Radar data encoding and decoding method, encoding device, and decoding device

Abstract

A radar data encoding method is provided, the method including the steps of: acquiring data to be encoded, the data to be encoded including first, second, and third attribute data of different modalities; encoding the first attribute data to obtain first encoded data; predicting the second and third attribute data based on the first attribute data to obtain predicted values ​​of the second and third attribute data; encoding the second attribute data based on the predicted value of the second attribute data to obtain second encoded data; and encoding the third attribute data based on the predicted value of the third attribute data to obtain third encoded data. According to the radar data encoding method provided in the present application, the compression efficiency of radar data is improved by cross-modal prediction.

Description

[Technical Field]

[0001] This application claims priority to Chinese Patent Application No. 202211362992.3, entitled "Radar Data Encoding and Decoding Method, Encoding Device, and Decoding Device," filed on November 2, 2022, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to the field of data compression technology, and in particular to a method, an apparatus, and a device for encoding and decoding radar data. [Background technology]

[0003] Point cloud data acquired by LiDAR plays an important role in autonomous driving algorithm modules (positioning, obstacle avoidance, etc.). In recent years, with the evolution of LiDAR technology, the durability of LiDAR has improved significantly, and the unit price has fallen to a level acceptable to the market. Various types of automotive-grade LiDAR are emerging one after another. LiDAR has now become the standard sensor for Level 4 autonomous driving and is adopted in many mass-produced vehicles.

[0004] At the same time, LIDAR data transmission is also a major challenge. With the pursuit of higher-resolution point clouds, the number of LIDAR lines is increasing. The transmission of raw data from a single LIDAR may require a bandwidth of over 100 Mbps. For security and reliability reasons, autonomous vehicles are usually equipped with multiple sensors, such as cameras and millimeter-wave radar. This puts greater pressure on data transmission and storage. To meet the data transmission and storage requirements of LIDAR, an efficient compression algorithm is essential. Summary of the Invention [Means for solving the problem]

[0005] SUMMARY OF THE INVENTION The embodiments of the present application provide a radar data encoding and decoding method, an encoding device, and a decoding device to improve the compression efficiency of radar data through cross-modal prediction.

[0006] According to a first aspect, the present application provides a radar data encoding method including the steps of: acquiring data to be encoded, the data being raw radar data or point cloud data; and, if the raw data is raw data, the data to be encoded includes first, second, and third attribute data of different modalities; encoding the first attribute data to obtain first encoded data; predicting the second and third attribute data based on the first attribute data to obtain predicted values ​​of the second and third attribute data; encoding the second attribute data based on the predicted value of the second attribute data to obtain second encoded data; and encoding the third attribute data based on the predicted value of the third attribute data to obtain third encoded data. The radar data encoding method provided in the present application improves radar data compression efficiency through cross-modal prediction.

[0007] According to the radar data encoding method provided in the present application, in order to improve the compression efficiency of radar data, correlations between attribute data of different modalities of radar data are used, and the attribute data of each modal is encoded via cross-modal prediction.

[0008] In a possible implementation form, the step of predicting the second attribute data and the third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data includes a step of setting an attribute value corresponding to a first moment of the first attribute data to 0, and setting both predicted values ​​corresponding to the first moments of the second attribute data and the third attribute data to 0, wherein the attribute data with a predicted value of 0 is configured to skip encoding.

[0009] For example, if the first attribute data, the second attribute data, and the third attribute data are signal strength, reflectivity, and distance information, respectively, and one piece of information is 0, such as distance, the signal strength and reflectivity corresponding to that distance are usually both 0. This is because a distance of 0 indicates that the signal is very weak and no useful information can be obtained. Therefore, the distance data is ignored, and correspondingly, both the signal strength and reflectivity corresponding to the distance data are indicated as 0. As another example, a signal strength of 0 indicates that there is no signal or that the signal is very weak and no useful information can be obtained. Therefore, both the reflectivity corresponding to the signal strength and the distance information are indicated as 0. Therefore, in order to improve the compression rate of radar data, it may be possible to perform cross-modal coding on these three pieces of information.

[0010] The encoder can skip data with an attribute value of 0 without encoding the data, and the position of the attribute data with a value of 0 in the data stream is not reserved. In this way, when the encoder encodes the multimodal attribute data, it can consider that no attribute data with a value of 0 exists. This can significantly reduce the amount of data. More 0 values ​​indicate higher compression efficiency, further saving bandwidth during data transmission.

[0011] In another possible implementation form, the step of predicting the second attribute data and the third attribute data based on the first attribute data to obtain predicted values ​​of the second attribute data and the third attribute data includes the steps of: obtaining a variation characteristic of the first attribute data; determining a predicted value of the second attribute data in the current frame based on the variation characteristic of the first attribute data and an attribute value of the second attribute data in the encoded frame; and determining a predicted value of the third attribute data in the current frame based on the variation characteristic of the first attribute data and an attribute value of the third attribute data in the encoded frame, wherein the variation characteristic indicates the variation characteristic of the first attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame.

[0012] For example, when the first attribute data is signal strength and the second attribute is distance, for the distance data that needs to be encoded, if the distance data at moment t is encoded, the optical flow I is first calculated by using the encoded signal strength data at moments t and t-1, and then the distance I at moment t is calculated by using the optical flow I at moment t and the encoded distance at moment t-1.

number

[0013] In another possible implementation form, a specific implementation form for obtaining the change characteristics of the first attribute data is to determine the change characteristics of the first attribute data based on the attribute values ​​of the attribute values ​​of the first attribute data in a frame corresponding to the encoded frame and the attribute values ​​of the first attribute data corresponding to the current frame.

[0014] Optionally, the first attribute data, the second attribute data and the third attribute data are each one of distance information, signal strength information, angle information and reflectance information.

[0015] In another possible implementation, when the data to be encoded is point cloud data, the data to be encoded includes fourth attribute data, fifth attribute data, and sixth attribute data of different modalities. The radar data encoding method provided in this embodiment of the present application further includes: separately encoding the fourth attribute data and the fifth attribute data to obtain fourth encoded data and fifth encoded data; predicting sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute; and encoding the sixth attribute based on the predicted value of the sixth attribute to obtain the sixth encoded data.

[0016] In this possible implementation, cross-modal prediction is implemented by using correlations between attribute data of different modalities of the lidar to improve prediction accuracy and directly predict the exact values ​​of some attribute data. More accurate prediction allows for a more concentrated distribution (around the value 0) of the coded residuals. In this way, compression performance is improved in cooperation with entropy coding.

[0017] In one example, a specific implementation form of predicting sixth attribute data based on fourth attribute data and fifth attribute data to obtain a predicted value of the sixth attribute is to obtain variation characteristics of the fourth attribute data and the fifth attribute data, and determine a predicted value of the sixth attribute data in a current frame based on the variation characteristics of the fourth attribute data and the fifth attribute data and an attribute value of the sixth attribute data in the encoded frame, where the variation characteristics indicate the variation characteristics of the fourth attribute data and the fifth attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame.

