Data compression method, device, electronic equipment, medium and program product

Abstract

The present disclosure provides a compression method and device for real-life business hall data, electronic equipment, medium and computer program product. The above-mentioned method and device can be used in the field of artificial intelligence technology. The compression method for real-life business hall data comprises: performing feature transformation on m point cloud data of a collected real-life business hall through a random sampling module to obtain local features of each point cloud data, wherein m is an integer greater than or equal to 1; performing feature enhancement on each local feature through a self-attention mechanism module and then aggregating to obtain global features of each point cloud data; compressing each global feature to obtain a compressed code stream of each point cloud data; and performing reverse decoding on each compressed code stream to obtain m reconstructed point cloud data of the real-life business hall.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, electronic device, medium, and computer program product for compressing real-world business hall data. Background Technology

[0002] With the development of metaverse and 5G technology, interactive multimedia technologies such as virtual reality, augmented reality, and mixed reality have received widespread attention, leading to the emergence of various immersive online virtual banking halls. Interactive media typically uses three-dimensional scenes as objects and three-dimensional models as data representation. Three-dimensional point cloud data has become the most representative type of three-dimensional data due to its numerous advantages, including non-contact acquisition, high precision, high efficiency, and high expressiveness. Specifically, LiDAR scanners can easily acquire online virtual bank halls based on three-dimensional point clouds, recreating real scenes 1:1 and providing customers with a more realistic digital banking experience through a six-degrees-of-freedom perspective. However, while the massive amount of point cloud data brings users a detailed and realistic immersive experience, it also poses significant challenges to data storage and transmission. Efficiently compressing point cloud data while ensuring visual quality for the human eye is a crucial research topic. Summary of the Invention

[0003] In view of this, the present disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for compressing real-world business hall data with high point cloud data transmission efficiency and high compression ratio distortion performance.

[0004] One aspect of this disclosure provides a method for compressing real-world business hall data, comprising: performing feature transformation on m point cloud data collected from a real-world business hall through a random sampling module to obtain local features of each point cloud data, wherein m is an integer greater than or equal to 1; aggregating each local feature after feature enhancement through a self-attention mechanism module to obtain global features of each point cloud data; compressing each global feature to obtain a compressed bitstream of each point cloud data; and performing reverse decoding on each compressed bitstream to obtain m reconstructed point cloud data of the real-world business hall.

[0005] According to the data compression method for real-world business halls disclosed herein, a random sampling module performs feature transformation on m point cloud data points of a real-world business hall, thereby initially reducing the amount of point cloud data and lowering the computational complexity of the network. The random sampling module fully preserves beneficial local features during the sampling optimization process, and these local features further learn high-dimensional features of beneficial information in subsequent network training, thus mitigating the impact of outliers. A self-attention mechanism module enhances and aggregates the transformed local features, improving the processing of massive point cloud data. The self-attention mechanism can quickly learn the correlation and differences between hidden layer features of the point cloud, aggregating important features to make the reconstructed point cloud data more realistic, thereby providing customers with a more immersive experience. The data compression method for real-world business halls disclosed herein can achieve end-to-end compression encoding of massive real-world business hall point cloud data, effectively improving the low storage and transmission efficiency of point cloud data and significantly enhancing the compression rate and distortion performance of point cloud data.

[0006] In some embodiments, each point cloud data of the real-world business hall includes location information and color information. The step of performing feature transformation on the m point cloud data of the collected real-world business hall through a random sampling module to obtain local features of each point cloud data includes: performing max pooling on each point cloud data of the real-world business hall to obtain a first local feature; using the color information of each point cloud data of the real-world business hall as a second local feature; and fusing the first local feature and the second local feature of each point cloud data to obtain local features of each point cloud data.

[0007] In some embodiments, performing max pooling on each point cloud data of the real-world business hall to obtain a first local feature includes: finding K nearest neighbor points for each point cloud data of the real-world business hall through K-nearest neighbor search, where K is an integer greater than or equal to 1; performing operations on each point cloud data and the K nearest neighbor points through a multilayer perceptron to obtain adaptive weights; multiplying the transpose matrix of each point cloud data with the adaptive weights and then fusing it with the second local feature of the point cloud data to obtain a pooled value; and performing max pooling on the pooled value to obtain the first local feature.

