An inter-frame encoding and decoding method for compressing 3D sonar point cloud data
Through the octree encoding and attribute encoding methods, combined with motion search and Hoffman encoding, the problem of difficulty in data storage and transmission of three-dimensional sonar point clouds is solved, and efficient data compression and rapid encoding and decoding are achieved, suitable for undersea three-dimensional sound Insert point cloud data.
Patent Information
- Application Number
- CN202111622588.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-28
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-12-28
AI Technical Summary
The prior art is difficult to effectively compress and transmit three-dimensional sonar point cloud data, resulting in difficulty in storage and transmission, and the existing methods are not suitable for subsea three-dimensional sonar point cloud data.
The octree encoding and attribute encoding methods are used to inter-frame code the three-dimensional sonar point cloud data, and the three-dimensional sonar point cloud data is compressed through motion search algorithms and Hoffman encoding to achieve efficient data storage and transmission.
It realizes efficient compression of three-dimensional sonar point cloud data, ensures the restoration quality of data, is suitable for sparse three-dimensional sonar point cloud data, with high compression efficiency and fast encoding and decoding speed.
Smart Images

Figure CN114143556B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of three-dimensional point cloud coding and decoding, and in particular relates to an inter-frame coding and decoding method for compressing three-dimensional sonar point cloud data. Background Art
[0002] The three-dimensional sonar system transmits fixed-frequency sound waves through the sound wave transmitting array, and receives the returned sound wave signals in the transmitting frequency band through the receiving array. After a series of signal processing, the lower computer obtains the underwater three-dimensional sonar point cloud data, and transmits the data to the upper computer through the transmission protocol for real-time image display.
[0003] Since the amount of 3D sonar point cloud data obtained by the 3D sonar system after acquisition and signal processing is very large, and the engineering requirements require real-time display of images and storage of single-frame data, this leads to certain difficulties in data storage and data transmission. Therefore, a 3D sonar point cloud data inter-frame encoding and decoding method is needed to compress the 3D point cloud data and realize the large-scale storage and rapid transmission of 3D point cloud data.
[0004] In 2020, MPEG approved two coding standards for point cloud compression. The first one is called V-PCC, which is video-based point cloud compression, and the second one is called G-PCC, which is geometry-based point cloud compression. The former projects the 3D point cloud into a 2D video sequence with multiple faces, and then compresses it through traditional video compression methods. It is suitable for dense point cloud data. The latter uses octrees to divide the 3D point cloud bounding box, and then performs geometric entropy coding, which is suitable for sparse point cloud data.
[0005] Patent document CN111899152A discloses a point cloud data compression method based on projection and video stitching, including: analyzing the geometric attribute characteristics of point cloud data to determine the corresponding projection strategy and the optimal projection angle; projecting the projection area at the same projection angle multiple times to obtain a set of two-dimensional images with spatial correlation at the same angle; stitching a set of two-dimensional images into a video file. This method is suitable for dense point cloud data, but not for the compression of submarine three-dimensional sonar point cloud data.
[0006] Patent document CN107633539A discloses a three-dimensional point cloud model data compression method based on quadrilateral facet segmentation, including: segmenting the surface of the input point cloud model into a number of quadrilateral facets; parameterizing and resampling the quadrilateral facets in turn to obtain two-dimensional image data; resampling the facets regularly, and each facet can be described by a height map after resampling, so that the three-dimensional data is converted into two-dimensional image data, reducing the amount of data; using the EZW algorithm to perform wavelet encoding on the two-dimensional image data; achieving progressive compression of the three-dimensional point cloud model; decoding and model reconstruction. This method realizes compression of the three-dimensional point cloud by converting the three-dimensional point cloud into a two-dimensional image according to the image compression method, and this method is not suitable for compression of submarine three-dimensional sonar point cloud data. Summary of the invention
[0007] In view of the above, the purpose of the present invention is to provide an inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data, which is suitable for the characteristics of three-dimensional sonar point cloud data and has the advantages of high compression efficiency, fast encoding and decoding speed, and high data restoration quality.