[0018] In another possible implementation form, a specific implementation form for obtaining the change characteristics of the fourth attribute data and the fifth attribute data includes a step of determining the change characteristics of the attribute values ​​of the fourth attribute data and the fifth attribute data based on the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data in a frame corresponding to the encoded frame and the attribute values ​​of the fourth attribute data and the fifth attribute data in a frame corresponding to the current frame.

[0019] According to a second aspect, an embodiment of the present application further provides a radar data decoding method, when the radar data is raw data, including: acquiring first encoded data; decoding the first encoded data to acquire first attribute data; predicting second attribute data and third attribute data based on the first attribute data to acquire predicted values ​​of the second attribute data and predicted values ​​of the third attribute data; decoding the second encoded data and the third encoded data to acquire remainder values ​​of the second attribute data and residual values ​​of the third attribute data; acquiring the second attribute data based on the predicted values ​​of the second attribute data and the residual value of the second attribute data; and acquiring the third attribute data based on the predicted values ​​of the third attribute data and the residual value of the third attribute data.

[0020] In a possible implementation, the radar data decoding method further provided in this embodiment of the present application, when the radar data is point cloud data, further includes: obtaining fourth encoded data and fifth encoded data; separately decoding the fourth encoded data and the fifth encoded data to obtain fourth attribute data and fifth attribute data; predicting sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data; decoding the sixth encoded data; and obtaining the sixth attribute data based on the predicted value of the sixth attribute data and a remainder value of the sixth attribute data to obtain a remainder value of the sixth attribute data.

[0021] According to a third aspect, an embodiment of the present application provides a radar data encoding device including an acquisition module, a first encoding module, and a first prediction module. The acquisition module is configured to acquire data to be encoded, where the data to be encoded is radar raw data or point cloud data. If the data to be encoded is raw data, the data to be encoded includes first, second, and third attribute data of different modalities. The first encoding module is configured to encode the first attribute data to obtain first encoded data. The first prediction module is configured to predict the second and third attribute data based on the first attribute data to obtain predicted values ​​of the second and third attribute data. The first encoding module is further configured to encode the second attribute data based on the predicted value of the second attribute data to obtain second encoded data, and encode the third attribute data based on the predicted value of the third attribute data to obtain third encoded data.

[0022] In another possible implementation form, the first prediction module is specifically configured to set an attribute value corresponding to the first moment of the first attribute data to 0, and set both predicted values ​​corresponding to the first moments of the second attribute data and the third attribute data to 0, and to skip encoding of attribute data having a predicted value of 0.

[0023] In another possible implementation form, the first prediction module is particularly configured to obtain variation characteristics of the first attribute data, determine a predicted value of the second attribute data in the current frame based on the variation characteristics of the first attribute data and an attribute value of the second attribute data in the encoded frame, and determine a predicted value of the third attribute data in the current frame based on the variation characteristics of the first attribute data and an attribute value of the third attribute data in the encoded frame, wherein the variation characteristics indicate the variation characteristics of the first attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame.

[0024] In another possible implementation form, the step of obtaining the change characteristics of the first attribute data includes a step of determining the change characteristics of the first attribute data based on an attribute value of the first attribute data in a frame corresponding to the current frame and an attribute value of the first attribute data in a frame corresponding to the encoded frame.

[0025] In another possible implementation, the first attribute data, the second attribute data, and the third attribute data are each one of distance information, signal strength information, angle information, and reflectance information.

[0026] In another possible implementation, when the data to be encoded is point cloud data, the data to be encoded includes fourth attribute data, fifth attribute data, and sixth attribute data of different modalities. The radar data encoding device provided in this embodiment of the present application further includes a second encoding module and a second prediction module. The second encoding module is configured to separately encode the fourth attribute data and the fifth attribute data to obtain fourth encoded data and fifth encoded data. The second prediction module is configured to predict the sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute. The second encoding module is further configured to encode the sixth attribute based on the predicted value of the sixth attribute to obtain sixth encoded data.

[0027] In another possible implementation form, the prediction module is particularly configured to obtain change characteristics of the fourth attribute data and the fifth attribute data, and determine a predicted value of the sixth attribute data in the current frame based on the change characteristics of the fourth attribute data and the fifth attribute data and the attribute value of the sixth attribute data in the encoded frame, wherein the change characteristics indicate change characteristics of the fourth attribute data and the fifth attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame.

[0028] In another possible implementation form, an implementation form for obtaining the change characteristics of the fourth attribute data and the fifth attribute data includes a step of determining the change characteristics of the attribute values ​​of the fourth attribute data and the fifth attribute data based on the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data in a frame corresponding to the encoded frame and the attribute values ​​of the fourth attribute data and the fifth attribute data in a frame corresponding to the current frame.

[0029] According to a fourth aspect, an embodiment of the present application further provides a radar data decoding device including a first acquisition module, a first decoding module, and a first prediction module. The first acquisition module is configured to acquire first encoded data when the radar data is raw data. The first decoding module is configured to decode the first encoded data to acquire first attribute data. The first prediction module is configured to predict second attribute data and third attribute data based on the first attribute data to acquire predicted values ​​of the second attribute data and predicted values ​​of the third attribute data. The first decoding module is further configured to decode the second encoded data and the third encoded data to acquire a remainder value of the second attribute data and a remainder value of the third attribute data, acquire the second attribute data based on the predicted value of the second attribute data and the remainder value of the second attribute data, and acquire the third attribute data based on the predicted value of the third attribute data and the remainder value of the third attribute data.

[0030] In another possible implementation, the radar data decoding device provided in this embodiment of the present application further includes a second acquisition module, a second decoding module, and a second prediction module. When the radar data is point cloud data, the second acquisition module is configured to acquire fourth encoded data and fifth encoded data. The second decoding module is configured to separately decode the fourth encoded data and the fifth encoded data to acquire fourth attribute data and fifth attribute data. The second prediction module is configured to predict sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data. The second decoding module is further configured to decode the sixth encoded data to obtain a remainder value of the sixth attribute data and obtain the sixth attribute data based on the predicted value of the sixth attribute data and the remainder value of the sixth attribute data.

[0031] According to a fifth aspect, the present application provides a computing device including a memory and a processor, wherein the memory stores executable code and the processor executes the executable code to perform the method provided in the first aspect and / or the method provided in the second aspect of the present application.

[0032] According to a sixth aspect, the present application provides a computer-readable storage medium storing a computer program which, when executed on a computer, enables the computer to perform the method provided in the first aspect and / or the method provided in the second aspect of the present application.

