Point cloud data preprocessing method and point cloud geometry encoding and decoding method and device

Abstract

The present invention discloses a method for preprocessing point cloud data and a method and device for point cloud geometry encoding and decoding. The preprocessing method includes: performing coordinate conversion on the original point cloud data to obtain a representation of the original point cloud in a cylindrical coordinate system; unfolding the cylindrical coordinate system to obtain a two-dimensional structure; and regularizing the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure. The encoding method includes: performing predictive encoding on the preprocessed original point cloud data to obtain a geometric information code stream. The point cloud geometry encoding method provided by the present invention performs regularized preprocessing on the original point cloud so that the point cloud presents a regular distribution in both the horizontal and vertical directions, better reflects the spatial correlation of the point cloud, and improves the encoding efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data processing, and in particular relates to a point cloud data preprocessing method and a point cloud geometry encoding and decoding method and device. Background Art

[0002] In the point cloud G-PCC (Geometry-based Point Cloud Compression) encoder framework, the geometric information of the point cloud and the attribute information corresponding to each point are encoded separately. Currently, the geometric coding and decoding of G-PCC can be divided into octree-based geometric coding and decoding and prediction tree-based geometric coding and decoding. The prediction tree-based geometric coding first sorts the input point cloud and establishes a prediction tree structure at the encoding end using two different methods. Then, based on the structure of the prediction tree, each node in the prediction tree is traversed, and the geometric position information of the node is predicted by selecting different prediction modes to obtain the prediction residual, and the geometric prediction residual is quantized using the quantization parameter. Finally, through continuous iteration, the prediction residual of the prediction tree node position information, the prediction tree structure, and the quantization parameters are encoded to generate a binary code stream.

[0003] Prediction tree coding based on LiDAR calibration information is a commonly used geometric coding method. For each LiDAR laser scanner, the acquisition points belonging to the same laser scanner should be regularly distributed in a cylindrical coordinate system. However, due to factors such as noise, measurement errors, and device jitter, actual data exhibits a non-uniform distribution, resulting in poor correlation between data, low prediction accuracy, and low coding efficiency.

[0004] However, point cloud encoding and decoding technology based on prediction trees only uses a subset of the LiDAR device's parameters to establish a tree structure. This tree structure doesn't fully reflect the spatial correlation of the point cloud, hindering point cloud prediction and entropy coding, thus affecting coding efficiency. Existing G-PCC methods only determine the relationship between each point and the laser scanner through vertical correction, requiring the introduction of additional variables to assist with encoding horizontal information during encoding. This increases the amount of information required and reduces geometric coding efficiency. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a point cloud geometry prediction encoding and decoding method and device based on a regularized structure. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0006] A point cloud data preprocessing method, comprising:

[0007] Perform coordinate transformation on the original point cloud data to obtain the representation of the original point cloud in the cylindrical coordinate system;

[0008] Expanding the cylindrical coordinate system to obtain a two-dimensional structure;

[0009] Regularization preprocessing is performed on the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure.

[0010] In one embodiment of the present invention, regularization preprocessing is performed on the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure, including:

[0011] The two-dimensional structure is adjusted according to a point-to-surface geometric distortion measure to obtain a regularized structure.

[0012] In one embodiment of the present invention, adjusting the two-dimensional structure according to a point-to-plane geometric distortion measure to obtain a regularized structure includes:

[0013] Find the point in the two-dimensional structure that is closest to the current point in terms of azimuth and elevation;

[0014] Construct a ray emitted from the origin through the angle information of the point closest to the current point;

[0015] constructing a plane based on the current point and its normal;

[0016] Find the intersection point of the ray and the plane, and record the distance from the origin to the intersection point;

[0017] Use this distance as the regularized radius of the current point from the center;

[0018] Repeat the above steps to complete the processing of all points and obtain the regularized structure of the original point cloud data.

[0019] In one embodiment of the present invention, performing regularization preprocessing on the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure further includes:

[0020] Adjusting the two-dimensional structure according to a point-to-point geometric distortion measure to obtain a regularized structure; or

[0021] Adjusting the two-dimensional structure according to the point-to-point and point-to-surface comprehensive distortion measures to obtain a regularized structure; or

[0022] The two-dimensional structure is adjusted according to a point-to-line geometric distortion measure to obtain a regularized structure.

