on the original point cloud data in a cylindrical coordinate system to obtain voxelized point cloud data; then, the computer device can perform a pooling operation on the voxelized point cloud data under the three coordinate axes of the cylindrical coordinate system to obtain plane features under three viewpoints in the cylindrical coordinate system as the multi-view feature plane.

[0079] The technical solution of this embodiment obtains voxelized point cloud data by dividing the original point cloud data into voxels in a cylindrical coordinate system; and obtains planar features under three perspectives in the cylindrical coordinate system by pooling the voxelized point cloud data under the three coordinate axes of the cylindrical coordinate system, thereby realizing efficient three-perspective conversion of the original point cloud data.

[0080] In an exemplary embodiment, feature extraction is performed on a multi-view feature plane to obtain a multi-view point cloud feature, including: encoding the multi-view feature plane through an encoder in a point cloud feature extraction model to obtain a feature plane encoding result; and decoding the feature plane encoding result through a decoder in the point cloud feature extraction model to obtain a multi-view point cloud feature at a target resolution.

[0081] In a specific implementation, when a computer device extracts features from a multi-view feature plane to obtain multi-view point cloud features, the computer device may extract features from the multi-view feature plane through a pre-trained point cloud feature extraction model: specifically, the computer device may encode the multi-view feature plane through an encoder in the point cloud feature extraction model to obtain a feature plane encoding result; then, the computer device may decode the feature plane encoding result through a decoder in the point cloud feature extraction model to obtain multi-view point cloud features at a target resolution.

[0082] The technical solution of this embodiment encodes the multi-view feature plane through the encoder in the point cloud feature extraction model, and decodes the feature plane encoding result through the decoder in the point cloud feature extraction model. It can realize the method of using a two-dimensional neural network to perform encoding and decoding operations on the multi-view feature plane to effectively capture the data features in the multi-view feature plane, and compress and reconstruct the key information in the multi-view feature plane, so as to extract representative features from the multi-view feature plane as multi-view point cloud features.

[0083] In an exemplary embodiment, feature extraction is performed on a two-dimensional semantic image to obtain multi-view image features, including: encoding the two-dimensional semantic image through an encoder in an image feature extraction model to obtain a semantic image encoding result; decoding the semantic image encoding result through a decoder in the image feature extraction model to obtain multi-view image features at a target resolution.

[0084] In a specific implementation, when a computer device extracts features from a two-dimensional semantic image to obtain multi-view image features, the computer device may extract features from the two-dimensional semantic image through a pre-trained image feature extraction model; specifically, the computer device may encode the two-dimensional semantic image through an encoder in the image feature extraction model to obtain a semantic image encoding result; then, the computer device may decode the semantic image encoding result through a decoder in the image feature extraction model to obtain multi-view image features at a target resolution.

[0085] The technical solution of this embodiment encodes the two-dimensional semantic image through the encoder in the image feature extraction model, and decodes the semantic image encoding result through the decoder in the image feature extraction model. It can realize the method of using a two-dimensional neural network to perform encoding and decoding operations on the two-dimensional semantic image to effectively capture the data features in the two-dimensional semantic image, and effectively compress and reconstruct the key information in the two-dimensional semantic image, so as to extract representative features from the two-dimensional semantic image as multi-view image features.

[0086] In order to facilitate the understanding of those skilled in the art, Figure 4 An exemplary schematic diagram of three-view feature encoding is provided. In which, the computer device can input the three-view feature plane 402 into the two-dimensional neural network A, encode the three-view feature plane 402 through the two-dimensional neural network A and decode the three-view point cloud feature 404; then, the computer device inputs the three-view point cloud feature 404 into the pre-trained generative model to obtain a two-dimensional semantic image 406; then, the computer device inputs the two-dimensional semantic image 406 into the two-dimensional neural network B, encodes the two-dimensional semantic image 406 through the two-dimensional neural network B and decodes the three-view image feature 408.