[0033] According to a seventh aspect, the present application provides a computer program or computer program product comprising instructions that, when executed, perform the method provided in the first aspect and / or the method provided in the second aspect of the present application. [Brief explanation of the drawings]

[0034] [Figure 1] 1 is a schematic flowchart of acquiring point cloud data by a lidar. [Figure 2] FIG. 1 is a diagram of the architecture of a lidar data transmission system. [Figure 3] 1 is a diagram of the structure of a codec according to an embodiment of the present application; [Figure 4] 1 is a schematic flowchart of a radar data encoding method according to an embodiment of the present application; [Figure 5] FIG. 10 is a diagram showing predicted values ​​of other attribute data when the signal intensity value is 0. [Figure 6] FIG. 10 is a diagram showing predicted values ​​of distance data when the signal strength value is not 0. [Figure 7] 1 is a flowchart of a radar data encoding method applied to point cloud data encoding according to an embodiment of the present application. [Figure 8] 1 is a flowchart of a radar data decoding method according to an embodiment of the present application. [Figure 9] 1 is a flowchart of a radar data decoding method according to an embodiment of the present application. [Figure 10] 1 is a diagram of the structure of a radar data encoding device according to an embodiment of the present application; [Figure 11] 1 is a diagram of the structure of a radar data decoding device according to an embodiment of the present application; [Figure 12] 1 is a diagram of the structure of a computing device according to one embodiment of the present application. DETAILED DESCRIPTION OF THE INVENTION

[0035] The technical solutions of the present application are further described in detail with reference to the accompanying drawings and embodiments below.

[0036] In order to facilitate understanding of the radar data encoding and decoding method, encoding device, and decoding device provided in the embodiments of the present application, the following briefly describes some technical terms in the embodiments of the present application.

[0037] Point cloud: A point cloud is a data set of points in a coordinate system. Point clouds contain a wealth of information, typically including the 3D coordinates of the points (X, Y, Z), but also color, classification values, intensity values, and other information. A dense point cloud can be used to reconstruct the 3D real world. Lidar point clouds are typically obtained by calibrating packet data.

[0038] Packet data: Packet data is a storage form of raw lidar data and is a sorted UDP data packet. Packet data includes information such as the distance of each point, rotation angle, laser ID, signal intensity, reflectivity, and timestamp. Packet data usually contains information in multiple emissions. Packet information can be directly input into the calibration algorithm, and after calibration, a point cloud is obtained. The calibration process is irreversible.

[0039] User Datagram Protocol (UDP) data packet: UDP protocol-based network transmission data. The direct output data of the lidar is sent in the form of UDP packets. The UDP data packets are parsed and sorted to obtain the packet data.

[0040] Lidar raw data: UDP data packets or packet data.

[0041] Single emission: This is the process in which the lidar emits all lasers and receives the reflections. The number of points that can be acquired during one emission is related to the number of radar lines. For example, a 32-line radar can theoretically emit 32 lasers and receive 32 reflections during one emission, resulting in 32 points.

[0042] Attribute: A single emission of radar corresponds to multiple pieces of information, such as distance, angle (azimuth, elevation), and signal strength (intensity), which are called attributes. The modal described in the embodiments of the present application is a type of attribute data. For example, distance, angle, and signal strength belong to different modal information. Various attribute information is directly stored in packet data. When packet data is converted into a point cloud, the distance attribute and angle attribute are converted into geometric attributes.

[0043] Lidar data compression methods can be broadly divided into two main types: one is compressing point clouds, and the other is compressing raw Lidar data, i.e., UDP data packets or packet data. Point cloud compression methods can be further categorized into two types. One is general point cloud compression algorithms that do not use prior knowledge that the point cloud is from Lidar, such as Google Draco or MPEG G-PCC (non-angular mode). The other is point cloud compression algorithms specifically designed for Lidar by using prior knowledge that the point cloud is from Lidar, such as MPEG LiDAR Codec and DJI's China Patent Application Publication No. 111699683.

[0044] Regarding the compression algorithm of lidar raw data, the packet data is rearranged using the radar operating principle and data format, and then the corresponding encoding and decoding algorithm (e.g., prediction and context) is designed, or existing image / video compression methods are used for compression.

[0045] Figure 1 is a schematic flowchart of acquiring a point cloud using a lidar. Point cloud data directly contains information such as the geometric coordinates x, y, and z and the reflectivity of each point, making it intuitive and directly input to algorithms for localization and target recognition. However, compared to raw data, some information and structure have been modified to aid in prediction. Raw lidar data is the actual data transmitted by autonomous vehicles. The data format and information within the data are more centralized. This facilitates single-modal or cross-modal prediction, which allows for more effective data compression.

[0046] Regardless of the method used, and whether point cloud or packet data is compressed, prediction is always one of the most important tools for improving compression efficiency. Most predictions in existing work are performed only in a single modality.

[0047] Point cloud compression solutions are common point cloud algorithms based on octrees. A typical example is MPEG G-PCC (angular mode is disabled). The essence of this algorithm is to enclose the entire point cloud in a bounding box and then perform octree subdivision on the point cloud. If there are still child nodes, further subdivisions are performed until the size of each node reaches the desired accuracy. By building a tree structure, lower-level geometric information can be used to predict higher-level geometric information. After the geometric information is encoded, the signal strength data is encoded. Similarly, the encoded signal strength data is used to encode the uncompressed signal strength data. In this solution, information from different modalities is compressed independently, and potential cross-modal correlations are not used.

[0048] Another compression solution is the MPEG G-PCC LiDAR codec. Like the previous solution, this also compresses point cloud data. However, this solution uses the prior knowledge that the point cloud is from a LiDAR, converts the point cloud back to the LiDAR raw data format, and then attempts to compress the raw data using existing tools such as CABAC.

[0049] As mentioned above, the process of calibrating packet data into a point cloud is lossy. Therefore, the main focus of this solution is how to convert the point cloud back into raw data as accurately as possible. The structure of the raw data, as well as the residuals of distance, angle, and coordinate, must be encoded, and then the attributes (e.g., signal strength) must be encoded. This solution also makes predictions in each modality independently, without using cross-modal attribute data correlations.

[0050] Lidar data (including raw data and point cloud data) has attribute data of multiple modalities, such as distance, signal strength, angle, and reflectivity. The data of multiple modalities are correlated to some extent. In the above solutions, the attribute data of each modal is predicted independently, and the correlation between the data of different modalities cannot be effectively utilized.

[0051] Embodiments of the present application provide a radar data encoding and decoding method, an encoding device, and a decoding device for performing cross-modal attribute data prediction using correlations between different modal attribute data of radar data and encoding the attribute data to improve data compression efficiency.

[0052] Specific implementation forms of the radar data encoding and decoding method, encoding device, and decoding device provided in the embodiments of the present application will be specifically described below.

[0053] As technology advances, affordable lidars that meet automotive regulations are gradually emerging. At the same time, lidars have brought qualitative and essential performance improvements in applications such as automatic emergency braking (AEB), gradually becoming a standard feature in mid- to high-end vehicles. Similar to on-board cameras, radar data also needs to be stored and transmitted. The large volume of data from multiple high-line-count lidars requires support from compression algorithms. Embodiments of the present application provide a series of cross-modal prediction methods for lidar data compression / encoding / decoding to improve compression efficiency by utilizing correlations between different modal information.

[0054] FIG. 2 illustrates a lidar data transmission system. The radar data encoding and decoding method, encoding device, and decoding device provided in the embodiments of the present application can be applied to the lidar data transmission system. As shown in FIG. 2, lidar data transmission mainly includes the following steps: radar data encoding, radar data encapsulation, channel transmission, radar data deencapsulation, and radar data decoding. The purpose of radar data encoding is to obtain an encoded file with a high compression rate for transmission and reduce the network transmission load. The purpose of radar data decoding is to decode the data stream (i.e., the encoded data) to reconstruct the radar data. The purpose of radar data encapsulation is to convert the encoded file into a format that is easy to transmit. In the embodiments of the present application, the radar data codec is improved, and the encoding efficiency is improved.