[0023] Another embodiment of the present invention further provides a point cloud geometry encoding method, characterized by comprising:

[0024] Get original point cloud data;

[0025] The raw point cloud data is subjected to regularization preprocessing using the preprocessing method described in the above embodiment to obtain a regularized structure;

[0026] Determining a prediction mode for each point in the regularized structure, and performing geometric prediction on each point using the selected prediction mode to obtain information to be encoded;

[0027] The information to be encoded is encoded in sequence to obtain a geometric information code stream.

[0028] In one embodiment of the present invention, determining a prediction mode for each point in the regularized structure and performing geometric prediction on each point using the selected prediction mode to obtain information to be encoded includes:

[0029] Establish a prediction tree structure based on lidar calibration information;

[0030] Selecting a prediction mode for each point according to the prediction tree structure;

[0031] Performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode to obtain a geometric prediction residual of each point;

[0032] The geometric prediction residual is used as part of the information to be encoded.

[0033] In one embodiment of the present invention, performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode to obtain the geometric prediction residual of each point includes:

[0034] According to the current node type and the selected prediction mode, the cylindrical coordinates (r, j, i) of the current node are predicted to obtain the predicted value (r', j', i') and the prediction residual (r r ,r j ,r i ), where the predicted value j' of the azimuth of the current point is calculated according to the following formula:

[0035] j'=j prev +n;

[0036] Among them, j prev Indicates the predicted azimuth of the current point; n indicates the number of points that need to be skipped between the parent node and the current point according to the scanning speed, and its predicted residual for n' represents the number of nodes adjacent to the current point that need to be skipped.

[0037] According to the Cartesian coordinates (x, y, z) of the current point and the predicted Cartesian coordinates Perform differential prediction to obtain the prediction residual (r x ,ry ,r z ).

[0038] Another embodiment of the present invention further provides a point cloud geometry encoding device, comprising:

[0039] A first data acquisition module is used to acquire original point cloud data;

[0040] A regularization module performs regularization preprocessing on the original point cloud data to obtain a regularized structure;

[0041] A first prediction module is configured to determine a prediction mode for each point in the regularized structure and perform geometric prediction on each point using the selected prediction mode to obtain information to be encoded;

[0042] The encoding module is used to encode the information to be encoded in sequence to obtain a geometric information code stream.

[0043] Yet another embodiment of the present invention further provides a point cloud geometry decoding method, comprising:

[0044] Obtaining a geometric information code stream and decoding it to obtain decoded data; wherein the decoded data includes a prediction mode of a current node;

[0045] Performing geometric prediction on the current node according to the prediction mode to obtain a prediction residual; wherein the prediction residual includes a prediction residual in a cylindrical coordinate system and a prediction residual in a Cartesian coordinate system;

[0046] Reconstructing a prediction tree structure according to the prediction residual in the cylindrical coordinate system, and performing coordinate transformation on the points in the prediction tree structure to obtain the predicted Cartesian coordinates of the current point;

[0047] Point cloud reconstruction is performed based on the prediction residual in the Cartesian coordinate system and the predicted Cartesian coordinates to obtain reconstructed point cloud data.

[0048] Another embodiment of the present invention further provides a point cloud geometry decoding device, comprising:

[0049] A second data acquisition module is used to acquire a geometric information code stream and decode it to obtain decoded data; wherein the decoded data includes a prediction mode of a current node;

[0050] a second prediction module, configured to perform geometric prediction on the current node according to the prediction mode to obtain a prediction residual; wherein the prediction residual includes a prediction residual in a cylindrical coordinate system and a prediction residual in a Cartesian coordinate system;

[0051] A prediction tree reconstruction module is used to reconstruct a prediction tree structure according to the prediction residual in the cylindrical coordinate system, and perform coordinate transformation on points in the prediction tree structure to obtain the predicted Cartesian coordinates of the current point;

[0052] The point cloud reconstruction module reconstructs the geometric point cloud according to the geometric prediction value and prediction residual of the current node to obtain reconstructed point cloud data.

[0053] Beneficial effects of the present invention:

[0054] 1. The point cloud data processing method provided by the present invention performs regular pre-processing on the original input point cloud, so that the point cloud presents a regular distribution in both the horizontal and vertical directions, better reflecting the spatial correlation of the point cloud, so as to facilitate the subsequent further processing of the point cloud data;

[0055] 2. When performing regularization processing, the present invention uses geometric distortion measurement to adjust the two-dimensional structure to ensure the quality of the point cloud model;

[0056] 3. The point cloud geometry coding method provided by the present invention performs regularization on the azimuth direction of the point cloud during pre-processing of the point cloud data. This eliminates the need to use additional code streams to encode auxiliary information in the azimuth direction during encoding, thus saving code streams and improving encoding efficiency.