[0008] In some embodiments, the self-attention mechanism module is an n-head self-attention mechanism module, where n is an integer greater than or equal to 1. The step of aggregating each local feature after feature enhancement by the self-attention mechanism module to obtain the global feature of each point cloud data includes: aggregating each local feature after feature enhancement in each of the n-head self-attention mechanism modules to obtain n sub-global features of each point cloud data, wherein the n-head self-attention mechanism module corresponds one-to-one with the n sub-global features; and concatenating the n sub-global features to obtain the global feature of each point cloud data.

[0009] In some embodiments, each local feature is enhanced and aggregated in each self-attention mechanism module to obtain a sub-global feature for each point cloud data, including: mapping each local feature into three linear matrices Query, Key, and Value in each self-attention mechanism module; and processing the three linear matrices using an attention mechanism to obtain the sub-global feature.

[0010] In some embodiments, the step of reverse decoding each of the compressed bitstreams to obtain m reconstructed point cloud data of the real-world business hall includes: first passing each of the compressed bitstreams through a self-attention mechanism module, and then through a random sampling module for decoding to obtain m reconstructed point cloud data of the real-world business hall.

[0011] Another aspect of this disclosure provides a compression device for real-world business hall data, comprising: a first processing module, configured to perform feature transformation on m point cloud data of the collected real-world business hall through a random sampling module to obtain local features of each point cloud data, wherein m is an integer greater than or equal to 1; a second processing module, configured to perform feature enhancement and aggregation on each local feature through a self-attention mechanism module to obtain global features of each point cloud data; a compression module, configured to compress each global feature to obtain a compressed bitstream of each point cloud data; and a decoding module, configured to perform reverse decoding on each compressed bitstream to obtain m reconstructed point cloud data of the real-world business hall.

[0012] Another aspect of this disclosure provides an electronic device including one or more processors and one or more memories, wherein the memories are used to store executable instructions that, when executed by the processor, implement the method described above.

[0013] Another aspect of this disclosure provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the method described above.

[0014] Another aspect of this disclosure provides a computer program product including a computer program comprising computer executable instructions that, when executed, implement the method described above. Attached Figure Description

[0015] The above and other objects, features and advantages of this disclosure will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0016] Figure 1 This illustration schematically shows an exemplary system architecture to which methods and apparatus can be applied according to embodiments of the present disclosure;

[0017] Figure 2 A flowchart illustrating a method for compressing real-world business hall data according to an embodiment of this disclosure is shown schematically.

[0018] Figure 3 This schematically illustrates a flowchart of a process, according to an embodiment of the present disclosure, of performing feature transformation on m point cloud data of a real-world business hall through a random sampling module to obtain local features of each point cloud data.

[0019] Figure 4 This schematically illustrates a flowchart of performing maximum pooling on each point cloud data of a real-world business hall according to an embodiment of the present disclosure to obtain a first local feature;

[0020] Figure 5 This schematically illustrates a flowchart of a process that aggregates each local feature after feature enhancement by a self-attention mechanism module, according to an embodiment of the present disclosure, to obtain the global features of each point cloud data.

[0021] Figure 6 This schematically illustrates a flowchart of a process in which each local feature is enhanced and aggregated in each head self-attention mechanism module according to an embodiment of the present disclosure to obtain a sub-global feature for each point cloud data.

[0022] Figure 7 A block diagram schematically illustrates a device for compressing real-world business hall data according to an embodiment of the present disclosure;

[0023] Figure 8 A block diagram of an electronic device according to an embodiment of the present disclosure is shown schematically. Detailed Implementation

[0024] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0025] In the technical solution disclosed herein, the acquisition, storage, and application of user personal information all comply with relevant laws and regulations, necessary confidentiality measures have been taken, and there is no violation of public order and good morals. In the technical solution disclosed herein, the acquisition, collection, storage, use, processing, transmission, provision, disclosure, and application of data all comply with relevant laws and regulations, necessary confidentiality measures have been taken, and there is no violation of public order and good morals.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0027] When using expressions such as "at least one of A, B, or C," it should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (e.g., "a system having at least one of A, B, or C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.). The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, features defined with "first" or "second" may explicitly or implicitly include one or more of the stated features.