[0055] The improved codec is shown in Figure 3. In the embodiment of the present application, the cross-modal prediction section in the prediction module is improved. In most of the above solutions, only intra-modal prediction is performed. In the embodiment of the present application, more modal and more complex prediction relationships and methods are introduced.

[0056] 4 is a schematic flowchart of a radar data encoding method according to an embodiment of the present application. This method can be implemented in the encoder of FIG. 3. As shown in FIG. 4, the radar data encoding method provided in this embodiment of the present application includes at least steps S401 to S404.

[0057] In step S401, data to be coded is obtained.

[0058] The encoding target data may be radar raw data or point cloud data. Here, the encoding target data that is radar raw data will be described first. The radar may optionally be a lidar or a microwave radar. Here, a lidar is used as an example. The lidar raw data is a UDP data packet or packet data.

[0059] The UDP data packet includes additional information and measurement data. The additional information includes a global positioning system (GPS) timestamp, echo, and universal time coordinated (UTC) used to obtain time information. The measurement data includes M data blocks, each corresponding to one lidar emission, during which one azimuth angle measurement data and N channel unit measurement data (representing that an N-line lidar has N laser emission modules) can be obtained. When the raw data is a UDP data packet, the UDP data packet needs to be processed to obtain the packet data.

[0060] The packet data includes multiple data blocks, each corresponding to one emission of the lidar, and each data block includes one angle measurement data and N groups of multiple types of attribute data of different modalities, such as distance data, signal strength data, angle data, and reflectivity data, with one-to-one correspondence between the multiple types of modal data.

[0061] The data to be encoded in this embodiment of the present application is measurement data collected by the lidar during M emissions, where M is equal to or greater than 1. The value of N, i.e., the number of groups of measurement data, is related to the number of lidar lines, which is not specifically limited in the present application.

[0062] After the packetized data is acquired, the measurement data in the packetized data may be pre-processed, for example, attribute data of different modalities in the measurement data may be sorted separately to facilitate encoding compression for the attribute data of different modalities.

[0063] In step S402, the first attribute data is encoded to obtain first encoded data.

[0064] As mentioned above, raw radar data includes attribute data of several different modalities, for example, range data, angle data, and signal strength data.

[0065] The attribute data of one of the modalities may be encoded first, for example, the signal strength data may be encoded first.

[0066] In this embodiment, the signal strength data may be encoded using multiple encoding schemes to obtain encoded signal strength data, for example, encoding schemes for the signal strength data include, but are not limited to, run-length encoding and entropy encoding.

[0067] In step S403, the second attribute data and the third attribute data are predicted based on the first attribute data to obtain predicted values ​​of the second attribute data and the third attribute data.

[0068] There are correlations between attribute data of multiple modals of radar data, so uncoded attribute data can be predicted using correlations between coded attribute data and cross-modal attribute data to obtain predicted values ​​of other uncoded attribute data.

[0069] For example, the attribute value corresponding to the first moment of the first attribute data is 0, and the predicted values ​​corresponding to the first moments of the second attribute data and the third attribute data are both 0, and the attribute data with a predicted value of 0 is configured to be skipped for encoding.

[0070] When the first attribute data, the second attribute data, and the third attribute data are signal strength, angle, and distance information, respectively, the signal strength is encoded first. When the angle and distance are encoded, the encoded signal strength data is referenced. If the signal strength of this radiation is 0, both the angle value and the distance value are 0. Therefore, the angle value and the distance value that are 0 may be directly skipped and do not need to be encoded (see FIG. 5).

[0071] This is because a signal strength of 0 indicates no signal or a very weak signal, and no useful information can be obtained. Therefore, both the angle and distance values ​​corresponding to the signal strength are shown as 0. Therefore, to improve the compression rate of radar data, it may be possible to perform cross-modal coding on these three values.

[0072] In other words, the encoder may skip data that is predicted to have an attribute value of 0 without encoding the data, and no positions for attribute data with a value of 0 are reserved in the data stream. In this way, when the encoder encodes the multimodal attribute data, attribute data with a value of 0 may be considered not to exist. This can significantly reduce the amount of data. More 0 values ​​indicate better compression efficiency, further saving bandwidth during data transmission.

[0073] For non-zero data in the first attribute data, predicted values ​​of other attribute data in the current frame can be predicted based on the variation characteristics of the first attribute data and the attribute values ​​of other attribute data in the encoded frame. Here, the current frame refers to the frame to be encoded. The encoded frame refers to the frame being encoded. The variation characteristics of the first attribute data indicate the variation characteristics of the first attribute data between two frames corresponding to the encoded frame and the current frame of the attribute data to be encoded.

[0074] For example, if the attribute data to be encoded is the second attribute data, the encoded frame is the Nth frame, and the current frame (also called the frame to be encoded) is the Mth frame, the change characteristic of the first attribute data is the change characteristic of the first attribute data between the Nth frame and the Mth frame.

[0075] The sequence of the coded frame and the current frame is related to the coding sequence of the second attribute data. If the second attribute data is coded from front to back, the coded frame is the Nth frame before the current frame (i.e., the past frame). If the second attribute data is coded from back to front, the coded frame is the Nth frame after the current frame (i.e., the future frame). In this embodiment of the present application, the order of the coded frames is not particularly limited.

[0076] Alternatively, the predicted value of the current frame may also be predicted by interpolation. For example, the encoded frames are frame E and frame F, and the current frame is frame H. Frame H is a frame between frame E and frame F. In this case, the variation characteristic of the first attribute data is the variation characteristic of the first attribute data from frame E and frame F to frame H. It will be understood that the attribute values ​​of moments described below have the same meaning as the attribute values ​​within a frame.

[0077] In one example, a predicted value of the second attribute data at a current moment is determined based on a change characteristic of the first attribute data and an attribute value of the second attribute data at a previous moment, and a predicted value of the third attribute data at a current moment is determined based on a change characteristic of the first attribute data and an attribute value of the third attribute data at a previous moment.

[0078] In one example, the variation characteristic of the first attribute data may be an optical flow corresponding to the first attribute data. As shown in Figure 6, for distance data that needs to be encoded, if distance data at moment t is encoded, the optical flow I is first calculated by using the encoded signal strength data at moments t and t-1, and then the distance data at moment t is calculated by

number

[0079] In steps S404 and S405, the second attribute data is encoded based on the predicted value of the second attribute data to obtain second encoded data, and the third attribute data is encoded based on the predicted value of the third attribute data to obtain third encoded data.

[0080] To obtain the encoded data, the attribute data is encoded based on the predicted values ​​of the other unencoded attribute data obtained in step S403.