[0057] 4. When the present invention performs geometric coding on the point cloud after regularization processing, it effectively utilizes the regularization structure and combines the horizontal and vertical directions for predictive coding, thereby improving the geometric coding efficiency.

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of a point cloud data preprocessing method provided by an embodiment of the present invention;

[0060] Figure 2 Schematic diagram of the structure of a laser radar provided by an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the original collected data distribution structure provided by an embodiment of the present invention;

[0062] Figure 4 is an expanded diagram of a cylindrical coordinate system provided by an embodiment of the present invention;

[0063] Figure 5 This is a schematic diagram of point-to-surface interpolation processing provided by an embodiment of the present invention.

[0064] Figure 6 This is a before-and-after comparison diagram of the regularization processing provided by an embodiment of the present invention;

[0065] Figure 7 Schematic diagram of a point cloud geometry encoding method provided by an embodiment of the present invention;

[0066] Figure 8 Schematic diagram of the structure of a point cloud geometry encoding device provided by an embodiment of the present invention;

[0067] Figure 9 Schematic diagram of a point cloud geometry decoding method provided by an embodiment of the present invention;

[0068] Figure 10 It is a schematic structural diagram of a point cloud geometry decoding device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0070] Example 1

[0071] See Figure 1 , Figure 1 Schematic diagram of a point cloud data preprocessing method provided by an embodiment of the present invention, comprising:

[0072] S1: Perform coordinate transformation on the original point cloud data to obtain the representation of the original point cloud in the cylindrical coordinate system.

[0073] Specifically, the original point cloud data is usually composed of a set of three-dimensional spatial points, each of which records its own geometric position information, as well as additional attribute information such as color, reflectivity, and normal. Among them, the geometric position information of the point cloud is generally represented based on the Cartesian coordinate system, that is, it is represented by the x, y, and z coordinates of the point. The original point cloud data can be obtained through 3D scanning equipment such as lidar, or through public data sets provided by various platforms. In this embodiment, the geometric position information of the original point cloud data obtained is represented as (x, y, z) based on the Cartesian coordinate system. It should be noted that the method of representing the geometric position information of the original point cloud data is not limited to Cartesian coordinates.

[0074] Specifically, before performing coordinate transformation on the original point cloud, it may also be subjected to row quantization and reordering processing to facilitate subsequent predictive coding.

[0075] See Figure 2 , Figure 2 is a schematic diagram of the structure of a laser radar provided by an embodiment of the present invention; the laser radar is composed of multiple laser scanners. When describing a laser radar, the so-called "line number" refers to the number of laser scanners contained therein. These laser scanners are distributed on both sides of the central axis of the laser radar and have different pitch angles θi , thereby acquiring spatial information about different objects in the vertical direction. Each can be considered a relatively independent acquisition system. During acquisition, through the rotation of the base, all laser scanners acquire position information of objects in space at a certain sampling rate.

[0076] In this embodiment, the Cartesian coordinates (x, y, z) of the original point cloud data can be converted into cylindrical coordinates (r, φ, i) according to an existing conversion formula to obtain the representation of the original point cloud in the cylindrical coordinate system.

[0077] See Figure 3 , Figure 3 This is a schematic diagram of the distribution structure of raw data in a cylindrical coordinate system, as provided by an embodiment of the present invention. Ideally, point cloud data acquired by a LiDAR is uniformly distributed in azimuth and elevation in a cylindrical coordinate system. However, due to factors such as noise, mechanical vibration, and alignment of the LiDAR's internal coordinate system, point clouds can exhibit non-uniform distribution.

[0078] S2: Expand the cylindrical coordinate system to obtain a two-dimensional structure.

[0079] In this embodiment, the structure and acquisition parameters of the laser radar are first used to determine the pitch angle θ and azimuth angle of each point after regularization. Here, the pitch angle θ can be directly obtained from the vertical acquisition range of each laser scanner in the calibration file, and the azimuth angle Sampling interval required Sure.

[0080] Specifically, see Figure 4 , Figure 4 This is an expanded diagram of the cylindrical coordinate system provided by an embodiment of the present invention, wherein the vertical spacing and the horizontal spacing can be transformed into a concept similar to the resolution in an image. Therefore, the vertical resolution theta and the horizontal resolution phi are respectively:

[0081] theta=laserNum;

[0082]

[0083] S3: performing regularization preprocessing on the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure.

[0084] After the vertical resolution and the horizontal resolution are determined in step S2, it is also necessary to determine the radius r of each point from the center after regularization.

[0085] The choice of radius directly determines whether the regularized point cloud can maintain the same geometric structure as the input point cloud. For applications such as recognition and autonomous driving, geometric distortion directly determines the performance of such applications. Therefore, to minimize the geometric distortion of the model, this embodiment uses the concept of geometric distortion measurement to perform nearest neighbor interpolation to calculate the r component of the corresponding points after regularization, thereby ensuring that the distortion of the geometric reconstruction quality in D2 (point to plane) can be controlled within a certain range.

[0086] In this embodiment, the two-dimensional structure may be adjusted according to the point-to-surface geometric distortion measure (D2) to obtain a regularized structure.

[0087] See Figure 5 , Figure 5 is a schematic diagram of point-to-surface interpolation processing provided by an embodiment of the present invention. Specifically,

[0088] 1. First, find the point in the 2D structure that is closest to the current point in terms of azimuth and elevation.

[0089] 2. Construct a ray from the origin using the angle information of the point;

[0090] 3. According to the current point p i and its normal to construct the plane;

[0091] 4. Find the intersection of the ray and the plane, and record the distance from the origin to the intersection.

[0092] 5. Use this distance information as the regularized radius of the current point from the center, which is the r component in the cylindrical coordinates.

[0093] 6. Repeat the above steps to complete the processing of all points and obtain the regularized structure of the original point cloud data.

[0094] At this point, the regularization preprocessing of the cylindrical coordinates (r, φ, i) of the point cloud has been completed, and the regularized coordinates (r, j, i) have been obtained. Figure 6 , Figure 6 3 is a before-and-after comparison diagram of the regularization processing provided by an embodiment of the present invention.

[0095] The regularized structure constructed through the above steps in this embodiment can ensure that the distortion of each point to the plane is zero, maintain the geometric structure information of the original point cloud, and have little impact on the performance of applications such as recognition and autonomous driving. In addition, this regularized structure is extremely friendly to the further processing of subsequent point clouds.

[0096] In another embodiment of the present invention, the two-dimensional structure can also be adjusted according to the point-to-point geometric distortion measure (D1) to obtain a regularized structure, which can ensure both the geometric D1 distortion measure and the model quality of the point cloud.

[0097] In addition, the two-dimensional structure can be adjusted according to the comprehensive distortion measures of point-to-point (D1) and point-to-surface (D2) to obtain a regularized structure; this can simultaneously ensure the overall distortion measures of the geometric D1 and D2 and the quality of the point cloud model.

[0098] Furthermore, the 2D structure can be adjusted based on the point-to-line geometric distortion measure to obtain a regularized structure. The statistical point-to-line geometric distortion measure is between the point-to-point and point-to-surface distortion measures. This ensures the overall distortion of the geometric D1 and D2 as well as the model quality of the point cloud.

[0099] The point cloud data processing method provided by the present invention performs regular preprocessing on the original input point cloud, so that the point cloud presents a regular distribution in both horizontal and vertical directions, increases the correlation between the data, and facilitates subsequent further processing of the point cloud data. When performing regular processing, geometric distortion measurement is used to perform regular processing in the horizontal direction to ensure the quality of the point cloud model.

[0100] Example 2

[0101] See Figure 7 , Figure 7 Schematic diagram of a point cloud geometry encoding method provided by an embodiment of the present invention, comprising the following steps:

[0102] Step 1: Get the original point cloud data.

[0103] In this embodiment, the original point cloud data is represented by Cartesian coordinates (x, y, z).

[0104] Step 2: Perform regularization preprocessing on the original point cloud data to obtain a regularized structure.

[0105] Specifically, the preprocessing method provided in the first embodiment may be used to perform regularization preprocessing on the original point cloud data to obtain a regularized structure.

[0106] More specifically, after regularization, the cylindrical coordinates (x, y, z) are converted into a regular (r, j, i) structure.

[0107] Step 3: Determine the prediction mode for each point in the regularized structure, and use the selected prediction mode to perform geometric prediction on each point to obtain the information to be encoded.

[0108] First, a prediction tree structure is established based on the lidar calibration information.

[0109] Then, the prediction mode of the current point is selected according to the prediction tree structure.

[0110] In this embodiment, a prediction tree is established by traversing in depth-first order, and each node in the tree can only be predicted by its ancestors.

[0111] Furthermore, this embodiment sets the following four prediction modes:

[0112] Mode 0: No prediction

[0113] Mode 1: Delta prediction (ie p0)

[0114] Mode2: Linear prediction (i.e. 2p0-p1)

[0115] Mode3: Parallelogram prediction (i.e. p0+p1-p2)

[0116] Where p0, p1, and p2 are the positions of the parent node, grandparent node, and great-grandparent node of the current node, respectively. The best prediction model can be selected for the current node based on the reconstruction quality.

