CN115496818B - Semantic graph compression method and device based on dynamic object segmentation - Google Patents
Semantic graph compression method and device based on dynamic object segmentation Download PDFInfo
- Publication number
- CN115496818B CN115496818B CN202211390992.4A CN202211390992A CN115496818B CN 115496818 B CN115496818 B CN 115496818B CN 202211390992 A CN202211390992 A CN 202211390992A CN 115496818 B CN115496818 B CN 115496818B
- Authority
- CN
- China
- Prior art keywords
- semantic
- dynamic object
- semantic graph
- graph
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/70—Labelling scene content, e.g. deriving syntactic or semantic representations
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a semantic graph compression method and a device based on dynamic object segmentation, wherein the method comprises the steps of firstly dividing a simulation scene into a static background and a dynamic object, and drawing the simulation scene to obtain a semantic graph; segmenting all dynamic objects in the semantic graph to obtain semantic subgraphs, and filling the static background semantic graph with adjacent pixels of the semantic subgraphs of the dynamic objects; and finally, coding all dynamic object semantic subgraphs and the filled static background semantic graph respectively by using a coding algorithm. The method separates the static background from the dynamic object, and splits a semantic graph into a static background semantic graph and a plurality of dynamic object semantic subgraphs, thereby reducing the pixel mutation in the semantic graph, increasing the continuity of data distribution and obviously improving the compression ratio of semantic graph coding.
Description
Technical Field
The invention relates to the field of image data compression, in particular to a semantic graph compression method and device based on dynamic object segmentation.
Background
As a core technology of computer vision and image understanding, an object detection algorithm is an important research direction in the field of computer vision and is also the basis of other complex visual tasks. The application scenes of the target detection algorithm comprise image description, scene understanding, image segmentation, target tracking, event detection and the like, and the target detection algorithm is also widely applied to judgment scenes such as automatic driving, medical images and unmanned aerial vehicle navigation.
The target detection algorithm based on machine learning needs to provide a large amount of truth data in the training process, and the truth data mainly comprises a semantic segmentation graph, namely a semantic graph. The semantic graph divides a picture or a video into a plurality of blocks according to the difference of the categories, so as to realize the classification of the pixel level. The semantic graph truth value is generally generated by adopting manual labeling or a simulation program automatically. The simulation algorithm has a high degree of customization and can generate massive data quickly, so that the simulation algorithm becomes one of important methods for generating a truth value data set of a semantic map. The traditional simulation semantic graph is not generally coded or directly coded by a compression algorithm such as run-length coding, and the characteristics of a simulation scene are not fully considered, so that the compression rate is low, and the efficiency of storage and data transmission is limited.
Disclosure of Invention
In order to solve the defects of the prior art and realize more efficient semantic graph compression, the invention adopts the following technical scheme:
a semantic graph compression method based on dynamic object segmentation comprises the following steps:
s1, initializing a simulation scene, wherein the simulation scene consists of a static background and a dynamic object;
s2, updating and drawing the simulation scene to obtain a semantic graph and two-dimensional bounding volumes of all dynamic objects under a semantic graph coordinate system;
s3, segmenting the semantic data of the dynamic object by using the two-dimensional bounding volume to form a plurality of dynamic object semantic subgraphs; filling the corresponding image area by using the adjacent pixels of each dynamic object semantic sub-image in the rest static background semantic graph;
and S4, respectively coding all dynamic object semantic subgraphs and the filled static background semantic graphs by using a coding algorithm.
Furthermore, all objects of the simulation scene in the S1 are assigned with an ID, and the IDs of the objects with the same semantic meaning are the same; each ID uniquely corresponds to one color, and different IDs correspond to different colors.
Further, updating the simulation scene in S2 includes updating the poses of all dynamic objects and rendering views; each pixel in the semantic graph uniquely corresponds to one object, and the color corresponding to the object ID is used for coloring.
Further, in S2, the two-dimensional bounding volume of the dynamic object in the semantic graph coordinate system is to contain all pixels of the dynamic object on the semantic graph.