[0008] To achieve the above-mentioned object of the invention, an embodiment provides an inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data, comprising the following steps:
[0009] For the current frame and reference frame of the three-dimensional sonar point cloud data, the current frame and the reference frame are octree-encoded, the reference frame is attribute-encoded, the occupancy code of each node in the octree is obtained, and the neighborhood-related entropy context of each node in the octree is obtained according to the occupancy code;
[0010] Calculate the voxel blocks and reference voxel blocks of each centroid of the point cloud in the current frame and the reference frame respectively, and use the motion search algorithm to search the reference frame for the reference voxel block corresponding to each centroid voxel block in the current frame according to the neighborhood correlation entropy context to form a matching voxel block pair;
[0011] For each matching voxel block pair, calculate the predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block, determine the predicted motion vectors and motion vector residuals of all voxel blocks of the point cloud where the centroid of the current frame is located based on the predicted motion vector, and calculate the attribute difference between each voxel block and the corresponding reference voxel block;
[0012] The octree coding of the reference frame is used as the octree coding of the current frame, and after attribute coding is performed on the predicted motion vector, motion vector residual and attribute difference of each voxel block of the current frame, Huffman coding is performed on the octree coding result and attribute coding result of the current frame;
[0013] During decoding, the octree coding result and attribute coding result of the reference frame are decoded to obtain the point positions and point attribute values in each reference voxel block of the reference frame, and the Huffman coding result of the current frame is decoded to obtain the actual motion vector and attribute difference. The actual motion vector and attribute difference are superimposed on the point positions and point attribute values in each reference voxel block of the reference frame to obtain the current frame.
[0014] In one embodiment, the three-dimensional sonar point cloud data is unstructured sparse point cloud data obtained after the data collected by the sonar receiving array is processed by signal processing, threshold filtering, and outlier removal steps, and the attribute is signal strength.
[0015] In one embodiment, in each node of the octree, the occupancy code is a flag code indicating whether each child node under the current node contains data, and the neighborhood-related entropy context is the occupancy code status of the nodes and their child nodes corresponding to the six cube spaces adjacent to the cube space corresponding to each node.
[0016] In one embodiment, the centroid of the point cloud in the current frame and the reference frame is determined by:
[0017] The center of mass is determined by taking the average of the sum of the coordinates of each point in each point cloud, or by taking the intensity of each point in each point cloud as a weight and taking the average of the weighted sum of the coordinates of each point to determine the center of mass.
[0018] In one embodiment, for the point cloud centroid that cannot be matched between the current frame and the reference frame, when the point cloud centroid is the point cloud centroid that disappears from the reference frame to the current frame, the predicted motion vector and the motion vector residual corresponding to the voxel block where the point cloud centroid is located are both 0, and the attribute difference is the point cloud centroid attribute of the reference frame.
[0019] In one embodiment, for the point cloud centroid that cannot be matched between the current frame and the reference frame, when the point cloud centroid is the point cloud centroid that is additional in the current frame compared to the reference frame, encoding is performed by selecting any blank leaf node in the octree structure of the reference frame to add a child node, wherein the predicted motion vector of the voxel block corresponding to each child node is the displacement vector of all points in the point cloud where the point cloud centroid is located in the current frame and each child node, the motion vector residual is 0, and the attribute difference is the negative attribute value of all points in the point cloud where the point cloud centroid is located in the current frame.
[0020] In one embodiment, the step of determining the predicted motion vectors and motion vector residuals of all voxel blocks of the point cloud where the centroid of the current frame is located according to the predicted motion vector includes:
[0021] The predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block is used as the predicted motion vector of all voxel blocks of the point cloud where the centroid of the current frame is located;
[0022] The difference between the predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block and the actual motion vectors of all voxel blocks in the point cloud where the centroid of the current frame is located is taken as the motion vector residual.