[0028] With the development of metaverse and 5G technology, interactive multimedia technologies such as virtual reality, augmented reality, and mixed reality have received widespread attention, leading to the emergence of various immersive online virtual banking halls. Interactive media typically uses three-dimensional scenes as objects and three-dimensional models as data representation. Three-dimensional point cloud data has become the most representative type of three-dimensional data due to its numerous advantages, including non-contact acquisition, high precision, high efficiency, and high expressiveness. Specifically, LiDAR scanners can easily acquire online virtual bank halls based on three-dimensional point clouds, recreating real scenes 1:1 and providing customers with a more realistic digital banking experience through a six-degrees-of-freedom perspective. However, while the massive amount of point cloud data brings users a detailed and realistic immersive experience, it also poses significant challenges to data storage and transmission. Efficiently compressing point cloud data while ensuring visual quality for the human eye is a crucial research topic.

[0029] This disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for compressing real-world business hall data. The method for compressing real-world business hall data includes: performing feature transformation on m point cloud data points collected from the real-world business hall using a random sampling module to obtain local features for each point cloud data point, where m is an integer greater than or equal to 1; aggregating each local feature after feature enhancement using a self-attention mechanism module to obtain global features for each point cloud data point; compressing each global feature to obtain a compressed bitstream for each point cloud data point; and performing reverse decoding on each compressed bitstream to obtain m reconstructed point cloud data points of the real-world business hall.

[0030] It should be noted that the compression method, apparatus, electronic device, computer-readable storage medium and computer program product of the real-world business hall data disclosed herein can be used in the field of artificial intelligence technology, or in any field other than the field of artificial intelligence technology, such as the financial field. The field of this disclosure is not limited here.

[0031] Figure 1 This illustration schematically depicts an exemplary system architecture 100 for compressing real-world business hall data according to embodiments of this disclosure, including methods, apparatus, electronic devices, computer-readable storage media, and computer program products. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this disclosure, in order to help those skilled in the art understand the technical content of this disclosure, but do not mean that the embodiments of this disclosure cannot be used in other devices, systems, environments or scenarios.

[0032] like Figure 1As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0033] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0034] Terminal devices 101, 102, and 103 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0035] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using terminal devices 101, 102, and 103 (for example only). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0036] It should be noted that the data compression method for real-time business halls provided in this embodiment can generally be executed by server 105. Correspondingly, the data compression device for real-time business halls provided in this embodiment can generally be located in server 105. The data compression method for real-time business halls provided in this embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105. Correspondingly, the data compression device for real-time business halls provided in this embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105.

[0037] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0038] The following will be based on Figure 1 The described scene, through Figures 2-6 The method for compressing real-world business hall data according to embodiments of this disclosure will be described in detail.

[0039] Figure 2 A flowchart illustrating a method for compressing real-world business hall data according to an embodiment of this disclosure is shown schematically.

[0040] like Figure 2 As shown, the data compression method for real-world business halls in this embodiment includes operations S210 to S240.

[0041] In operation S210, the collected m point cloud data points of the real-world business hall are processed by a random sampling module to perform feature transformation, obtaining the local features of each point cloud data point, where m is an integer greater than or equal to 1. For example, the m point cloud data points can be represented as P (m，6) ={p1, p2, ..., p i , ..., p m}

[0042] As one possible implementation method, each point cloud data of the real-world business hall includes location information and color information. For example, the i-th point cloud data p i It can be represented as p i (x, y, z, R, G, B), where x, y, and z represent the 3D position information of the i-th point cloud, and R, G, and B represent the color information of the i-th point cloud, which can be understood as the proportion information of the three primary colors. For example... Figure 3 As shown, operation S210 transforms the m point cloud data collected from the real-world business hall through a random sampling module to obtain the local features of each point cloud data, including operations S211 to S213.