Further, in S3, the semantic data of the dynamic object is segmented by using the two-dimensional bounding volume in S2 to form a plurality of semantic subgraphs, which is specifically realized by the following sub-steps:
(1) Initializing corresponding semantic subgraph data according to the size of a two-dimensional bounding volume of each dynamic object, wherein each element of the semantic subgraph is initialized to be (R: 0, G:0, B: 0);
(2) Traversing each pixel of the semantic graph, and for each pixel, performing the following processing:
and judging whether the pixel coordinate of the pixel is positioned in the range of a two-dimensional surrounding body of a certain dynamic object, if so, calculating the relative coordinate of the pixel under the semantic sub-image coordinate system of the dynamic object, writing the relative coordinate into a corresponding element of the semantic sub-image of the dynamic object, otherwise, ignoring the pixel, and continuously traversing the next pixel to obtain the semantic sub-image of the certain dynamic object.
Further, in S3, the remaining static background semantic graphs are obtained by filling corresponding image regions with pixels adjacent to each dynamic object semantic sub-graph, through the following sub-steps:
(1) Searching edge pixels around the dynamic object semantic subgraph, and if the edge pixels exceed the range of the semantic graph, determining the edge pixels to be invalid;
(2) And counting all effective edge pixels, and filling an image area corresponding to the dynamic object semantic sub-image in the semantic graph according to the values of the edge pixels.
Further, in S4, the encoding result of each dynamic object semantic subgraph is accompanied by corresponding two-dimensional bounding volume information.
The device comprises one or more processors and is used for realizing the semantic graph compression method based on the dynamic object segmentation.
A computer-readable storage medium, on which a program is stored which, when executed by a processor, implements a semantic graph compression method based on dynamic object segmentation.
The invention has the following beneficial effects:
the invention discloses a semantic graph compression method based on dynamic object segmentation, which comprises the steps of firstly dividing a simulation scene into a static background and a dynamic object, and drawing the simulation scene to obtain a semantic graph; segmenting all dynamic objects in the semantic graph to obtain semantic subgraphs, and filling the static background semantic graph with adjacent pixels of the semantic subgraphs of the dynamic objects; and finally, coding all dynamic object semantic subgraphs and the filled static background semantic graphs respectively by using a coding algorithm. The method separates the static background from the dynamic object, and splits a semantic graph into a static background semantic graph and a plurality of dynamic object semantic subgraphs, thereby reducing the pixel mutation in the semantic graph, increasing the continuity of data distribution and obviously improving the compression ratio of semantic graph coding.
Drawings
FIG. 1 is a diagram of steps of a semantic graph compression method based on dynamic object segmentation in an exemplary embodiment.
FIG. 2 is a diagram of a static background and dynamic objects of a simulation scene in an exemplary embodiment.
FIG. 3 is a diagram of semantic graph dynamic object segmentation and compression in an exemplary embodiment.
FIG. 4 is a schematic diagram of a semantic graph compression apparatus based on dynamic object segmentation in an exemplary embodiment.
Detailed Description
For the purposes of promoting an understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description of the invention, taken in conjunction with the accompanying drawings and examples, it being understood that the specific embodiments described herein are illustrative of the invention and are not intended to be exhaustive. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the scope of the present invention.
In one embodiment, as shown in fig. 1, a semantic graph compression method and apparatus based on dynamic object segmentation is provided, in which a simulation scene is first divided into two parts, namely a static background and a dynamic object, and the simulation scene is drawn to obtain a semantic graph; segmenting all dynamic objects in the semantic graph to obtain semantic subgraphs, and filling the static background semantic graph with adjacent pixels of the semantic subgraphs of the dynamic objects; and finally, coding all dynamic object semantic subgraphs and the filled static background semantic graph respectively by using a coding algorithm.
The method specifically comprises the following steps:
step 1, initializing a simulation scene, wherein the simulation scene is composed of a static background and a dynamic object.