[0117] Next, geometric prediction is performed on each point in the prediction tree structure according to the selected prediction mode to obtain the geometric prediction residual of each point. In this embodiment, the geometric prediction includes cylindrical coordinate prediction and Cartesian coordinate prediction. The specific process is as follows:

[0118] According to the current node type and the selected prediction mode, the cylindrical coordinates (r, j, i) of the current node are predicted to obtain the predicted value (r', j', i') and the prediction residual (r r ,r j ,r i );

[0119] If the current node is the root node of the prediction tree, the prediction mode selects Mode0, that is, the cylindrical coordinates of the current node are not predicted, and the corresponding cylindrical coordinate prediction value is (r min ,j prev ,i prev ), where r min The minimum value of the r component after the coordinate transformation of the entire point cloud. If the current node has no parent node, then j prev ,i prev Set to 0, otherwise it is the cylindrical coordinate component of the parent node. The Cartesian coordinate prediction value is the inverse transformation of the cylindrical coordinates (r, j, i) of the point

[0120] If the current node is not the root node and the prediction mode is Mode1, the cylindrical coordinates of the current point are obtained by multiplying the cylindrical coordinates of its parent node (r min ,j prev ,i prev ) is used to predict the cylindrical coordinates of the current point (r', j', i'), and the Cartesian coordinates are obtained by inverse conversion of the original cylindrical coordinates (r, j, i)

[0121] If the current node is not the root node and the prediction mode is Mode2 or Mode3, the cylindrical coordinates of the current point are predicted by the corresponding prediction method, and the predicted value of the cylindrical coordinates of the current point is (r', j', i'). Similarly, the Cartesian coordinate prediction value is obtained by inverse conversion of the original cylindrical coordinates (r, j, i)

[0122] Using the best prediction mode of the current node, the cylindrical coordinates (r, j, i) of the current point are predicted to obtain the corresponding prediction residual (r r ,r j ,r i ).

[0123] It should be noted that the predicted value j' of the azimuth angle of the current point is calculated according to the following formula:

[0124] j'=j prev +n;

[0125] Among them, j prev Represents the predicted azimuth of the current point; n represents the number of points that need to be skipped between the parent node and the current point according to the scanning speed. If the Laser (laser scanner) of the current node is i, and the adjacent Laser of the current node is i+1 and has completed encoding and decoding, then the parameter n can be further used to perform differential prediction on the corresponding position node n' where Laser is i+1 to obtain the prediction residual of the number of points that need to be skipped, that is:

[0126] In addition, since in this embodiment, geometric prediction coding is performed based on a regularized structure, encoding and decoding are performed in sequence according to each Laser, it is necessary to temporarily store the node j component that has been encoded when Laser is i, so as to predict the node j component at the corresponding position when the encoding Laser is i+1~N.

[0127] Use the Cartesian coordinates (x, y, z) of the current point and the predicted Cartesian coordinates Perform differential prediction to obtain the prediction residual (r x ,r y ,r z ).

[0128] Finally, the prediction residual (r r ,r j ,r i ), prediction residuals in Cartesian coordinate system (r x ,r y ,r z ), and the prediction residuals for the number of points to be skipped Together with other parameters that need to be encoded, such as the number of child nodes of the current node, the prediction mode of the current node, and other information, one of them is used as the information to be encoded.

[0129] Step 4: Encode the information to be encoded in sequence to obtain the geometric information code stream.

[0130] Specifically, for each node's information to be encoded, first, the number of child nodes of the current node needs to be encoded, and then the prediction mode of the current node and the corresponding (r r ,r j ,r i ) and (r x ,r y ,r z ) prediction residuals and prediction residuals for points that need to be skipped

[0131] At this point, the geometric prediction coding of the point cloud is completed.

[0132] The point cloud geometric coding method provided in this embodiment performs regularization processing on the original input point cloud so that the point cloud presents a regular distribution in both the horizontal and vertical directions, thereby increasing the correlation between the data and improving the coding efficiency. At the same time, due to the regularization processing of the point cloud in the azimuth direction, there is no need to use an additional code stream to encode the auxiliary information in the azimuth direction during encoding, which saves the code stream and improves the coding efficiency.

[0133] When performing geometric coding on points after regularization processing, this embodiment effectively utilizes the regularization structure and combines horizontal and vertical directions for predictive coding, thereby improving the geometric coding efficiency.