[0043] In operation S211, a maximum pooling operation is performed on each point cloud data of the real-world business hall to obtain the first local feature.

[0044] As an feasible approach, such as Figure 4 As shown, operation S211 performs maximum pooling on each point cloud data of the real-world business hall to obtain the first local feature, including operations S2111 to S2114.

[0045] In operation S2111, K nearest neighbor search is used to find K nearest neighbor points for each point cloud data of the real-world business hall, where K is an integer greater than or equal to 1. Here, the K nearest neighbor points of the i-th point cloud data can be represented as...

[0046] In operation S2112, an adaptive weight is obtained by processing each point cloud data point and its K nearest neighbors using a multilayer perceptron. The adaptive weight can be used... express, It can be obtained using formula (1).

[0047]

[0048] Where ||·|| is the Euclidean distance between nearest neighbors. This indicates feature fusion.

[0049] In operation S2113, the transpose matrix of each point cloud data is multiplied by the adaptive weight and then fused with the second local feature of the point cloud data to obtain the pooling value.

[0050] In operation S2114, max pooling is performed on the pooled values ​​to obtain the first local feature. The first local feature can be used... express, It can be obtained using formula (2).

[0051]

[0052] in, Indicates the second local feature. This represents the pooling value.

[0053] By operating S2111 to S2114, it is easy to perform maximum pooling operation on each point cloud data of the real-world business hall to obtain the first local feature.

[0054] In operation S212, the color information of each point cloud data of the real-world business hall is used as the second local feature.

[0055] In operation S213, the first and second local features of each point cloud data are fused to obtain the local features of each point cloud data. The local features of each point cloud data can be used... express, It can be obtained using formula (3).

[0056]

[0057] By operating S211 to S213, it is easy to perform feature transformation on the m point cloud data of the collected real-world business hall through the random sampling module to obtain the local features of each point cloud data.

[0058] In operation S220, each local feature is enhanced by the self-attention mechanism module and then aggregated to obtain the global features of each point cloud data.

[0059] As an implementable approach, the self-attention mechanism module is an n-head self-attention mechanism module, where n is an integer greater than or equal to 1. For example... Figure 5As shown, operation S220 aggregates each local feature after feature enhancement through the self-attention mechanism module to obtain the global features of each point cloud data, including operations S221 and S222.

[0060] In operation S221, each local feature is enhanced and aggregated in each self-attention mechanism module of the n-head self-attention mechanism module to obtain n sub-global features for each point cloud data. The n-head self-attention mechanism module corresponds one-to-one with the n sub-global features.

[0061] As a feasible approach, such as Figure 6 As shown, operation S221 performs feature enhancement on each local feature in each head self-attention mechanism module and then aggregates them to obtain a sub-global feature for each point cloud data, including operation S2211 and operation S2212.

[0062] In operation S2211, each local feature is mapped to three linear matrices Query, Key, and Value in each head self-attention mechanism module. Query, Key, and Value can be obtained using formula (4).

[0063]

[0064] Among them, W q W k W V These are the weight matrices, yes The matrix representation of .

[0065] In operation S2212, the three linear matrices are processed using an attention mechanism to obtain sub-global features. Specifically, the hidden layer features of the i-th unit in the (l+1)-th layer can be determined according to formula (5). Hidden features It can be used as a sub-global feature.

[0066]

[0067] Among them, Q 1 Represents the Query at level 1, K 1 V represents the key of level 1. 1 This represents the Value at level 1. Represents the hidden layer features of the i-th unit in the i-th layer.

[0068] Therefore, by operating on S2211 and S2212, it is convenient to perform feature enhancement on each local feature and then aggregate it in each head self-attention mechanism module to obtain a sub-global feature for each point cloud data. Here, the sub-global feature for each point cloud data obtained by aggregating each local feature after feature enhancement in the i-th head self-attention mechanism module can be represented as head. i .