As shown in fig. 2, in the present embodiment, the simulation scene is a digital twin scene of a certain park, and is composed of several types of models, such as an oblique photography model, an art handmade model, and a procedural generation model, which are classified into two types, namely a static background and a dynamic object. The static background comprises sky, terrain, road surface, road teeth, street lamps, buildings, plants and the like, and the dynamic object comprises a non-motor vehicle, a pedestrian and the like.
The image rendering interface of the simulation program adopts an Open-source Graphics rendering Library OpenGL (Open Graphics Library), which is a cross-language, cross-platform Application Programming Interface (API) for rendering 2D and 3D vector Graphics. The OpenGL interface is typically used to interact with a Graphics Processing Unit (GPU) to implement hardware-accelerated rendering, which includes nearly 350 different function call components to draw from simple graphics bits to complex three-dimensional scenes; openGL is commonly used in CAD, virtual reality, scientific visualization programs, and electronic game development, and contains 7 major functions: establishing a 3D model, graph transformation, a color mode, illumination and material setting, texture mapping, an image enhancement function, an extended function of bitmap display and a double-cache function.
All objects of the simulation scene are assigned with an ID, and the IDs of the objects with the same type of semantics are the same, wherein the same type of semantics refers to the objects belonging to the same type in a target detection algorithm; each ID uniquely corresponds to one color, and different IDs correspond to different colors. The correspondence between the IDs and the colors of the different objects is shown in table 1.
And a semantic graph simulation camera is also placed in the simulation scene, and a semantic graph is generated by a scene drawing method. Setting the field angle of the semantic graph simulation camera to be 72 degrees, the aspect ratio to be 16, and the resolution of the target image to be 1280 × 720. The semantic graph simulation camera also comprises a motion control module which controls the motion logic of the semantic graph simulation camera in a scene: a motion track is preset, each point of the motion track comprises a group of pose data, the semantic graph simulation camera moves at a constant speed according to the preset speed along the motion track, and a semantic graph is generated by drawing at fixed time intervals.
TABLE 1 correspondence of ID and color of different objects
And 2, updating and drawing the simulation scene to obtain a semantic graph and two-dimensional bounding volumes of all dynamic objects under a semantic graph coordinate system.
The two-dimensional enclosure may be a two-dimensional enclosure frame, such as a rectangular frame, or may be other regular shapes, such as a circle, an ellipse, etc., or other irregular shapes.
Semantic segmentation is an important ring in image processing and computer vision techniques with respect to image understanding. The semantic segmentation classifies each pixel point in the image, determines the category of each point, such as the category of the background, the edge or the body, and separates the example segmentation. Semantic segmentation does not separate instances of the same class; it is only the class of each pixel that is of interest, and if there are two objects of the same class in the input object, the segmentation itself will not distinguish them as separate objects. The semantic segmentation algorithm outputs a semantic segmentation image, which is called a semantic image for short.
In the embodiment, the semantic graph divides one picture or video into a plurality of blocks according to the difference of the categories, so as to realize the classification of the pixel level; each pixel in the semantic graph obtained by drawing uniquely corresponds to an object, and the color corresponding to the object ID is used for coloring.
In the embodiment, the updating of the simulation scene comprises the pose updating of all dynamic objects and semantic graph simulation cameras, and the updating frequency is set to be 60HZ; all dynamic objects including non-motor vehicles, motor vehicles and pedestrians are skeleton skin models, and the pose updating comprises the following two steps:
(1) Receiving the state data of the dynamic object from the server, analyzing the position and rotation data in the state data, and setting the state data of the dynamic object to a corresponding dynamic object according to the object identifier character string;
(2) And performing animation skinning calculation based on the GPU according to the current simulation time and the animation state of the dynamic object, and updating each vertex of the dynamic object model.