[0081] For example, the unencoded second attribute data is distance data. The remainder of the distance data may be calculated based on the predicted value of the distance data obtained through prediction and the actual value (also referred to as real value) of the distance data, and then the distance data is encoded based on context and entropy encoding to obtain encoded data of the distance data. Similarly, the encoded data of the unencoded third attribute data is obtained to compress the radar raw data.

[0082] In other words, according to the radar data encoding method provided in this embodiment of the present application, when raw radar data is encoded, attribute data of any of multiple modal attributes may be pre-encoded, and then cross-modal prediction is performed on any attribute data among other attribute data using the encoded attribute data to improve prediction accuracy. More accurate prediction allows the remainder that needs to be encoded to be distributed more intensively (close to the value 0). In this way, the compression performance of raw radar data is improved in cooperation with entropy encoding.

[0083] In another embodiment, the radar data encoding method provided in this embodiment of the present application may also be applied to encoding and compressing point cloud data.

[0084] 7 is a flowchart of a radar data encoding method applied to point cloud data encoding according to an embodiment of the present application. As shown in FIG. 7, the method includes at least steps S701 to S704.

[0085] In step S701, point group data to be encoded is obtained.

[0086] The point cloud data to be encoded can be obtained by calibrating the packet data. The point cloud data includes multi-modal attribute data, such as geometric data (x, y, z), or referred to as position data, signal strength, and reflectivity.

[0087] In step S702, the fourth attribute data and the fifth attribute data are coded separately to obtain fourth coded data and fifth coded data.

[0088] The fourth attribute data and the fifth attribute data may be any two of a plurality of cross-modal attribute data (for example, geometric data, signal intensity, and reflectance) included in the point cloud data.

[0089] For example, the geometric data and the signal strength are coded separately to obtain coded data corresponding to the geometric data and coded data corresponding to the signal strength data.

[0090] The particular scheme for encoding the geometric data and signal strength is not limited in this embodiment of the present application and may be any suitable encoding scheme.

[0091] For example, the geometric data is first coded, then the signal strength is predicted using the coded geometric data to obtain a predicted value of the signal strength data, and then the remainder of the predicted value of the signal strength and the actual value is calculated, and entropy coding is performed based on the coding remainder.

[0092] If the geometric data in the same group is a value of 0, the predicted value of the signal strength is determined to be a value of 0. For the geometric data that is not a value of 0, the predicted value of the signal strength can be predicted according to a preset polynomial. For example, the polynomial may be y=a*x+b, where y represents the geometric data, x represents the predicted value of the signal strength, and a and b are parameters that are preset according to the correlation between the geometric data and the signal strength.

[0093] If the geometric data is not 0, the variation characteristics of the geometric data may be further determined based on the geometric data of the current moment and the previous moment, and the predicted value of the signal strength of the current moment is predicted based on the variation characteristics of the geometric data and the value of the signal strength data of the previous moment. For example, when encoding the signal strength of frame t, the encoded geometric data of frames t and t-1 are used to determine a matching relationship I={x, y, z} consisting of a total of three-dimensional components. t-1 →{x,y,z} t may be constructed, and a predicted value of the signal strength data at moment t may be calculated with reference to the coded signal strength data at moment t-1.

number

[0094] In step S703, sixth attribute data is predicted based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute.

[0095] The predicted value of the other attribute data in the current frame may be predicted using the change characteristics of the fourth attribute data and the fifth attribute data and the attribute values ​​of the other attribute data in the encoded frame. Here, the current frame refers to the frame to be encoded. The encoded frame refers to the frame being encoded. The change characteristics of the fourth attribute data and the fifth attribute data indicate the change characteristics of the fourth attribute data and the fifth attribute data between two frames corresponding to the encoded frame and the current frame of the attribute data to be encoded.

[0096] For example, if the attribute data to be encoded is the sixth attribute data, the encoded frame is the Nth frame, and the current frame (also called the frame to be encoded) is the Mth frame, the change characteristics of the fourth attribute data and the fifth attribute data are the change characteristics of the fourth attribute data and the fifth attribute data between the Nth frame and the Mth frame.

[0097] The sequence of the coded frame and the current frame is related to the coded sequence of the sixth attribute data. If the sixth attribute data is coded from front to back, the coded frame is the Nth frame before the current frame (i.e., the past frame). If the sixth attribute data is coded from back to front, the coded frame is the Nth frame after the current frame (i.e., the future frame). In this embodiment of the present application, the order of the coded frames is not particularly limited.

[0098] In one example, the predicted value of the sixth attribute data of the current moment may be determined using the change characteristics of the fourth attribute data and the fifth attribute data and the attribute value of the sixth attribute data of the previous moment.

[0099] For example, the fourth attribute data is geometry data, the fifth attribute data is signal intensity data, and the sixth attribute data is reflectivity data. If reflectivity in frame t is encoded, a matching relation I={x, y, z, intensity} containing a total of four dimensional components can be obtained by using the encoded geometry and signal intensity data in frames t and t-1. t-1 →{x,y,z,intensity} t may be constructed and the reflectance at moment t predicted by using the encoded reflectance at moment t-1:

number

[0100] That is, cross-modal prediction may be performed on attribute data of the other modal predicted using two types of encoded attribute data.

[0101] In step S704, the sixth attribute is encoded based on the predicted value of the sixth attribute to obtain sixth encoded data.

[0102] To obtain the coded data, the attribute data is coded based on the predicted values ​​of the other uncoded attribute data obtained in step S703.

[0103] For example, the sixth attribute data that is not coded is reflectance data. The remainder of the reflectance data may be calculated based on the predicted value obtained by the coding and the actual value (also called real value) of the reflectance data, and the reflectance data is coded based on the context and entropy coding to obtain coded data of the reflectance data.

[0104] The radar data encoding method provided in this embodiment of the present application may be applied to encoding and compressing radar point cloud data. When radar data is encoded, the correlation between different modal information of the lidar is better utilized. Any two attribute data among multiple modal attributes are pre-encoded, and cross-modal prediction is performed on any attribute data among the other attribute data using the encoded attribute data to improve prediction accuracy. More accurate prediction allows the remainder that needs to be encoded to be distributed more intensively (near the value 0). In this way, the compression performance of radar point cloud data is improved in cooperation with entropy encoding.

[0105] Correspondingly, an embodiment of the present application further provides a radar data decoding method, which can be applied to a decoder to decode encoded data encoded according to the radar data encoding method provided in the embodiment of the present application and reconstruct radar data.

[0106] 8 is a flowchart of a radar data decoding method according to an embodiment of the present application. The method is used to decode the encoded data of radar raw data to reconstruct the radar raw data. The method can be implemented in the decoder shown in FIG. 3. As shown in FIG. 8, the radar data decoding method provided in this embodiment of the present application includes at least steps S801 to S806.

[0107] In step S801, the first encoded data is obtained.

[0108] The first encoded data is encoded data of the first attribute data of the radar raw data, that is, encoded data of the signal strength data described above.

[0109] In step S802, the first encoded data is decoded to obtain first attribute data.

[0110] When decoding is performed in a decoder, first, the encoded data corresponding to the first attribute data is decoded to reconstruct the first attribute data.

[0111] In step S803, the second attribute data and the third attribute data are predicted based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data.