[0069] In operation S222, the n sub-global features are concatenated to obtain the global features of each point cloud data. The global features of each point cloud data can be represented by H, which can be obtained by formula (6).

[0070] H = Concat(head1, ..., head) i , ..., head n (6)

[0071] By operating on S221 and S222, it is easy to aggregate the features of each local feature after feature enhancement by the self-attention mechanism module, thus obtaining the global features of each point cloud data. Parallel execution of the multi-head attention mechanism ensures the stability of the learning process.

[0072] In operation S230, each global feature is compressed to obtain a compressed bitstream of each point cloud data.

[0073] By operating the S240, each compressed bitstream is decoded in reverse to obtain m reconstructed point cloud data of the real-world business hall.

[0074] As one possible approach, operation S240 performs reverse decoding on each compressed bitstream to obtain m reconstructed point cloud data of the real-world business hall, including operation S241.

[0075] In operation S241, each compressed bitstream first passes through the self-attention mechanism module, and then through the random sampling module for decoding, to obtain m reconstructed point cloud data of the real-world business hall. Operation S241 also facilitates reverse decoding of each compressed bitstream to obtain the m reconstructed point cloud data of the real-world business hall.

[0076] According to the data compression method for real-world business halls disclosed herein, a random sampling module performs feature transformation on m point cloud data points of a real-world business hall, thereby initially reducing the amount of point cloud data and lowering the computational complexity of the network. The random sampling module fully preserves beneficial local features during the sampling optimization process, and these local features further learn high-dimensional features of beneficial information in subsequent network training, thus mitigating the impact of outliers. A self-attention mechanism module enhances and aggregates the transformed local features, improving the processing of massive point cloud data. The self-attention mechanism can quickly learn the correlation and differences between hidden layer features of the point cloud, aggregating important features to make the reconstructed point cloud data more realistic, thereby providing customers with a more immersive experience. The data compression method for real-world business halls disclosed herein can achieve end-to-end compression encoding of massive real-world business hall point cloud data, effectively improving the low storage and transmission efficiency of point cloud data and significantly enhancing the compression rate and distortion performance of point cloud data.

[0077] Based on the above-mentioned method for compressing real-world service hall data, this disclosure also provides a device for compressing real-world service hall data. The following will be combined with... Figure 7 The compression device 10 for real-world business hall data is described in detail.

[0078] Figure 7 The diagram illustrates the structure of a data compression device 10 for a real-world business hall according to an embodiment of the present disclosure.

[0079] The compression device 10 for real-time business hall data includes a first processing module 1, a second processing module 2, a compression module 3, and a decoding module 4.

[0080] The first processing module 1 is used to perform operation S210: the m point cloud data of the collected real-world business hall are processed by the random sampling module to perform feature transformation to obtain the local features of each point cloud data, where m is an integer greater than or equal to 1.

[0081] The second processing module 2 is used to perform operation S220: after each local feature is enhanced by the self-attention mechanism module, it is aggregated to obtain the global features of each point cloud data.

[0082] Compression module 3 is used to perform operation S230: compress each global feature to obtain a compressed bitstream of each point cloud data.

[0083] Decoding module 4 is used to perform operation S240: reverse decoding of each compressed bitstream to obtain m reconstructed point cloud data of the real-world business hall.

[0084] According to some embodiments of this disclosure, each point cloud data of the real-world business hall includes location information and color information, and the first processing module may include a first determining unit, a second determining unit, and a fusion unit.

[0085] The first determining unit is used to perform maximum pooling operation on each point cloud data of the real-world business hall to obtain the first local feature.

[0086] The second determining unit is used to use the color information of each point cloud data of the real-world business hall as the second local feature.

[0087] The fusion unit is used to fuse the first local features and the second local features of each point cloud data to obtain the local features of each point cloud data.

[0088] According to some embodiments of this disclosure, the first determining unit may include a first operating element, a second operating element, a third operating element, and a fourth operating element.

[0089] The first operating element is used to find K nearest neighbor points for each point cloud data of the real-world business hall through K-nearest neighbor search, where K is an integer greater than or equal to 1.

[0090] The second operating element is used to operate on each point cloud data and K nearest neighbor points through a multilayer perceptron to obtain adaptive weights.

[0091] The third operational element is used to multiply the transpose matrix of each point cloud data with the adaptive weight and then fuse it with the second local feature of the point cloud data to obtain the pooling value.

[0092] The fourth operating element is used to perform maximum pooling on the pooled value to obtain the first local feature.

[0093] According to some embodiments of this disclosure, the self-attention mechanism module is an n-head self-attention mechanism module, where n is an integer greater than or equal to 1, and the second processing module may include a third determining unit and a splicing unit.

[0094] The third determining unit is used to perform feature enhancement on each local feature in each of the n-head self-attention mechanism modules and then aggregate them to obtain n sub-global features for each point cloud data. The n-head self-attention mechanism modules correspond to the n sub-global features.

[0095] The splicing unit is used to splice together n sub-global features to obtain the global features of each point cloud data.

[0096] According to some embodiments of this disclosure, the third determining unit may include a mapping element and a processing element.

[0097] The mapping element is used to map each local feature into three linear matrices, Query, Key, and Value, in each head self-attention mechanism module.

[0098] The processing element is used to process three linear matrices using an attention mechanism to obtain sub-global features.

[0099] According to some embodiments of this disclosure, the decoding module may include a fourth determining unit.

[0100] The fourth determining unit is used to decode each compressed bitstream first through the self-attention mechanism module and then through the random sampling module to obtain m reconstructed point cloud data of the real-scene business hall.

[0101] According to the real-world business hall data compression device 10 of this disclosure, the feature transformation of m point cloud data of the real-world business hall is performed by a random sampling module, which can initially reduce the amount of point cloud data and reduce the computational complexity of the network. The random sampling module fully retains beneficial local features during the sampling optimization process. The local features are further learned in the subsequent network training to learn the high-dimensional features of beneficial information, thereby mitigating the impact of outliers. The self-attention mechanism module performs feature enhancement and aggregation on the local features after feature transformation, which can improve the processing of massive point cloud data. The self-attention mechanism can quickly learn the correlation and differences between hidden layer features of the point cloud and aggregate important features, making the reconstructed point cloud data more realistic. This can provide customers with a more realistic immersive experience. The real-world business hall data compression method of this disclosure can realize end-to-end compression encoding of massive real-world business hall point cloud data, which can effectively improve the problem of low storage and transmission efficiency of point cloud data and effectively improve the compression rate distortion performance of point cloud data.

[0102] Furthermore, according to embodiments of this disclosure, any multiple modules among the first processing module 1, the second processing module 2, the compression module 3, and the decoding module 4 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module.

[0103] According to embodiments of this disclosure, at least one of the first processing module 1, the second processing module 2, the compression module 3, and the decoding module 4 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in any one of the three methods of software, hardware, and firmware, or in a suitable combination of any of these.

[0104] Alternatively, at least one of the first processing module 1, the second processing module 2, the compression module 3, and the decoding module 4 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0105] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing the above-described method according to an embodiment of the present disclosure.

[0106] like Figure 8 As shown, an electronic device 900 according to an embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage portion 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0107] RAM 903 stores various programs and data required for the operation of electronic device 900. Processor 901, ROM 902, and RAM 903 are interconnected via bus 904. Processor 901 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 902 and / or RAM 903. It should be noted that the programs may also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0108] According to embodiments of this disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to a bus 904. The electronic device 900 may also include one or more of the following components connected to the I / O interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 910 as needed so that computer programs read from it can be installed into the storage section 908 as needed.

[0109] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0110] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 902 and / or RAM 903 and / or one or more memories other than ROM 902 and RAM 903 described above.

[0111] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods of the embodiments of this disclosure.