The semantic graph simulation camera updates the position and the posture of the semantic graph simulation camera by sampling a preset motion track, and the method comprises the following specific steps of:
(1) Simulating the motion speed of a camera according to the current simulation time and a preset semantic graph, and calculating the path length of the current motion;
(2) If the path length is larger than the preset motion track length, stopping the simulation process and exiting the program; otherwise, sampling a preset motion track by using the path length, and setting the obtained position and posture to the semantic graph simulation camera.
And starting to execute the drawing operation after the poses of all the dynamic objects and the semantic graph simulation cameras are updated. Before drawing each scene object, calling a glUniform1i function to transfer the ID of the object as a uniform variable to a pixel shader, and calling a glDrawElements function to draw vertex data of the model. In the pixel shader, the calculation formula of the output pixel color PixelColor and the object ID is as follows:
PixelColor.r=ID*20
PixelColor.g=ID*20
PixelColor.b=ID*20
in addition to obtaining drawn semantic graph data, the embodiment also generates a two-dimensional bounding box of all dynamic objects in a semantic graph coordinate system, and an area covered by the two-dimensional bounding box includes all pixels of the dynamic objects on the semantic graph; for each dynamic object, the specific generation algorithm of the two-dimensional bounding box is as follows:
(1) Traversing all vertexes of the dynamic object, respectively calculating the maximum value and the minimum value in three axial directions, and constructing an axially Aligned three-dimensional Bounding Box (AABB, axis-Aligned Bounding Box), wherein the three-dimensional Bounding Box can contain all vertexes of the dynamic object;
(2) Traversing 8 three-dimensional corner points of the three-dimensional bounding box, and transforming the three-dimensional corner points into two-dimensional corner points under a semantic map coordinate system through a model matrix, a view matrix, a perspective matrix and a viewport transformation matrix in sequence;
(3) Traversing the two-dimensional angular points obtained in the last step, respectively calculating the maximum value and the minimum value in two axial directions, and constructing an axially aligned two-dimensional surrounding frame containing all the two-dimensional angular points, wherein the two-dimensional surrounding frame is the two-dimensional surrounding frame of the dynamic object.
Step 3, segmenting the semantic data of the dynamic object by using the two-dimensional bounding volume in the step 2 to form a plurality of semantic subgraphs; and filling the corresponding image area by using adjacent pixels of each dynamic object semantic sub-image in the rest static background semantic graph.
In this embodiment, as shown in fig. 3, the two-dimensional bounding box of the dynamic object is used to segment semantic data of all dynamic objects from the original semantic graph to form a semantic subgraph; storing the semantic graph data and the semantic subgraph data by using a two-dimensional array, wherein elements of the two-dimensional array are RGB color values of pixels;
initializing corresponding semantic subgraph data according to the size of a two-dimensional bounding volume of each dynamic object, wherein each element of the semantic subgraph is initialized to be (R: 0, G:0, B: 0); traversing each pixel of the original semantic graph, and performing the following processing for each pixel:
judging whether the pixel coordinate of the pixel is located in the range of a two-dimensional surrounding body of a certain dynamic object, if so, calculating the relative coordinate of the pixel under the semantic subgraph coordinate system of the dynamic object, and writing the relative coordinate into the corresponding element of the semantic subgraph of the dynamic object; otherwise, the pixel is ignored, and the next pixel is continuously traversed to obtain the semantic subgraph of a certain dynamic object.
For each dynamic object semantic sub-graph, searching all edge pixels of a circle of the dynamic object semantic sub-graph around the original semantic graph, and if the edge pixels exceed the range of the semantic graph, determining that the edge pixels are invalid; and selecting the mode of all effective edge pixels as specified adjacent pixels, and filling the corresponding image area of the dynamic object semantic subgraph in the original semantic graph with the adjacent pixels. Wherein, the edge pixel can be one circle or a plurality of circles; in the case of specifying adjacent pixels, the average number and the median number of all effective edge pixels may be selected in addition to the mode of all effective edge pixels.
In this embodiment, if no dynamic object exists in the simulation scene, the above steps of dynamic object segmentation and original semantic graph filling are not performed.
And 4, respectively coding all the dynamic object semantic subgraphs and the filled static background semantic graphs by using a coding algorithm.