[0112] Thereafter, when other attribute data, such as the encoded data of the second attribute data and the encoded data of the third attribute data, are decoded, the attribute data of the first attribute data, such as the signal strength data described above, is referenced. If the signal strength data in this radiation is 0, both the corresponding angle and the corresponding distance value are set to 0.

[0113] For non-zero values ​​of signal strength data, the angle and distance are predicted with reference to the prediction methods described above to obtain predicted values ​​of angle data and predicted values ​​of distance data.

[0114] In step S804, the second coded data and the third coded data are decoded to obtain a remainder value of the second attribute data and a remainder value of the third attribute data.

[0115] The coded data of the second attribute data and the coded data of the third attribute data are decoded separately to obtain a remainder value of the second attribute data and a remainder value of the third attribute data.

[0116] Next, in steps S805 and S806, second attribute data is obtained based on the predicted value of the second attribute and the remainder value of the second attribute, and third attribute data is obtained based on the predicted value of the third attribute and the remainder value of the third attribute. Finally, the radar raw data is reconstructed.

[0117] 9 is a flowchart of a radar data decoding method according to an embodiment of the present application. The method is used to decode the encoded data of radar point cloud data to reconstruct the radar point cloud data. The method can be implemented in the decoder shown in FIG. 3. As shown in FIG. 9, the radar data decoding method provided in this embodiment of the present application includes at least steps S901 to S905.

[0118] In step S901, the fourth encoded data and the fifth encoded data are obtained.

[0119] The fourth and fifth coded data are coded data of the fourth and fifth attribute data of the radar point cloud data, respectively, i.e., coded data of the above-mentioned geometric data and coded data of the signal intensity data.

[0120] In step S902, the fourth encoded data and the fifth encoded data are decoded separately to obtain fourth attribute data and fifth attribute data.

[0121] When decoding is performed in a decoder, first, the encoded data corresponding to the fourth attribute data and the fifth attribute data is decoded to reconstruct the fourth attribute data and the fifth attribute data.

[0122] In step S903, the sixth attribute data is predicted based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data.

[0123] Next, when the encoded data of other attribute data, for example, the sixth attribute data, is decoded, the fourth attribute data and the fifth attribute data, for example, the geometric data and the signal strength data, are referenced. If both the geometric data and the signal strength data in this radiation are 0, the corresponding reflectance is set to 0.

[0124] For non-zero values ​​in the geometry and signal intensity data, the reflectance is predicted with reference to the prediction methods described above to obtain predicted values ​​of the reflectance.

[0125] In step S904, the sixth encoded data is decoded to obtain a remainder value of the sixth attribute data.

[0126] The encoded data of the sixth attribute data is decoded to obtain a remainder value of the sixth attribute data.

[0127] Then, in the following step S905, the sixth attribute data is acquired based on the predicted value of the sixth attribute data and the remainder value of the sixth attribute data, and finally the radar point cloud data is reconstructed.

[0128] Based on the same concept as the aforementioned embodiment of the radar data encoding method, an embodiment of the present application further provides a radar data encoding device 1000. The radar data encoding device 1000 includes units or modules configured to implement the steps of the radar data encoding method shown in Figures 3 to 7.

[0129] 10 is a diagram of the structure of a radar data encoding apparatus according to one embodiment of the present application. The apparatus may be located in any apparatus, device, platform, or device cluster having computing capabilities, such as a computer or a server. The apparatus may be located in the encoder shown in FIG. 3 to implement radar data encoding and compression.

[0130] 10 , the radar data encoding device 1000 includes at least an acquisition module 1001, a first encoding module 1002, and a first prediction module 1003. The acquisition module 1001 is configured to acquire encoding target data, which may be radar raw data or point cloud data. If the encoding target data is raw data, the encoding target data may include first attribute data, second attribute data, and third attribute data of different modalities. The first encoding module 1002 is configured to encode the first attribute data to obtain first encoded data. The first prediction module 1003 is configured to predict the second attribute data and the third attribute data based on the first attribute data to obtain predicted values ​​of the second attribute data and the third attribute data. The first encoding module 1002 is further configured to encode the second attribute data based on the predicted value of the second attribute data to obtain second encoded data, and to encode the third attribute data based on the predicted value of the third attribute data to obtain third encoded data.

[0131] In another possible implementation form, the first prediction module 1003 is specifically configured such that the attribute value corresponding to the first moment of the first attribute data is 0, and the predicted values ​​corresponding to the first moment of the second attribute data and the third attribute data are both 0, and the attribute data having the predicted value 0 is configured to skip encoding.

[0132] In another possible implementation form, the first prediction module 1003 is particularly configured to: obtain a variation characteristic of the first attribute data; determine a predicted value of the second attribute data of the current moment based on the variation characteristic of the first attribute data and an attribute value of the second attribute data of the previous moment; and determine a predicted value of the third attribute data of the current moment based on the variation characteristic of the first attribute data and an attribute value of the third attribute data of the previous moment.

[0133] In another possible implementation, obtaining the change characteristic of the first attribute data includes determining the change characteristic of the first attribute data based on an attribute value of the first attribute data of the current moment and an attribute value of the first attribute data of the previous moment.

[0134] In another possible implementation, the first attribute data, the second attribute data, and the third attribute data are each one of distance information, signal strength information, angle information, and reflectance information.

[0135] In another possible implementation, when the data to be encoded is point cloud data, the data to be encoded includes fourth, fifth, and sixth attribute data of different modalities. The radar data encoding device 1000 provided in this embodiment of the present application further includes a second encoding module 1004 and a second prediction module 1005. The second encoding module 1004 is configured to separately encode the fourth attribute data and the fifth attribute data to obtain fourth encoded data and fifth encoded data. The second prediction module 1005 is configured to predict the sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute. The second encoding module is further configured to encode the sixth attribute based on the predicted value of the sixth attribute to obtain sixth encoded data.

[0136] In another possible implementation form, the second prediction module 1005 is particularly configured to obtain change characteristics of the fourth attribute data and the fifth attribute data, and determine a predicted value of the sixth attribute data of the current moment based on the change characteristics of the fourth attribute data and the fifth attribute data and the attribute value of the sixth attribute data of the previous moment.

[0137] In another possible implementation form, an implementation form for obtaining the change characteristics of the fourth attribute data and the fifth attribute data is to determine the change characteristics of the attribute values ​​of the fourth attribute data and the fifth attribute data based on the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data of a previous moment and the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data of a current moment.

[0138] The radar data encoding device 1000 according to this embodiment of the present application can correspondingly perform the methods described in the embodiments of the present application, and the above and other operations and / or functions of the modules in the radar data encoding device 1000 are separately used to implement corresponding procedures of the methods of Figures 3 to 7. For the sake of brevity, the details will not be described again here.

[0139] Correspondingly, an embodiment of the present application further provides a radar data decoding device 1100. The radar data decoding device 1100 includes units or modules configured to implement the steps of the radar data decoding method shown in Figures 8 and 9.