[0087] In an exemplary embodiment, obtaining original point cloud data in a target space scene includes: sampling point cloud data in the target space scene according to preset sampling configuration information to obtain sampled data; and obtaining the sampled data as original point cloud data.

[0088] The sampling configuration information may refer to configuration information for setting sampling parameters.

[0089] The sampling parameters may include sampling density, sampling method, etc. The sampling method includes uniform sampling, random sampling, importance sampling, etc.

[0090] In a specific implementation, when the computer device is acquiring the original point cloud data in the target space scene, the computer device may acquire the sampling configuration information pre-configured by the user; then, the computer device may sample the point cloud data in the target space scene according to the sampling parameters such as the sampling method and sampling density indicated by the sampling configuration information to acquire the sampled data. The computer device uses the acquired sampled data as the original point cloud data.

[0091] The technical solution of this embodiment, in the process of acquiring the original point cloud data in the target space scene, samples the point cloud data in the target space scene according to preset sampling configuration information, and uses the sampled data as the original point cloud data, thereby effectively reducing the data processing amount of the computer equipment while avoiding the impact on the accuracy of the spatial semantic occupancy prediction as much as possible, thereby improving the efficiency of the spatial semantic occupancy prediction.

[0092] In an exemplary embodiment, Figure 5 As shown, a spatial semantic occupancy prediction system is provided, including: a point cloud sampling unit, a three-view conversion unit, a three-view feature encoding unit, a query point feature encoding unit, and a three-dimensional semantic occupancy prediction output unit;

[0093] The point cloud sampling unit is used to sample the point cloud data in the target space scene to obtain the original point cloud data in the target space scene;

[0094] A three-view conversion unit is used to perform three-view conversion on the original point cloud data in the target space scene to obtain a three-view feature plane;

[0095] A three-view feature encoding unit is used to extract features from the three-view feature plane to obtain three-view point cloud features; input the three-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, extract features from the two-dimensional semantic image to obtain three-view image features; and fuse the three-view point cloud features and the three-view image features to obtain three-view fusion features;

[0096] The query point feature encoding unit is used to obtain the corresponding fusion features under each viewing angle in the three-view fusion features corresponding to the query point; and fuse the corresponding fusion features under each viewing angle to obtain the three-dimensional features corresponding to the query point;

[0097] The three-dimensional semantic occupancy prediction output unit is used to input the three-dimensional features of the query point into the pre-trained three-dimensional feature classification model and output the prediction result.

[0098] It should be noted that the specific limitations of the above steps can refer to the specific limitations of a spatial semantic occupancy prediction method mentioned above.

[0099] The above-mentioned spatial semantic occupancy prediction system obtains the original point cloud data in the target spatial scene, and converts the original point cloud data into a multi-view feature plane, extracts features from the multi-view feature plane, and obtains multi-view point cloud features; obtains a two-dimensional semantic image by inputting the multi-view point cloud features into a pre-trained visual generative model, extracts features from the two-dimensional semantic image, and obtains multi-view image features; then fuses the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; for any query point in the target spatial scene, outputs the spatial semantic occupancy prediction result according to the multi-view fusion features corresponding to the query point; in this way, the use of multiple The two-dimensional point cloud features of the perspective and the two-dimensional semantic image output by the visual generative model are then encoded and decoded to obtain the two-dimensional image features to reduce the information loss in the projection process. This can further enhance the model's ability to model complex three-dimensional scenes while maintaining high efficiency. At the same time, multi-perspective point cloud features and multi-perspective image features are fused in parallel, and the features of the query points on the three-perspective planes are aggregated by addition. This can effectively solve the problem that the two-dimensional feature method is difficult to depict the fine structure in complex three-dimensional scenes, effectively realize the use of two-dimensional features to model complex three-dimensional scenes, reduce the computational and storage burden of the model, and thus improve the efficiency of spatial semantic occupancy prediction based on the model.