[0023] In one embodiment, the octree encoding result and attribute encoding result of the reference frame at the encoding end and the Huffman encoding result of the current frame are transmitted to the decoding end in the form of data stream;
[0024] The decoding end separates the received data stream into a reference frame code stream and a current frame code stream according to the resolution type, wherein the reference frame code stream includes the octree coding result and the attribute coding result of the reference frame, and the current frame code stream includes the Huffman coding result.
[0025] In one embodiment, during decoding, the Huffman coding result of the current frame is decoded to obtain a predicted motion vector, a motion vector residual, and an attribute difference;
[0026] According to the sum of the predicted motion vector and the motion vector residual as the actual motion vector of each point in each voxel block, the actual motion vector is superimposed on the position of each point in each reference voxel of the reference frame to obtain the position of each point in each voxel block of the current frame;
[0027] The attribute values of each point in each reference voxel of the reference frame are superimposed on the attribute difference to obtain the attribute values of each point in each voxel block of the current frame.
[0028] Compared with the prior art, the present invention has the following beneficial effects:
[0029] The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data provided in the embodiment compresses the three-dimensional sonar point cloud data by means of geometric coding and attribute coding divided by an octree structure. While ensuring a high degree of restoration of the three-dimensional sonar point cloud data, it greatly compresses the size of the three-dimensional sonar point cloud data during data storage and transmission, and has very important engineering use value. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0031] Figure 1 This is the system architecture of the three-dimensional sonar system provided by the embodiment.
[0032] Figure 2 It is a flowchart of an inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data provided by an embodiment. DETAILED DESCRIPTION
[0033] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.
[0034] Figure 1 This is the system architecture of the three-dimensional sonar system provided in the embodiment. Figure 1 As shown, the three-dimensional sonar system includes a lower computer and a host computer, wherein the sound wave transmitting array in the lower computer transmits sound waves of a fixed frequency, and receives the sound wave signals of the returned frequency band through the sound wave receiving array. After the sound wave signals undergo a series of signal processing, threshold filtering, and outlier removal in the FPGA, the three-dimensional sonar point cloud data of the underwater environment is obtained. After the three-dimensional sonar point cloud data is compressed, the compression coding result is transmitted to the host computer via UDP. After the compression coding result is decoded by the PC of the host computer, the three-dimensional point cloud data is obtained for real-time display of images and other operations.
[0035] Figure 2 1 is a flow chart of an inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data provided by an embodiment. Figure 2 As shown, the inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data provided by the embodiment includes an encoding operation at the encoding end and a decoding operation at the decoding end, wherein the encoding operation is performed at Figure 1 The FPGA implementation of the three-dimensional sonar system shown in the figure is used for decoding. Figure 1 The PC in the three-dimensional sonar system shown is implemented. The encoding operation and the decoding operation are described in detail below.
[0036] like Figure 2 As shown, the encoding operation includes:
[0037] S101, for a current frame and a reference frame of the three-dimensional sonar point cloud data, octree encoding is performed on the current frame and the reference frame, and attribute encoding is performed on the reference frame.
[0038] Most of the 3D sonar point cloud data obtained by the 3D sonar are echoes reflected by the water environment. The 3D sonar point cloud data is obtained after filtering the data below the threshold. The attribute of the 3D sonar point cloud data is represented by the signal strength. In the embodiment, the first frame of the collected adjacent multiple frames of data is used as a reference frame, and the subsequent frames are encoded with the first frame as the reference frame.
[0039] Since the three-dimensional sonar point cloud data is sparse, a geometry-based point cloud compression method is applicable. Therefore, in the embodiment, an octree encoding method is used to geometrically encode the reference frame and the current frame. When the current frame and the reference frame are divided into octrees, the geometric center of the three-dimensional sonar detection range is used as the origin, and the division is continued until the leaf node contains 1 or 0 data.
[0040] In the embodiment, attribute encoding is directly performed on the attribute value (signal strength value) of each point in the reference frame.
[0041] S102, performing octree encoding on the current frame and the reference frame, obtaining an occupancy code of each node in the octree, and obtaining a neighborhood-related entropy context of each node in the octree according to the occupancy code.
[0042] In the embodiment, after the current frame and the reference frame are octree-encoded, the data contained in the child nodes of each node are recorded, and if the child node contains data, the occupancy code is 1, otherwise it is 0, so that each node obtains an 8-bit occupancy code about its child node. Then, the neighborhood-related entropy context of each node in the octree is obtained according to the occupancy code, specifically including: for each node, the occupancy code states of the nodes and their child nodes corresponding to the six cubic voxel spaces adjacent to the cubic voxel space corresponding to each node are used as the neighborhood-related entropy context of each node.
[0043] S103, matching the point cloud centroids of the current frame and the reference frame according to the neighborhood correlation entropy context.
[0044] In an embodiment, the voxel blocks and reference voxel blocks of each centroid of the point cloud in the current frame and the reference frame are calculated respectively, and then, based on the neighborhood correlation entropy context, a motion search algorithm is used to search the reference frame for the reference voxel block corresponding to the voxel block where the centroid of the current frame is located. The voxel block and the corresponding reference voxel block form a matching voxel block pair to achieve matching of the point cloud centroid, and the centroids that can be successfully matched and the centroids that cannot be successfully matched are separated for different subsequent processing.
[0045] In an embodiment, the method for determining the center of mass of the point cloud in the current frame and the reference frame includes: determining the center of mass by taking the average of the coordinates of each point in each point cloud, or using the intensity of each point in each point cloud as a weight, taking the average of the weighted sum of the coordinates of each point and determining the center of mass.
[0046] It should be noted that, through the above point cloud determination method, it can be known that each point cloud corresponds to a centroid, and each point cloud will be divided into multiple voxel blocks when being voxelized.
[0047] S103, calculating the predicted motion vector, motion vector residual, attribute difference according to the matching result of the point cloud centroid, and constructing attribute data.
[0048] For the centroid of the point cloud that can be matched, that is, the obtained matching voxel block pair, the predicted motion vectors and motion vector residuals of all voxel blocks of the point cloud where the centroid of the current frame is located are used as the predicted motion vectors of all voxel blocks of the point cloud where the centroid of the current frame is located from the predicted motion vectors of the voxel block of the current frame to the reference voxel block of the reference frame. The specific process includes: using the predicted motion vectors of the voxel block where the centroid of the current frame is located to the corresponding reference voxel block as the predicted motion vectors of all voxel blocks of the point cloud where the centroid of the current frame is located; using the difference between the predicted motion vectors of the voxel block where the centroid of the current frame is located to the corresponding reference voxel block and the actual motion vectors of all voxel blocks of the point cloud where the centroid of the current frame is located as the motion vector residual. In the embodiment, the attribute difference between each voxel block of the current frame and the corresponding reference voxel block is calculated based on the attribute values of each point in the reference frame and the attribute values of each point in the current frame.
[0049] Since the predicted motion vectors of all voxel blocks in each point cloud are calculated through the voxel block where the centroid is located, it is only necessary to record one predicted motion vector data. At the same time, each point cloud generally represents an object. The relative error of the object as a whole during the movement will be very small. By recording the motion vector residual, the amount of data can be greatly reduced.
[0050] For the unsuccessful matching of the point cloud centroid, when the point cloud centroid is the point cloud centroid that disappears from the reference frame to the current frame, the predicted motion vector and the motion vector residual corresponding to the voxel block where the point cloud centroid is located are both 0, and the attribute difference is the point cloud data attribute of the reference frame.
[0051] For the point cloud centroid that is not matched successfully, when the point cloud centroid is the point cloud centroid that is more in the current frame than in the reference frame, the method of adding a child node to any blank leaf node in the octree structure of the reference frame is adopted for encoding, that is, this part of the point cloud data is added to the blank leaf node of the existing octree in the octree structure, that is, the original blank leaf node is changed into a parent node, and this part of the point cloud data is organized as a child node of the parent node. The predicted motion vector of the voxel block corresponding to each child node is the displacement vector of all points in the point cloud where the point cloud centroid is located in the current frame and each child node, the motion vector residual is 0, and the attribute difference is the negative attribute value of all points in the point cloud where the point cloud centroid is located in the current frame.
[0052] In the embodiment, the predicted motion vector, the motion vector residual and the attribute difference of each voxel block are combined into attribute data.
[0053] S104, after encoding the attribute data, perform Huffman encoding in combination with the geometric encoding result.
[0054] In the embodiment, the octree code of the reference frame is used as the octree code of the current frame, and the predicted motion vector, motion vector residual and attribute difference of each voxel block of the current frame are Huffman encoded, and the Huffman encoding result is uploaded to the host computer in the form of a data stream.
[0055] In the embodiment, most of the point cloud centroids that can be matched in the current frame and the reference frame satisfy similar motion relationships and the attribute values do not change strongly. Therefore, for a large amount of repeated data or data close to 0 generated by encoding the current frame using predicted motion vectors, motion vector residuals, and attribute differences, Huffman coding is used to compress the data.
[0056] In the embodiment, the octree encoding result and the attribute encoding result of the reference frame are also uploaded to the host computer in the form of data stream.
[0057] like Figure 2 As shown, the decoding operation includes:
[0058] S201, obtaining a data stream, and separating a reference frame code stream and a current frame code stream according to a resolution type.
[0059] In the embodiment, the obtained reference frame code stream includes an octree coding result and an attribute coding result of the reference frame, and the obtained current frame code stream includes a Huffman coding result.
[0060] S202, performing octree decoding and attribute decoding processing on the reference frame code stream.
[0061] In the embodiment, octree decoding is performed on the octree encoding result in the reference frame code stream to obtain the point position in each reference voxel block of the reference frame, and attribute decoding is performed on the encoding result in the reference frame code stream to obtain the point attribute information in each reference voxel block of the reference frame.
[0062] S203, performing Huffman decoding processing on the current frame code stream.
[0063] Decode the Huffman coding result of the current frame to obtain the predicted motion vector, motion vector residual and attribute difference.
[0064] S204, obtaining current frame data by vector superposition according to the decoding results of S202 and S203.
[0065] In the embodiment, the actual motion vector of each point in each voxel block is taken as the sum of the predicted motion vector and the motion vector residual, and the actual motion vector is superimposed on the position of each point in each reference voxel of the reference frame to obtain the position of each point in each voxel block of the current frame;
[0066] The attribute values of each point in each reference voxel of the reference frame are superimposed on the attribute difference to obtain the attribute values of each point in each voxel block of the current frame.
[0067] The inter-frame encoding and decoding method for compressing three-dimensional point cloud data provided by the above embodiment has the characteristics of high compression efficiency, fast encoding and decoding speed, and high data restoration quality. The method is suitable for the compression of spatially sparse three-dimensional point cloud data, and has high compression efficiency. The method can be applied to the encoding and decoding of most spatially sparse three-dimensional point cloud data, which is conducive to the storage and efficient transmission of three-dimensional point clouds.
[0068] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. An inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data, characterized in that: The following steps are involved: For the current frame and reference frame of the three-dimensional sonar point cloud data, the current frame and the reference frame are octree-encoded, the reference frame is attribute-encoded, the occupancy code of each node in the octree is obtained, and the neighborhood-related entropy context of each node in the octree is obtained according to the occupancy code; Calculate the voxel blocks and reference voxel blocks of each centroid of the point cloud in the current frame and the reference frame respectively, and use the motion search algorithm to search the reference frame for the reference voxel blocks corresponding to each centroid of the voxel block in the current frame according to the neighborhood correlation entropy context, to form a matching voxel block pair, and for the point cloud centroids that cannot be matched between the current frame and the reference frame, when the point cloud centroid is the point cloud centroid that is extra in the current frame compared with the reference frame, select any blank leaf node in the octree structure of the reference frame to add a child node for encoding, where the predicted motion vector of the voxel block corresponding to each child node is the displacement vector of all points in the point cloud where the point cloud centroid of the current frame is located and each child node, the motion vector residual is 0, and the attribute difference is the negative attribute value of all points in the point cloud where the point cloud centroid of the current frame is located; For each matching voxel block pair, a predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block is calculated, and the predicted motion vectors and motion vector residuals of all voxel blocks of the point cloud where the centroid of the current frame is located are determined based on the predicted motion vector, and the attribute difference between each voxel block and the corresponding reference voxel block is calculated, wherein the predicted motion vectors and motion vector residuals of all voxel blocks of the point cloud where the centroid of the current frame is located are determined based on the predicted motion vector, including: taking the predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block as the predicted motion vector of all voxel blocks of the point cloud where the centroid of the current frame is located; taking the difference between the predicted motion vector from the voxel block where the centroid of the current frame is located to the corresponding reference voxel block and the actual motion vectors of all voxel blocks of the point cloud where the centroid of the current frame is located as the motion vector residual; The octree code of the reference frame is used as the octree code of the current frame, and the predicted motion vector, motion vector residual and attribute difference of each voxel block of the current frame are Huffman coded; During decoding, the octree coding result and the attribute coding result of the reference frame are decoded to obtain the point positions and point attribute values in each reference voxel block of the reference frame, the Huffman coding result of the current frame is decoded to obtain the actual motion vector and the attribute difference, and the actual motion vector and the attribute difference are superimposed on the point positions and the point attribute values in each reference voxel block of the reference frame to obtain the current frame, including: decoding the Huffman coding result of the current frame to obtain the predicted motion vector, the motion vector residual and the attribute difference; taking the sum of the predicted motion vector and the motion vector residual as the actual motion vector of each point in each voxel block, superimposing the actual motion vector on the position of each point in each reference voxel of the reference frame to obtain the position of each point in each voxel block of the current frame; superimposing the attribute difference on the attribute value of each point in each reference voxel of the reference frame to obtain the attribute value of each point in each voxel block of the current frame.
2. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: The three-dimensional sonar point cloud data is unstructured sparse point cloud data obtained after the data collected by the sonar receiving array is processed by signal processing, threshold filtering, and outlier removal steps, and the attribute is signal strength.
3. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: In each node of the octree, the occupancy code is a flag code indicating whether each child node under the current node contains data, and the neighborhood-related entropy context is the occupancy code status of the nodes and their child nodes corresponding to the six cube spaces adjacent to the cube space corresponding to each node.
4. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: The centroid of the point cloud in the current frame and the reference frame is determined by: The center of mass is determined by taking the average of the sum of the coordinates of each point in each point cloud, or by taking the intensity of each point in each point cloud as a weight and taking the average of the weighted sum of the coordinates of each point to determine the center of mass.
5. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: For the point cloud centroid that cannot be matched between the current frame and the reference frame, when the point cloud centroid is the point cloud centroid that disappears from the reference frame to the current frame, the predicted motion vector and the motion vector residual corresponding to the voxel block where the point cloud centroid is located are both 0, and the attribute difference is the point cloud centroid attribute of the reference frame.
6. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: The octree encoding result and attribute encoding result of the reference frame at the encoding end and the Huffman encoding result of the current frame are transmitted to the decoding end in the form of data stream; The decoding end separates the received data stream into a reference frame code stream and a current frame code stream according to the resolution type, wherein the reference frame code stream includes the octree coding result and the attribute coding result of the reference frame, and the current frame code stream includes the Huffman coding result.
7. The inter-frame encoding and decoding method for compressing three-dimensional sonar point cloud data according to claim 1, characterized in that: The first frame among the collected adjacent multiple frames of data is used as a reference frame, and subsequent frames use the first frame as a reference frame for current frame encoding.
Citation Information
Patent Citations
Three-dimensional point cloud model data compression method based on four-side patch segmentation
CN107633539A
Point cloud data compression method and system based on projection and video stitching
CN111899152A
Depth map sequence fractal coding method based on motion vectors of color video
CN103581647A
Point cloud geometric coding method, decoding method and equipment
CN113284250A
Point cloud data transmission device, point cloud data transmission method, point cloud data reception device, and point cloud data reception method
CN113615204A