In the embodiment, all dynamic object semantic subgraphs and the filled static background semantic graph are compressed and encoded by using the stroke coding; run-Length Encoding (RLE) is a form of lossless data compression in which runs of data (sequences of identical data values occurring in many consecutive data elements) are stored as a single data value and count, rather than as original runs. This is most effective for data containing many such runs, such as simple graphical images, like icons, line drawings and animations. For files without too many repeated characters, run length encoding may instead increase the file size.
In addition, the run-length coding is sensitive to transmission errors, and if one bit of the code is wrong, the correctness of the whole coding sequence is affected, so that the run-length coding cannot restore the original data, and therefore, the errors are generally controlled within one row and one column by using a row synchronization and column synchronization method. In summary, run-length coding is well suited for compressing highly repetitive computer-generated images, especially binary images, and is also suitable for image data such as semantic maps.
All dynamic object semantic subgraphs and the filled static background semantic graphs are compressed by using stroke coding, and the coding and compression processes are consistent; here, taking the filled static background semantic graph (background semantic graph for short) as an example, the compression process is described as follows:
creating a variable Header, and initializing the variable Header to be the first pixel of the background semantic graph; creating a variable Count, and initializing to 0; creating an empty one-dimensional array for storing the compressed result data, wherein the array elements are 24-bit integer values; traversing each pixel of the background semantic graph, and executing the following operations for each pixel: judging whether the color value of the current pixel is the same as the Header or not, and if so, adding one to the Count; otherwise, writing the current Count value into the compression result array, combining the RGB three channel values of the color value of the Header into an integer, writing the integer into the compression result array, assigning the color value of the current pixel to the Header, resetting the Count value to 0, and continuously traversing the next pixel.
The compression result of the dynamic object semantic subgraph needs to store two-dimensional bounding volume information of the dynamic object semantic subgraph on the original semantic graph besides the compressed data of the run-length coding. The resulting compression result data includes three parts:
(1) A stroke coding compression array of the background semantic graph;
(2) Compressing arrays of the stroke codes of all dynamic object semantic subgraphs;
(3) And (3) two-dimensional bounding volume information of all dynamic object semantic subgraphs.
In this embodiment, the complete compression result data is transmitted to the algorithm end through the network, and the algorithm end restores the original semantic graph through the corresponding decompression method: firstly, decompressing background semantic graph data according to a stroke coding compression array of a background semantic graph, decompressing all dynamic object semantic sub-graph data according to the stroke coding compression array of all dynamic object semantic sub-graphs, and finally covering all dynamic object semantic sub-graph data to corresponding positions of the background semantic graph data according to two-dimensional bounding volume information of all dynamic object semantic sub-graphs to obtain original semantic graph data;
the specific method for decompressing the original data according to the run length coding compression array is as follows:
(1) Creating a variable Count, and initializing to 0; reading the first array element and assigning a value to the Count;
(2) Reading subsequent Count array elements, splitting each array element into RGB three channel values to form a pixel color value, and writing the pixel color value into original data;
(3) If all the array elements have been traversed, the decompression step is completed, and the program is exited; otherwise, reading the next array element, assigning value to Count, and continuing to execute the step 2.
After the decompression of all the dynamic object semantic subgraphs and the background semantic graph is finished, the final semantic graph splicing step is executed: traversing the two-dimensional bounding volume information of all the dynamic object semantic subgraphs, traversing all the pixels of the dynamic object semantic subgraph corresponding to the current two-dimensional bounding volume, transforming the coordinates of the current pixels in the dynamic object semantic subgraph to the coordinates of the background semantic graph by using the two-dimensional bounding volume, and writing the current pixels into the pixels at the corresponding positions of the background semantic graph.
The application effect of the final semantic graph compression method based on dynamic object segmentation is shown in table 2.
