Method for generating three-dimensional space semantic occupancy grids at low cost
By obtaining point cloud data and obstacle labeling information, removing obstacles and segmenting point clouds using ground segmentation algorithms, combining obstacle labeling to generate three-dimensional space semantic occupancy grids, solving the cost problem in the existing technology and achieving low cost , efficiently occupying grid generation, suitable for autonomous driving systems.
Patent Information
- Application Number
- CN202411676442.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-22
- Publication Date
- 2025-01-17
AI Technical Summary
In the prior art, the cost of generating three-dimensional space semantics occupying grids is relatively high, and it is time-consuming and costly to directly mark the point cloud, making it difficult to meet the low-cost needs of autonomous driving systems.
By obtaining point cloud data and obstacle labeling information, remove obstacle labeling point clouds and use ground segmentation algorithm to divide background point clouds into ground points and non-ground points, and reintegrate obstacle labeling, and finally use the occupying grid generation algorithm to generate Generate semantic segmentation point clouds.
It reduces the cost of occupancy of grid production, greatly saves labeling time, improves generation efficiency, maintains the accuracy and reliability of data sets, and is suitable for data set requirements of different sizes and complexities.
Smart Images
Figure CN119323676A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method for generating a three-dimensional space semantic occupancy grid at low cost. Background Art
[0002] In autonomous driving, the 3D semantic occupancy grid is an important technology for environmental perception and map modeling. This technology divides the environment into a 3D grid and maintains an occupancy state for each grid, which is usually used to indicate whether the grid is occupied and the type of object occupied. This enables real-time perception and understanding of roads, obstacles and other environmental features. This method is widely used in autonomous driving, mainly in two aspects: environmental perception and map modeling and positioning.
[0003] Environmental perception: Autonomous vehicles need to accurately perceive the surrounding environment, including roads, obstacles, pedestrians, etc. By occupying the grid, the vehicle can update the state of the surrounding environment in real time and perform obstacle avoidance, path planning and driving decisions accordingly.
[0004] Map modeling and positioning: The occupancy grid map is not only used to perceive the environment, but also to build a map of the vehicle's surroundings, and combined with positioning information, to achieve accurate positioning of the vehicle. This is very important for navigation, map updates, and long-term path planning.
[0005] Generally, generating a 3D space semantic occupancy grid requires a semantically segmented point cloud. If the point cloud is directly semantically segmented and annotated, the annotation cost is high. Therefore, the present invention proposes a method for generating a 3D space semantic occupancy grid at low cost for autonomous driving. Summary of the invention
[0006] In order to solve these problems, the present invention discloses a method for low-cost generation of three-dimensional space semantic occupancy grids for autonomous driving, aiming to reduce the production cost of three-dimensional space occupancy grids.
[0007] To achieve the above object, the present invention provides the following technical solutions: A method for generating a three-dimensional space semantic occupancy grid at low cost, the method comprising: S1: Obtain point cloud data and corresponding obstacle annotation information. Obstacle annotations usually include people and different types of vehicles. The obstacle annotation point cloud is the foreground point cloud. S2: Obstacle annotation point clouds are removed from the point cloud data according to the positions of the obstacle annotations to obtain background point clouds; S3: Use the ground segmentation algorithm to segment the background point cloud generated by S2, and divide all points in the background point cloud into ground points and non-ground points; S4: The ground point cloud, non-ground point cloud, and obstacle annotation point cloud are reintegrated into a complete point cloud according to the semantics of their respective parts. At this time, the complete point cloud contains the semantics of ground points, non-ground points, and obstacle types; S5: Use the occupancy grid generation algorithm to generate an occupancy grid for the semantic segmentation point cloud generated by S4.
[0008] Compared with the prior art, the beneficial effects of the present invention are: the present invention provides a method for low-cost generation of a three-dimensional semantic occupancy grid for autonomous driving. Low cost: Compared with directly semantically segmenting and annotating point clouds, the cost of occupancy grid production can be greatly reduced. High efficiency: Through the method of the present invention, occupancy grid datasets can be generated more efficiently, saving annotation time. Scalability: The technical solution has certain flexibility and scalability, and can adapt to the needs of datasets of different scales and complexities. Accuracy: Although it is a low-cost method, the present invention still maintains the accuracy and reliability of the dataset, which helps to improve the performance and safety of the autonomous driving system. Easy to implement: The technical solution is simple in design, easy to implement and operate, and is conducive to promotion and application to actual scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Figure 1 The present invention is a flowchart of a method for generating a three-dimensional space semantic occupancy grid at low cost according to an embodiment of the present invention. DETAILED DESCRIPTION
[0010] In order to make the technical problems, technical solutions and beneficial effects to be solved by 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 embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0011] See also Figure 1 As an embodiment of the present invention, a method for generating a three-dimensional space semantic occupancy grid at low cost comprises: S1: Obtain point cloud data and corresponding obstacle annotation information. Obstacle annotations usually include people and different types of vehicles. The obstacle annotation point cloud is the foreground point cloud. S2: Obstacle annotation point clouds are removed from the point cloud data according to the positions of the obstacle annotations to obtain background point clouds; S3: Use the ground segmentation algorithm to segment the background point cloud generated by S2, and divide all points in the background point cloud into ground points and non-ground points; S4: The ground point cloud, non-ground point cloud, and obstacle annotation point cloud are reintegrated into a complete point cloud according to the semantics of their respective parts. At this time, the complete point cloud contains the semantics of ground points, non-ground points, and obstacle types; S5: Use the occupancy grid generation algorithm to generate an occupancy grid for the semantic segmentation point cloud generated by S4.
[0012] Usually, generating a semantic occupancy grid in three-dimensional space requires a semantically segmented point cloud. If the point cloud is directly semantically segmented and labeled, the labeling cost is high. The present invention only needs to be generated through an algorithm based on the point cloud obstacle labeling, which is about 10 times cheaper than the point cloud semantic labeling in the current existing labeling market. First, obstacles are removed according to the point cloud obstacle labeling, and the remaining point cloud is segmented into ground points and non-ground points using the ground segmentation algorithm. The point cloud obstacles are then refilled back into the point cloud. At this time, the point cloud semantic segmentation is completed, and the point cloud semantic segmentation is used to generate the occupancy grid.
[0013] It should be noted that, in this article, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.
[0014] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for generating a three-dimensional space semantic occupancy grid at low cost, characterized in that: The method comprises: S1: Obtain point cloud data and corresponding obstacle annotation information. Obstacle annotations are people and different types of vehicles. The obstacle annotation point cloud is the foreground point cloud. S2: Obstacle annotation point clouds are removed from the point cloud data according to the positions of the obstacle annotations to obtain background point clouds; S3: Use the ground segmentation algorithm to segment the background point cloud generated by S2, and divide all points in the background point cloud into ground points and non-ground points; S4: The ground point cloud, non-ground point cloud, and obstacle annotation point cloud are reintegrated into a complete point cloud according to the semantics of their respective parts. At this time, the complete point cloud contains the semantics of ground points, non-ground points, and obstacle types; S5: Use the occupancy grid generation algorithm to generate an occupancy grid for the semantic segmentation point cloud generated by S4.