[0140] 11 is a diagram of the structure of a radar data decoding apparatus according to an embodiment of the present application. The apparatus may be located in any apparatus, device, platform, or device cluster having computing capabilities, such as a computer or a server. The apparatus may be located in the decoder shown in FIG. 3 to decode radar encoded data and reconstruct radar data.

[0141] 11, the radar data decoding device 1100 includes at least a first acquisition module 1101, a first decoding module 1102, and a first prediction module 1103. When the radar data is raw data, the first acquisition module 1101 is configured to acquire first encoded data. The first decoding module 1102 is configured to decode the first encoded data to acquire first attribute data. The first prediction module 1103 is configured to predict second attribute data and third attribute data based on the first attribute data to obtain predicted values ​​of the second attribute data and the third attribute data. The first decoding module 1102 is further configured to: decode the second encoded data and the third encoded data to obtain a remainder value of the second attribute data and a remainder value of the third attribute data; obtain the second attribute data based on the predicted value of the second attribute data and the remainder value of the second attribute data; and obtain the third attribute data based on the predicted value of the third attribute data and the remainder value of the third attribute data.

[0142] In another possible implementation, the radar data decoding device 1100 provided in this embodiment of the present application further includes a second acquisition module 1104, a second decoding module 1105, and a second prediction module 1106. When the radar data is point cloud data, the second acquisition module 1104 is configured to acquire fourth encoded data and fifth encoded data. The second decoding module 1105 is configured to separately decode the fourth encoded data and the fifth encoded data to obtain fourth attribute data and fifth attribute data. The second prediction module 1106 is configured to predict sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data. The second decoding module is further configured to decode the sixth encoded data to obtain a remainder value of the sixth attribute data, and obtain the sixth attribute data based on the predicted value of the sixth attribute data and the remainder value of the sixth attribute data.

[0143] The radar data decoding device 1100 according to this embodiment of the present application can correspondingly perform the methods described in the embodiments of the present application, and the above and other operations and / or functions of the modules in the radar data decoding device 1100 are separately used to implement the corresponding procedures of the methods of Figures 8 and 9. For the sake of brevity, the details will not be described again here.

[0144] Based on the same concept as the above-mentioned method embodiment, an embodiment of the present application further provides a computing device. The electronic device includes at least a processor and a memory. The memory stores a program. When the processor executes the program, the units or modules of the method steps shown in Figures 3 to 9 can be implemented.

[0145] FIG. 12 is a diagram of the structure of a computing device according to one embodiment of the present application.

[0146] 12, computing device 1200 includes at least one processor 1201, memory 1202, and communication interface 1203. Processor 1201, memory 1202, and communication interface 1203 may be communicatively connected via a wire (e.g., a bus) or wirelessly. Communication interface 1203 is configured to receive data (e.g., radar data) transmitted by another device. Memory 1202 stores computer instructions, and processor 1201 executes the computer instructions to perform the method in the aforementioned method embodiments.

[0147] It is to be understood that in this embodiment of the present application, processor 1201 may be a central processing unit CPU, or processor 1201 may be another general-purpose processor, a digital signal processor (DSP), or an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or another programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc. The general-purpose processor may be a microprocessor, or any conventional processor, etc.

[0148] The memory 1202 may include read-only memory and random access memory and may provide instructions and data to the processor 1201. The memory 1202 may further include non-volatile random access memory.

[0149] The memory 1202 may be volatile or nonvolatile memory, or may include both volatile and nonvolatile memory. Nonvolatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may be random access memory (RAM), which acts as an external cache. By way of example and not limitation, many forms of RAM may be used, such as static random access memory (static RAM, SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchlink dynamic random access memory (synchlink DRAM, SLDRAM), and direct rambus random access memory (direct rambus RAM, DR RAM).

[0150] It should be understood that the computing device 1200 according to this embodiment of the present application can execute and implement the methods shown in Figures 3 to 9 in the embodiments of the present application. For detailed descriptions of the implementation forms of the methods, please refer to the above descriptions. For the sake of brevity, the details will not be described again here.

[0151] One embodiment of the present application provides a computer-readable storage medium, which stores a computer program, the computer instructions of which, when executed by a processor, perform the aforementioned method.

[0152] One embodiment of the present application provides a chip, the chip including at least one processor and an interface, the at least one processor using the interface to determine program instructions or data, the at least one processor configured to execute the program instructions to implement the aforementioned method.

[0153] One embodiment of the present application provides a computer program or computer program product, which comprises instructions that, when executed, enable a computer to perform the method described above.

[0154] Those skilled in the art will further recognize that, in combination with the examples described in the embodiments disclosed herein, the units and algorithm steps can be implemented by electronic hardware, computer software, or a combination of the two. To clearly explain the compatibility between hardware and software, the configurations and steps of each example are generally described in the foregoing specification based on their functions. Whether a function is performed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can implement the described functions using various methods for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0155] The steps of a method or algorithm described in the embodiments disclosed herein may be implemented by hardware and software modules executed by a processor, or a combination of hardware and software modules executed by a processor. The software modules may be comprised in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0156] The above specific embodiments further describe the objectives, technical solutions, and beneficial effects of the present application. It should be understood that the above descriptions are only specific implementation forms of the present application and are not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, or improvements made without departing from the spirit and principles of this application shall fall within the protection scope of this application. [Explanation of symbols]

[0157] 1000 Radar data encoding device 1001 Acquisition Module 1002 first encoding module 1003 First Prediction Module 1004 Second Encoding Module 1005 Second Prediction Module 1100 Radar data decoder 1101 First Acquisition Module 1102 first decoding module 1103 First Prediction Module 1104 Second Acquisition Module 1105 Second Decryption Module 1106 Second Prediction Module 1200 computing devices 1201 processor 1202 memory 1208 Communication Interface

Claims

1. 1. A radar data encoding method, comprising: acquiring encoding target data, the encoding target data being radar raw data or point cloud data, and when the encoding target data is the raw data, the encoding target data including first attribute data, second attribute data, and third attribute data of different modals; encoding the first attribute data to obtain first encoded data; predicting the second attribute data and the third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data; encoding the second attribute data based on the predicted value of the second attribute data to obtain second encoded data; encoding the third attribute data based on the predicted value of the third attribute data to obtain third encoded data; A method comprising:

2. the step of predicting the second attribute data and the third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data, an attribute value corresponding to a first moment in the first attribute data being set to 0, and predicted values ​​corresponding to the first moments in the second attribute data and the third attribute data being set to 0, wherein the attribute data having a predicted value of 0 is configured to skip encoding; 2. The method of claim 1, comprising:

3. the step of predicting the second attribute data and the third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data, acquiring a change characteristic of the first attribute data; determining a predicted value of the second attribute data in a current frame based on the variation characteristic of the first attribute data and an attribute value of the second attribute data in an encoded frame; determining a predicted value of the third attribute data in the current frame based on the variation characteristic of the first attribute and an attribute value of the third attribute data in the encoded frame; Including, the variation characteristic indicates a variation characteristic of the first attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame; 3. The method according to claim 1 or 2.

4. The step of acquiring a change characteristic of the first attribute data includes: determining the variation characteristics of the first attribute data based on attribute values ​​of the first attribute data in the encoded frame and attribute values ​​of the first attribute data in the current frame; 4. The method of claim 3, comprising:

5. 5. The method of claim 1, wherein the first attribute data, the second attribute data, and the third attribute data are each one of distance data, signal strength data, angle data, and reflectivity data.

6. When the encoding target data is the point cloud data, the encoding target data includes fourth attribute data, fifth attribute data, and sixth attribute data of different modals; The method comprises: separately encoding the fourth attribute data and the fifth attribute data to obtain fourth encoded data and fifth encoded data; predicting the sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute; encoding the sixth attribute based on the predicted value of the sixth attribute to obtain the sixth encoded data; further comprising:

6. The method according to any one of claims 1 to 5.

7. the step of predicting the sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute, acquiring change characteristics of the fourth attribute data and the fifth attribute data; determining a predicted value of the sixth attribute data in the current frame based on the change characteristics of the fourth attribute data and the fifth attribute data and an attribute value of the sixth attribute data in an encoded frame; Including, the change characteristics indicate change characteristics of the fourth attribute data and the fifth attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame. The method of claim 6.

8. the step of acquiring change characteristics of the fourth attribute data and the fifth attribute data, determining the change characteristics of the attribute values ​​of the fourth attribute data and the fifth attribute data based on the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data in the encoded frame and the attribute values ​​of the fourth attribute data and the attribute values ​​of the fifth attribute data in the current frame; 8. The method of claim 7, comprising:

9. 1. A radar data decoding method, comprising: If the radar data is raw data, obtaining first encoded data; decoding the first encoded data to obtain first attribute data; predicting second attribute data and third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data; decoding the second coded data and the third coded data to obtain a remainder value of the second attribute data and a remainder value of the third attribute data; obtaining the second attribute data based on the predicted value of the second attribute data and the remainder value of the second attribute data; acquiring the third attribute data based on the predicted value of the third attribute data and the remainder value of the third attribute data; A method comprising:

10. If the radar data is point cloud data, acquiring fourth encoded data and fifth encoded data; separately decoding the fourth encoded data and the fifth encoded data to obtain fourth attribute data and fifth attribute data; predicting sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data; decoding the sixth encoded data to obtain a remainder value of the sixth attribute data; acquiring the sixth attribute data based on the predicted value of the sixth attribute data and the remainder value of the sixth attribute data; 10. The method of claim 9, comprising:

11. A radar data encoding device, comprising: an acquisition module configured to acquire encoding target data, the encoding target data being radar raw data or point cloud data, and when the encoding target data is the raw data, the encoding target data including first attribute data, second attribute data, and third attribute data of different modals; a first encoding module configured to encode the first attribute data to obtain first encoded data; a first prediction module configured to predict the second attribute data and the third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data; Equipped with The first encoding module: encoding the second attribute data based on the predicted value of the second attribute data to obtain second encoded data; Encoding the third attribute data based on the predicted value of the third attribute data to obtain third encoded data. further configured as follows: Device.

12. 12. The apparatus of claim 11, wherein the first prediction module is further configured to set an attribute value corresponding to a first moment in the first attribute data to 0, and to set both predicted values ​​corresponding to the first moments in the second attribute data and the third attribute data to 0, and to skip encoding of attribute data with a predicted value of 0.

13. The first prediction module: acquiring a change characteristic of the first attribute data; determining a predicted value of the second attribute data in a current frame based on the variation characteristic of the first attribute data and an attribute value of the second attribute data in an encoded frame, where N is a positive integer; determining a predicted value of the third attribute data in the current frame based on the variation characteristic of the first attribute and an attribute value of the third attribute data in the encoded frame; further configured as follows: the variation characteristic indicates a variation characteristic of the first attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame; 13. Apparatus according to claim 11 or 12.

14. Acquiring the change characteristic of the first attribute data includes: determining the variation characteristic of the first attribute data based on an attribute value of the first attribute data in the encoded frame and an attribute value of the first attribute data in the current frame; 14. The apparatus of claim 13, comprising:

15. The apparatus of claim 11 , wherein the first attribute data, the second attribute data, and the third attribute data are each one of distance data, signal strength data, angle data, and reflectivity data.

16. When the encoding target data is the point cloud data, the encoding target data includes fourth attribute data, fifth attribute data, and sixth attribute data of different modals; The device comprises: a second encoding module configured to separately encode the fourth attribute data and the fifth attribute data to obtain fourth encoded data and fifth encoded data; a second prediction module configured to predict the sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute; Furthermore, the second encoding module is further configured to encode the sixth attribute based on the predicted value of the sixth attribute to obtain the sixth encoded data.

16. Apparatus according to any one of claims 11 to 15.

17. The second prediction module: acquiring change characteristics of the fourth attribute data and the fifth attribute data; determining a predicted value of the sixth attribute data in the current frame based on the change characteristics of the fourth attribute data and the fifth attribute data and the attribute value of the sixth attribute data in the encoded frame; further configured as follows: the change characteristics indicate change characteristics of the fourth attribute data and the fifth attribute data between a frame corresponding to the encoded frame and a frame corresponding to the current frame.

17. The apparatus of claim 16.

18. Obtaining the change characteristics of the fourth attribute data and the fifth attribute data includes: determining the change characteristics of the attribute value of the fourth attribute data and the fifth attribute data based on the attribute value of the fourth attribute data and the attribute value of the fifth attribute data in the encoded frame and the attribute value of the fourth attribute data and the attribute value of the fifth attribute data in the current frame; 18. The apparatus of claim 17, comprising:

19. A radar data decoding device, comprising: a first acquisition module configured to acquire first encoded data when the radar data is raw data; a first decoding module configured to decode the first encoded data to obtain first attribute data; a first prediction module configured to predict second attribute data and third attribute data based on the first attribute data to obtain a predicted value of the second attribute data and a predicted value of the third attribute data; Equipped with The first decoding module: decoding the second coded data and the third coded data to obtain a remainder value of the second attribute data and a remainder value of the third attribute data; acquiring the second attribute data based on the predicted value of the second attribute data and the remainder value of the second attribute data; acquiring the third attribute data based on the predicted value of the third attribute data and the remainder value of the third attribute data; further configured as follows: Device.

20. a second acquisition module configured to acquire fourth encoded data and fifth encoded data when the radar data is point cloud data; a second decoding module configured to separately decode the fourth encoded data and the fifth encoded data to obtain fourth attribute data and fifth attribute data; a second prediction module configured to predict sixth attribute data based on the fourth attribute data and the fifth attribute data to obtain a predicted value of the sixth attribute data; Furthermore, The second decoding module: Decoding the sixth encoded data to obtain a remainder value of the sixth attribute data; acquiring the sixth attribute data based on the predicted value of the sixth attribute data and the remainder value of the sixth attribute data; further configured as follows:

20. The apparatus of claim 19.

21. 11. A computing device comprising: a memory; and a processor, wherein the memory stores executable code and the processor executes the executable code to perform the method of any one of claims 1 to 10.

22. 11. A computer-readable storage medium storing a computer program that, when executed on a computer, enables the computer to perform the method of any one of claims 1 to 10.

Images