[0100] In another embodiment, Figure 6 As shown, a spatial semantic occupancy prediction method is provided, which is applied to Figure 1 The terminal in is used as an example to illustrate, including the following steps:

[0101] Step S602: sampling to obtain the original input point cloud. Specifically, sampling is performed on the point cloud data in the target space scene to obtain the original point cloud data in the target space scene.

[0102] Step S604, three-view conversion. Specifically, the original point cloud data is divided into voxels in a cylindrical coordinate system to obtain voxelized point cloud data; the voxelized point cloud data is pooled under three coordinate axes of the cylindrical coordinate system to obtain plane features under three view angles in the cylindrical coordinate system.

[0103] Step S606, three-view feature encoding. Specifically, extract features from the three-view feature plane to obtain three-view point cloud features; input the three-view point cloud features into the pre-trained visual generative model to obtain a two-dimensional semantic image, extract features from the two-dimensional semantic image to obtain three-view image features; fuse the three-view point cloud features and the three-view image features to obtain three-view fusion features.

[0104] Step S608, query point feature encoding. Specifically, obtain the corresponding fusion features of the three-view fusion features corresponding to the query point under each view; fuse the corresponding fusion features under each view to obtain the three-dimensional features corresponding to the query point. Input the three-dimensional features of the query point into the pre-trained three-dimensional feature classification model, and output the prediction result.

[0105] It should be noted that the specific limitations of the above steps can be found in the specific limitations of a spatial semantic occupancy prediction method mentioned above, which will not be repeated here.

[0106] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0107] Based on the same inventive concept, the embodiment of the present application also provides a spatial semantic occupancy prediction device for implementing the spatial semantic occupancy prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more spatial semantic occupancy prediction device embodiments provided below can refer to the limitations of the spatial semantic occupancy prediction method above, and will not be repeated here.

[0108] In an exemplary embodiment, Figure 7 As shown, a spatial semantic occupancy prediction device is provided, comprising:

[0109] An acquisition module 710 is used to acquire original point cloud data in a target space scene;

[0110] A conversion module 720, configured to convert the original point cloud data into a multi-view feature plane, perform feature extraction on the multi-view feature plane, and obtain multi-view point cloud features;

[0111] A generation module 730 is used to input the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and perform feature extraction on the two-dimensional semantic image to obtain multi-view image features;

[0112] A fusion module 740 is used to fuse the multi-view point cloud features and the multi-view image features to obtain a multi-view fusion feature;

[0113] The output module 750 is used to output a spatial semantic occupancy prediction result for any query point located in the target space scene according to the multi-view fusion feature corresponding to the query point.

[0114] In one embodiment, the output module 750 is specifically used to obtain the corresponding fusion features of the multi-view fusion features corresponding to the query point under each perspective; fuse the corresponding fusion features under each perspective to obtain the three-dimensional features corresponding to the query point; input the three-dimensional features of the query point into a pre-trained three-dimensional feature classification model to obtain the spatial semantic occupancy prediction result.

[0115] In one embodiment, the conversion module 720 is specifically used to perform voxel division on the original point cloud data in a cylindrical coordinate system to obtain voxelized point cloud data; perform pooling operation on the voxelized point cloud data under three coordinate axes of the cylindrical coordinate system to obtain plane features under three perspectives in the cylindrical coordinate system; and obtain the plane features under the three perspectives as the multi-perspective feature plane.

[0116] In one of the embodiments, the conversion module 720 is specifically used to encode the multi-view feature plane through an encoder in the point cloud feature extraction model to obtain a feature plane encoding result; and decode the feature plane encoding result through a decoder in the point cloud feature extraction model to obtain the multi-view point cloud features at the target resolution.

[0117] In one embodiment, the generation module 730 is specifically used to encode the two-dimensional semantic image through the encoder in the image feature extraction model to obtain a semantic image encoding result; and decode the semantic image encoding result through the decoder in the image feature extraction model to obtain the multi-view image features at the target resolution.

[0118] In one of the embodiments, the acquisition module 710 is specifically used to sample the point cloud data in the target space scene according to preset sampling configuration information to obtain sampling data; the sampling data is obtained as the original point cloud data.