Table 2 shows the compression ratio comparison of image compression using the conventional three methods and the method of the present invention
| RLE | TRLE | FastPFor | Ours | |
| Compression ratio | 22.4% | 14.6% | 9.3% | 5.9% |
Table 2 compares the compression ratios of four Encoding methods, i.e., the conventional Run-Length Encoding (RLE), the Turbo Run-Length Encoding (TRLE), the fast integer sequence compression (FastPFor), and the semantic graph compression based on dynamic object segmentation. It can be seen that, because the characteristics of the simulation scene are fully utilized, the moving objects and the static objects are separated, the continuity of the semantic graph data is maximized so as to facilitate the subsequent compression and encoding, and the compression ratio of the semantic graph compression method based on the dynamic object segmentation is the lowest, thereby embodying the beneficial effects of the invention.
Corresponding to the foregoing embodiment, the present invention further provides an embodiment of a semantic graph compression apparatus based on dynamic object segmentation, as shown in fig. 4, the apparatus includes one or more processors, and is configured to implement the above laser radar point cloud generation method.
The semantic graph compressing device based on dynamic object segmentation of the present invention can be applied to any device with data processing capability, such as a computer or other devices. The apparatus embodiments may be implemented by software, or by hardware, or by a combination of hardware and software.
The software implementation is taken as an example, and as a device in a logical sense, a processor of any device with data processing capability reads corresponding computer program instructions in the nonvolatile memory into the memory for operation. From the hardware layerIn addition to the processor, the memory, the network interface, and the nonvolatile memory, any device with data processing capability in which the apparatus is located in the embodiment may further include other hardware generally according to an actual function of the any device with data processing capability, which is not described herein again.
The implementation process of the functions and actions of each unit in the above device is specifically described in the implementation process of the corresponding step in the above method, and is not described herein again.
For the device embodiment, since it basically corresponds to the method embodiment, reference may be made to the partial description of the method embodiment for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. One of ordinary skill in the art can understand and implement it without inventive effort.
The embodiment of the present invention further provides a computer-readable storage medium, on which a program is stored, and when the program is executed by a processor, the method for compressing a semantic graph based on dynamic object segmentation in the foregoing embodiments is implemented.
The computer readable storage medium may be an internal storage unit, such as a hard disk or a memory, of any data processing capability device described in any of the foregoing embodiments. The computer readable storage medium may also be an external storage device such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash memory card (Flash card), etc. provided on the device. Further, the computer readable storage medium may include both internal storage units and external storage devices of any data processing capable device. The computer readable storage medium is used to store the computer program, and other programs and data needed by any of the data processing capable devices, and may also be used to temporarily store data that has been or will be output.
The above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and these modifications or substitutions do not depart from the scope of the embodiments of the present invention in nature.
Claims (8)
1. A semantic graph compression method based on dynamic object segmentation is characterized by comprising the following steps:
s1, initializing a simulation scene, wherein the simulation scene consists of a static background and a dynamic object;
s2, updating and drawing the simulation scene to obtain a semantic graph and two-dimensional bounding volumes of all dynamic objects under a semantic graph coordinate system;
s3, segmenting the semantic data of the dynamic object by using the two-dimensional bounding volume to form a plurality of dynamic object semantic subgraphs; filling the corresponding image area with the adjacent pixels of each dynamic object semantic subgraph in the rest static background semantic graph;
in the step S3, the semantic data of the dynamic object is segmented by using the two-dimensional bounding volume in the step S2 to form a plurality of semantic subgraphs, and the semantic subgraphs are specifically realized by the following substeps:
(1) Initializing corresponding semantic subgraph data according to the size of a two-dimensional bounding volume of each dynamic object, wherein each element of the semantic subgraph is initialized to be (R: 0, G:0, B: 0);
(2) Traversing each pixel of the semantic graph, and for each pixel, performing the following processing:
judging whether the pixel coordinate of the pixel is located in the range of a two-dimensional surrounding body of a certain dynamic object, if so, calculating the relative coordinate of the pixel under the semantic sub-image coordinate system of the dynamic object, writing the relative coordinate into a corresponding element of the semantic sub-image of the dynamic object, otherwise, ignoring the pixel, and continuously traversing the next pixel to obtain the semantic sub-image of the certain dynamic object;
and S4, respectively coding all the dynamic object semantic subgraphs and the filled static background semantic graph by using a coding algorithm.