[0134] In another embodiment of the present invention, step 3 may also use the existing geometric coding mode based on the prediction number to convert the Cartesian coordinates (x, y, z) into cylindrical coordinates. Then make a prediction and get the predicted value and and the prediction residual (r r ,r φ ,r i ) and (r x ,r y ,rz ), and uses lossless encoding to The component encodes the number of points n that are skipped, and the specific process is not described in detail here.

[0135] Example 3

[0136] On the basis of the above-mentioned embodiment 2, a mode switch may be further provided to guide whether to initialize the original point cloud data during the entire encoding process.

[0137] Specifically, the gps (Geometry Parameter Set Syntax) parameter set introduces the flag geom_enable_regular_flag, which is used to determine whether the regularized preprocessing scheme of the present invention is enabled in the entire encoding process. When geom_enable_regular_flag is 1, it is enabled; otherwise, it is disabled. See Table 1 for details.

[0138] When the regularized preprocessing scheme is enabled, the point cloud data can be preprocessed using the point cloud data preprocessing method provided in the above-mentioned embodiment 1, and then the point cloud data can be predictively encoded using the encoding scheme provided in embodiment 2 or the existing geometric prediction encoding method to improve encoding efficiency.

[0139] When the regularized pre-processing solution is not enabled, the prediction method provided in the above-mentioned second embodiment may also be directly used to predict and encode the original point cloud data.

[0140] Example 4

[0141] Based on the above-mentioned embodiment 2, this embodiment provides a point cloud geometry encoding device. Figure 8 , Figure 8 : is a schematic structural diagram of a point cloud geometry encoding device provided by an embodiment of the present invention, comprising:

[0142] A first data acquisition module 11 is used to acquire original point cloud data;

[0143] The regularization module 12 performs regularization preprocessing on the original point cloud data to obtain a regularized structure;

[0144] A first prediction module 13 is configured to determine a prediction mode for each point in the regularized structure and perform geometric prediction on each point using the selected prediction mode to obtain information to be encoded;

[0145] The encoding module 14 is used to encode the information to be encoded in sequence to obtain a geometric information code stream.

[0146] The device provided in this embodiment can implement the encoding method provided in the above-mentioned embodiment 2, and the specific implementation process will not be repeated here.

[0147] Example 5

[0148] See Figure 9 , Figure 9 Schematic diagram of a point cloud geometry decoding method provided by an embodiment of the present invention, including:

[0149] Step 1: Obtain the geometric information code stream and decode it to obtain decoded data; wherein the decoded data includes the prediction mode of the current node.

[0150] Step 2: Perform geometric prediction on the current node according to the prediction mode to obtain prediction residuals; wherein the prediction residuals include prediction residuals in a cylindrical coordinate system and prediction residuals in a Cartesian coordinate system.

[0151] According to the prediction mode adopted by the current point obtained by decoding and the reconstructed cylindrical coordinates of its ancestors, the cylindrical coordinates of the current point are predicted and the corresponding prediction values are generated. The specific process is as follows:

[0152] If the current node is the root node of the prediction tree, the prediction mode selects Mode0, that is, the cylindrical coordinates of the current node are not predicted, and the corresponding cylindrical coordinate prediction value is (r min ,j prev ,i prev ), where r min The minimum value of the r component after the coordinate transformation of the entire point cloud. If the current node has no parent node, then j prev ,i prev Set to 0, otherwise it is the cylindrical coordinate component of the parent node. The Cartesian coordinate prediction value is the inverse transformation of the original cylindrical coordinates (r, j, i) of the point

[0153] If the current node is not the root node and the prediction mode is Mode1, the cylindrical coordinates of the current point are obtained by multiplying the cylindrical coordinates of its parent node (r min ,j prev ,i prev ) is used to predict the cylindrical coordinates of the current point (r', j', i'), and the Cartesian coordinates are obtained by inverse conversion of the original cylindrical coordinates (r, j, i)

[0154] If the current node is not the root node and the prediction mode is Mode2 or Mode3, the cylindrical coordinates of the current point are predicted by the corresponding prediction method, and the predicted value of the cylindrical coordinates of the current point is (r', j', i'). Similarly, the Cartesian coordinate prediction value is obtained by inverse conversion of the original cylindrical coordinates (r, j, i)

[0155] Using the best prediction mode of the current node, the cylindrical coordinates (r, j, i) of the current point are predicted to obtain the corresponding prediction residual (r r ,r j ,r i ).

[0156] As with the encoding end, it should be noted that the predicted value j' of the azimuth angle of the current point is calculated according to the following formula:

[0157] j'=j prev +n;

[0158] Among them, j prev Indicates the predicted azimuth of the current point; n indicates the number of points that need to be skipped between the parent node and the current point according to the scanning speed. Note that if the laser of the current node is i, and the adjacent laser of the current node is i-1 and has completed encoding and decoding, then the parameter n is recovered using the corresponding position node n' of the laser i-1, that is:

[0159] At this point, the prediction residuals in the cylindrical coordinate system, the number of points to be skipped, and the prediction residuals in the Cartesian coordinate system are obtained.

[0160] Step 3: Reconstruct the prediction tree structure based on the prediction residual in the cylindrical coordinate system, and perform coordinate transformation on the points in the prediction tree structure to obtain the predicted Cartesian coordinates of the current point.

[0161] Specifically, the cylindrical coordinate residual (r r ,r j ,r i ) and the predicted cylindrical coordinates of the current point (r', j', i') to calculate the reconstructed cylindrical coordinates of the current point (r, j, i).

[0162] (r,j,i)=(r',j',i')+(r r ,r j ,r i ).

[0163] After obtaining the reconstructed cylindrical coordinates of the current point, the position of the current point in the prediction tree can be further determined based on the reconstructed cylindrical coordinates (r, j, i), thereby reconstructing the prediction tree.

[0164] Convert the reconstructed cylindrical coordinates (r, j, i) of the current point into Cartesian coordinates according to the following formula This is the predicted Cartesian coordinate of the current point.

[0165]

[0166]

[0167]

[0168]

[0169] Here, i is the LaserID corresponding to the point. The prior information of each Laser is different, that is, the elevation angle θ and the height zLaser in the vertical direction are different. Therefore, the elevation angle corresponding to the i-th Laser is θ(i), and the height in the vertical direction is zLaser(i).

[0170] Step 4: Reconstruct the point cloud based on the prediction residual and the predicted Cartesian coordinates in the Cartesian coordinate system to obtain reconstructed point cloud data.

[0171] The Cartesian coordinate residual (r x ,r y ,r z ) and the predicted Cartesian coordinates of the current point Compute the reconstructed Cartesian coordinates (x,y,z) of the current point.

[0172]

[0173] At this point, the decoding of the point cloud is completed, and the reconstructed voxelized point cloud is obtained.

[0174] Example 6

[0175] Based on the above-mentioned fifth embodiment, this embodiment provides a point cloud geometry decoding device. Figure 10 , Figure 10 : This is a schematic diagram of the structure of a point cloud geometry decoding device provided by an embodiment of the present invention, comprising:

[0176] The second data acquisition module 21 is used to acquire the geometric information code stream and decode it to obtain decoded data; wherein the decoded data includes the prediction mode of the current node;

[0177] The second prediction module 22 is configured to perform geometric prediction on the current node according to the prediction mode to obtain a prediction residual; wherein the prediction residual includes a prediction residual in a cylindrical coordinate system and a prediction residual in a Cartesian coordinate system;

[0178] The prediction tree reconstruction module 23 is used to reconstruct the prediction tree structure according to the prediction residual in the cylindrical coordinate system, and perform coordinate transformation on the points in the prediction tree structure to obtain the predicted Cartesian coordinates of the current point;

[0179] The point cloud reconstruction module 24 reconstructs the geometric point cloud according to the geometric prediction value and prediction residual of the current node to obtain reconstructed point cloud data.

[0180] The device provided in this embodiment can implement the decoding method provided in the fourth embodiment above, and the specific implementation process will not be repeated here.

[0181] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

[0182] Appendix 1: Geometry parameter set syntax

[0183]

[0184]

Claims

1. A point cloud data preprocessing method, characterized in that: include: Perform coordinate transformation on the original point cloud data to obtain the representation of the original point cloud in the cylindrical coordinate system; Expanding the cylindrical coordinate system to obtain a two-dimensional structure; Performing regularization preprocessing on the two-dimensional structure based on a geometric distortion measure to obtain a regularized structure; The performing regularization preprocessing on the two-dimensional structure based on the geometric distortion measure to obtain a regularized structure includes: Find the point in the two-dimensional structure that is closest to the current point in terms of azimuth and elevation; Construct a ray emitted from the origin through the angle information of the point closest to the current point; constructing a plane based on the current point and its normal; Find the intersection point of the ray and the plane, and record the distance from the origin to the intersection point; Use this distance as the regularized radius of the current point from the center; Repeat the above steps to complete the processing of all points and obtain the regularized structure of the original point cloud data.

2. The point cloud data preprocessing method according to claim 1, characterized in that: The performing regularization preprocessing on the two-dimensional structure based on the geometric distortion measure to obtain a regularized structure further includes: Adjusting the two-dimensional structure according to a point-to-point geometric distortion measure to obtain a regularized structure; or Adjusting the two-dimensional structure according to the point-to-point and point-to-surface comprehensive distortion measures to obtain a regularized structure; or The two-dimensional structure is adjusted according to a point-to-line geometric distortion measure to obtain a regularized structure.

3. A point cloud geometry encoding method, characterized in that: include: Get original point cloud data; Performing regularization preprocessing on the original point cloud data using the preprocessing method according to claim 1 or 2 to obtain a regularized structure; Determining a prediction mode for each point in the regularized structure, and performing geometric prediction on each point using the selected prediction mode to obtain information to be encoded; The information to be encoded is encoded in sequence to obtain a geometric information code stream.

4. The point cloud geometry encoding method according to claim 3, characterized in that: Determining a prediction mode for each point in the regularized structure and performing geometric prediction on each point using the selected prediction mode to obtain information to be encoded, including: Establish a prediction tree structure based on lidar calibration information; Selecting a prediction mode for each point according to the prediction tree structure; Performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode to obtain a geometric prediction residual of each point; The geometric prediction residual is used as part of the information to be encoded.

5. The point cloud geometry encoding method according to claim 4, characterized in that: Performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode to obtain a geometric prediction residual of each point, including: According to the current node type and the selected prediction mode, the cylindrical coordinates of the current node are Make a prediction and get the predicted value in the cylindrical coordinate system of the current node and prediction residuals ; Among them, the predicted value of the azimuth of the current point Calculate using the following formula: ； in, Indicates the predicted azimuth of the current point; Indicates the number of points that need to be skipped between the parent node and the current point according to the scanning speed, and its prediction residual for , Indicates the number of points that need to be skipped among the encoded nodes adjacent to the current point; According to the Cartesian coordinates of the current point with the predicted Cartesian coordinates Perform differential prediction to obtain the prediction residual in the Cartesian coordinate system .

6. A point cloud geometry encoding device, characterized in that: include: A first data acquisition module (11) is used to acquire original point cloud data; A regularization module (12), configured to perform regularization preprocessing on the original point cloud data using the preprocessing method according to claim 1 or 2 to obtain a regularized structure; A first prediction module (13) is used to determine a prediction mode for each point in the regularized structure and perform geometric prediction on each point using the selected prediction mode to obtain information to be encoded; The encoding module (14) is used to encode the information to be encoded in sequence to obtain a geometric information code stream.

7. A point cloud geometry decoding method, characterized in that: include: Obtaining a geometric information code stream and decoding it to obtain decoded data; wherein the decoded data includes a prediction mode of the current node; the geometric information code stream is obtained based on the point cloud geometry encoding method according to any one of claims 3 to 5; Performing geometric prediction on the current node according to the prediction mode to obtain a prediction residual; wherein the prediction residual includes a prediction residual in a cylindrical coordinate system and a prediction residual in a Cartesian coordinate system; Reconstructing a prediction tree structure according to the prediction residual in the cylindrical coordinate system, and performing coordinate transformation on the points in the prediction tree structure to obtain the predicted Cartesian coordinates of the current point; Point cloud reconstruction is performed based on the prediction residual in the Cartesian coordinate system and the predicted Cartesian coordinates to obtain reconstructed point cloud data.

8. A point cloud geometry decoding device, characterized in that: include: a second data acquisition module (21), configured to acquire a geometric information code stream and decode it to obtain decoded data; wherein the decoded data includes a prediction mode of a current node; and the geometric information code stream is obtained based on the point cloud geometric coding method according to any one of claims 3 to 5; A second prediction module (22) is used to perform geometric prediction on the current node according to the prediction mode to obtain a prediction residual; wherein the prediction residual includes a prediction residual in a cylindrical coordinate system and a prediction residual in a Cartesian coordinate system; A prediction tree reconstruction module (23) is used to reconstruct a prediction tree structure based on the prediction residual in the cylindrical coordinate system, and perform coordinate transformation on points in the prediction tree structure to obtain predicted Cartesian coordinates of the current point; The point cloud reconstruction module (24) reconstructs the geometric point cloud according to the geometric prediction value and the prediction residual of the current node to obtain reconstructed point cloud data.

Images