[0119] Each module in the above-mentioned spatial semantic occupancy prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0120] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a spatial semantic occupancy prediction method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0121] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0122] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above-mentioned spatial semantic occupancy prediction method. The steps of the spatial semantic occupancy prediction method here may be the steps of the spatial semantic occupancy prediction method in each of the above-mentioned embodiments.

[0123] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the above-mentioned spatial semantic occupancy prediction method. The steps of the spatial semantic occupancy prediction method here can be the steps of the spatial semantic occupancy prediction method in each of the above-mentioned embodiments.

[0124] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, causes the processor to perform the steps of the above-mentioned spatial semantic occupancy prediction method. The steps of the spatial semantic occupancy prediction method here may be the steps of the spatial semantic occupancy prediction method in each of the above-mentioned embodiments.

[0125] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0126] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magneto resistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0127] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0128] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A spatial semantic occupancy prediction method, characterized in that: The method comprises: Obtain the original point cloud data in the target space scene; The original point cloud data is converted into a multi-view feature plane, and features are extracted from the multi-view feature plane to obtain multi-view point cloud features; Inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and performing feature extraction on the two-dimensional semantic image to obtain multi-view image features; Fusion of the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; For any query point located in the target space scene, a spatial semantic occupancy prediction result is output according to the multi-view fusion feature corresponding to the query point.

2. The method according to claim 1, characterized in that Outputting a spatial semantic occupancy prediction result according to the multi-view fusion feature corresponding to the query point includes: Obtaining the fusion features corresponding to each perspective in the multi-perspective fusion features corresponding to the query point; Fusing the corresponding fusion features under each viewing angle to obtain the three-dimensional feature corresponding to the query point; The three-dimensional features of the query point are input into a pre-trained three-dimensional feature classification model to obtain the spatial semantic occupancy prediction result.

3. The method according to claim 1, characterized in that The converting the original point cloud data into a multi-view feature plane comprises: Dividing the original point cloud data into voxels in a cylindrical coordinate system to obtain voxelized point cloud data; A pooling operation is performed on the voxelized point cloud data under three coordinate axes of the cylindrical coordinate system to obtain plane features under three viewing angles in the cylindrical coordinate system; The plane features under the three viewing angles are obtained as the multi-view feature plane.

4. The method according to claim 1, characterized in that The extracting features from the multi-view feature plane to obtain multi-view point cloud features includes: Encoding the multi-view feature plane through an encoder in a point cloud feature extraction model to obtain a feature plane encoding result; The feature plane encoding result is decoded by a decoder in the point cloud feature extraction model to obtain the multi-view point cloud features at the target resolution.

5. The method according to claim 4, characterized in that The step of extracting features from the two-dimensional semantic image to obtain multi-view image features includes: Encoding the two-dimensional semantic image through an encoder in an image feature extraction model to obtain a semantic image encoding result; The semantic image encoding result is decoded by a decoder in the image feature extraction model to obtain the multi-view image features at the target resolution.

6. The method according to any one of claims 1 to 5, characterized in that: The step of obtaining the original point cloud data in the target space scene includes: According to preset sampling configuration information, sampling the point cloud data in the target space scene to obtain sampling data; The sampling data is obtained as the original point cloud data.

7. A spatial semantic occupancy prediction device, characterized in that: The device comprises: An acquisition module is used to acquire the original point cloud data in the target space scene; A conversion module, used for converting the original point cloud data into a multi-view feature plane, performing feature extraction on the multi-view feature plane, and obtaining multi-view point cloud features; A generation module, used for inputting the multi-view point cloud features into a pre-trained visual generative model to obtain a two-dimensional semantic image, and performing feature extraction on the two-dimensional semantic image to obtain multi-view image features; A fusion module, used for fusing the multi-view point cloud features and the multi-view image features to obtain multi-view fusion features; The output module is used to output a spatial semantic occupancy prediction result for any query point located in the target space scene according to the multi-view fusion feature corresponding to the query point.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images