2. The semantic graph compression method based on dynamic object segmentation according to claim 1, wherein all objects of the simulation scene in S1 are assigned with an ID, and IDs of objects having the same semantic class are the same; each ID uniquely corresponds to one color, and different IDs correspond to different colors.
3. The semantic graph compression method based on dynamic object segmentation according to claim 1, wherein the updating of the simulation scene in S2 includes pose updating of all dynamic objects and rendering view angles; each pixel in the semantic graph uniquely corresponds to one object, and the color corresponding to the object ID is used for coloring.
4. The method according to claim 1, wherein in S2, the two-dimensional bounding volume of the dynamic object in the semantic map coordinate system contains all pixels of the dynamic object on the semantic map.
5. The semantic map compression method based on dynamic object segmentation according to claim 1, wherein in the step S3, the remaining static background semantic map fills the corresponding image region with the adjacent pixels of each dynamic object semantic sub-map, and is implemented by the following sub-steps:
(1) Searching edge pixels around the dynamic object semantic subgraph, and if the edge pixels exceed the range of the semantic graph, determining the edge pixels to be invalid;
(2) And counting all effective edge pixels, and filling the corresponding image area of the dynamic object semantic subgraph in the semantic graph according to the values of the edge pixels.
6. The semantic graph compression method based on dynamic object segmentation according to claim 1, wherein the encoding result of each dynamic object semantic subgraph in S4 is accompanied by corresponding two-dimensional bounding volume information.
7. A semantic graph compression device based on dynamic object segmentation is characterized by comprising one or more processors and being used for realizing the semantic graph compression method based on dynamic object segmentation in any one of claims 1 to 6.
8. A computer-readable storage medium, on which a program is stored, which, when executed by a processor, implements the semantic graph compression method based on dynamic object segmentation according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211390992.4A CN115496818B (en) | 2022-11-08 | 2022-11-08 | Semantic graph compression method and device based on dynamic object segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211390992.4A CN115496818B (en) | 2022-11-08 | 2022-11-08 | Semantic graph compression method and device based on dynamic object segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115496818A CN115496818A (en) | 2022-12-20 |
| CN115496818B true CN115496818B (en) | 2023-03-10 |
Family
ID=85115811
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211390992.4A Active CN115496818B (en) | 2022-11-08 | 2022-11-08 | Semantic graph compression method and device based on dynamic object segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115496818B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111739037A (en) * | 2020-07-31 | 2020-10-02 | 之江实验室 | Semantic segmentation method for indoor scene RGB-D image |
| CN111918071A (en) * | 2020-06-29 | 2020-11-10 | 北京大学 | Data compression method, device, equipment and storage medium |
| CN112132897A (en) * | 2020-09-17 | 2020-12-25 | 中国人民解放军陆军工程大学 | Visual SLAM method based on deep learning semantic segmentation |
| CN114915786A (en) * | 2022-04-26 | 2022-08-16 | 哈尔滨工业大学（深圳） | Asymmetric semantic image compression method for scene of Internet of things |
| CN115049832A (en) * | 2022-06-27 | 2022-09-13 | 西安交通大学 | Scene semantic segmentation method, system, terminal device and readable storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180161986A1 (en) * | 2016-12-12 | 2018-06-14 | The Charles Stark Draper Laboratory, Inc. | System and method for semantic simultaneous localization and mapping of static and dynamic objects |