[0112] When the computer program is executed by the processor 901, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0113] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via the communication section 909, and / or installed from a removable medium 911. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0114] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, it performs the functions defined in the system of this disclosure embodiment. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0115] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0116] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0117] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0118] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this disclosure is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A method for compressing real-world business hall data, characterized in that, include: The m point cloud data collected from the real-world business hall are processed by a random sampling module to perform feature transformation, thereby obtaining the local features of each point cloud data, where m is an integer greater than or equal to 1. Each local feature is enhanced by a self-attention mechanism module and then aggregated to obtain the global features of each point cloud data. Each of the global features is compressed to obtain a compressed bitstream of each point cloud data; and Each of the compressed bitstreams is decoded in reverse to obtain m reconstructed point cloud data of the real-world business hall. Each point cloud data point in the real-world business hall includes location information and color information. The m point cloud data points collected from the real-world business hall are subjected to feature transformation through a random sampling module to obtain local features of each point cloud data point, including: Perform maximum pooling on each point cloud data of the real-world business hall to obtain the first local feature; The color information of each point cloud data point in the real-world business hall is used as a second local feature; and The first local feature and the second local feature of each point cloud data are fused to obtain the local features of each point cloud data.

2. The method according to claim 1, characterized in that, The step of performing maximum pooling on each point cloud data point of the real-world business hall to obtain the first local feature includes: K-nearest neighbor search is used to find K nearest neighbor points for each point cloud data of the real-world business hall, where K is an integer greater than or equal to 1; Adaptive weights are obtained by operating on each point cloud data and the K nearest neighbors using a multilayer perceptron. The transpose matrix of each point cloud data point is multiplied by the adaptive weight and then fused with the second local feature of that point cloud data to obtain a pooling value; and The pooled value is subjected to max pooling to obtain the first local feature.

3. The method according to claim 1, characterized in that, The self-attention mechanism module is an n-head self-attention mechanism module, where n is an integer greater than or equal to 1. The process of aggregating each local feature after feature enhancement by the self-attention mechanism module to obtain the global features of each point cloud data includes: In each of the n-head self-attention mechanism modules, each local feature is enhanced and then aggregated to obtain n sub-global features for each point cloud data, wherein each n-head self-attention mechanism module corresponds one-to-one with the n sub-global features; and The n sub-global features are concatenated to obtain the global features of each point cloud data.

4. The method according to claim 3, characterized in that, In each self-attention mechanism module, each local feature is enhanced and then aggregated to obtain a sub-global feature for each point cloud data, including: In each self-attention mechanism module, each local feature is mapped to three linear matrices: Query, Key, and Value; and The three linear matrices are processed using an attention mechanism to obtain sub-global features.

5. The method according to claim 1, characterized in that, The reverse decoding of each compressed bitstream yields m reconstructed point cloud data points of the real-world business hall, including: Each compressed bitstream is first processed by a self-attention mechanism module and then by a random sampling module to obtain m reconstructed point cloud data of the real-world business hall.

6. A data compression device for a real-scene business hall, characterized in that, include: The first processing module is used to perform feature transformation on m point cloud data collected from the real-world business hall through a random sampling module to obtain local features of each point cloud data, where m is an integer greater than or equal to 1. The second processing module is used to perform feature enhancement and aggregation on each of the local features through the self-attention mechanism module to obtain the global features of each point cloud data. A compression module, configured to compress each of the global features to obtain a compressed bitstream of each point cloud data; and The decoding module is used to perform reverse decoding on each of the compressed bitstreams to obtain m reconstructed point cloud data of the real-world business hall. Each point cloud data point in the real-world business hall includes location information and color information. The m point cloud data points collected from the real-world business hall are subjected to feature transformation through a random sampling module to obtain local features of each point cloud data point, including: Perform maximum pooling on each point cloud data of the real-world business hall to obtain the first local feature; The color information of each point cloud data point in the real-world business hall is used as a second local feature; and The first local feature and the second local feature of each point cloud data are fused to obtain the local features of each point cloud data.

7. An electronic device, characterized in that, include: One or more processors; One or more memories are provided for storing executable instructions that, when executed by the processor, implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores executable instructions that, when executed by a processor, implement the method according to any one of claims 1 to 5.

9. A computer program product, characterized in that, The method includes a computer program comprising one or more executable instructions that, when executed by a processor, implement the method according to any one of claims 1 to 5.

Images