| EP3660787A4 (en) * | 2017-07-25 | 2021-03-03 | Cloudminds (Shenzhen) Robotics Systems Co., Ltd. | Training data generation method and generation apparatus, and image semantics segmentation method therefor |
| DE102020209024A1 (en) * | 2020-07-20 | 2022-01-20 | Robert Bosch Gesellschaft mit beschränkter Haftung | Method for generating a surveillance image |
| US12062174B2 (en) * | 2020-09-15 | 2024-08-13 | Sri International | Fully automated multimodal system architecture for semantic segmentation of large-scale 3D outdoor point cloud data |
-
2022
- 2022-11-08 CN CN202211390992.4A patent/CN115496818B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111918071A (en) * | 2020-06-29 | 2020-11-10 | 北京大学 | Data compression method, device, equipment and storage medium |
| CN111739037A (en) * | 2020-07-31 | 2020-10-02 | 之江实验室 | Semantic segmentation method for indoor scene RGB-D image |
| CN112132897A (en) * | 2020-09-17 | 2020-12-25 | 中国人民解放军陆军工程大学 | Visual SLAM method based on deep learning semantic segmentation |
| CN114915786A (en) * | 2022-04-26 | 2022-08-16 | 哈尔滨工业大学（深圳） | Asymmetric semantic image compression method for scene of Internet of things |
| CN115049832A (en) * | 2022-06-27 | 2022-09-13 | 西安交通大学 | Scene semantic segmentation method, system, terminal device and readable storage medium |
Non-Patent Citations (4)
| Title |
|---|
| Real-time Semantic Segmentation of Crop and Weed for Precision Agriculture Robots Leveraging Background Knowledge in CNNs;Andres Milioto,et al.;《IEEE Xplore》;20180913;全文 * |
| 基于环境语义信息的同步定位与地图构建方法综述;李小倩 等;《工程科学学报》;20210630;全文 * |
| 面向内容的工程图识别与理解综述;路通 等;《图学学报》;20121031;全文 * |
| 面向实时视频流分析的边缘计算技术;杨铮 等;《中国科学：信息科学》;20220110;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115496818A (en) | 2022-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112884875B (en) | Image rendering method, device, computer equipment and storage medium | |
| US9852544B2 (en) | Methods and systems for providing a preloader animation for image viewers | |
| US7940279B2 (en) | System and method for rendering of texel imagery | |
| US20230108967A1 (en) | Micro-meshes, a structured geometry for computer graphics | |
| US9928643B2 (en) | Hierarchical continuous level of detail for three-dimensional meshes | |
| US11450057B2 (en) | Hardware acceleration for ray tracing primitives that share vertices | |
| KR20160033128A (en) | Sparse gpu voxelization for 3d surface reconstruction | |
| US20050017971A1 (en) | Ray tracing hierarchy | |
| Duguet et al. | Flexible point-based rendering on mobile devices | |
| CN1655191A (en) | Programmable graphic hardware based rapid voxel method for polygonal grid model | |
| US20150332481A1 (en) | Indexed uniform styles for stroke rendering | |
| CN114419099B (en) | Method for capturing motion trail of virtual object to be rendered | |
| CN109979013B (en) | Three-dimensional face mapping method and terminal equipment | |
| CN113648655B (en) | Virtual model rendering method and device, storage medium and electronic equipment | |
| CN114863038A (en) | Real-time dynamic free visual angle synthesis method and device based on explicit geometric deformation | |
| CN115428027A (en) | Neural opaque point cloud | |
| CN110930492B (en) | Model rendering method, device, computer readable medium and electronic equipment | |
| US10347034B2 (en) | Out-of-core point rendering with dynamic shapes | |
| WO2019088865A1 (en) | Method and system for removing hidden surfaces from a three-dimensional scene | |
| CN115496818B (en) | Semantic graph compression method and device based on dynamic object segmentation | |
| CN110136235B (en) | Three-dimensional BIM model shell extraction method and device and computer equipment | |
| WO2021155318A1 (en) | Method for improved handling of texture data for texturing and other image processing tasks | |
| US20240320903A1 (en) | Methods and systems for generating enhanced light texture data | |
| Poux | How to represent 3D Data? | |
| WO2023184139A1 (en) | Methods and systems for rendering three-dimensional scenes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |