Laser Point Cloud Labeling via 3D Scene Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for labeling laser point clouds in automatic driving systems are inefficient due to the manual point-by-point labeling of a large number of laser points, leading to low speed and accuracy in identifying target objects.
Innovation Solution
A method and device that convert laser point clouds into a 3D scene with a coordinate system, allowing for the mapping and labeling of laser points in a 3D environment, enabling centralized visualization and rapid identification of target types through the use of multiple cameras and interactive labeling tools like 3D selecting boxes and coloring brushes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual point-by-point labeling is used for laser point clouds, then labeling can be performed with simple tools, but labeling speed and efficiency are extremely low
Solution Approach 1:
The patent transforms 2D/3D scattered laser points into a structured 3D scene representation with spatial relationships. By constructing a three-dimensional scene model that organizes points according to their spatial coordinates and visual characteristics, the system enables efficient batch processing and visual judgment, dramatically improving labeling speed while maintaining manageable system complexity through standardized scene construction algorithms
2Measurement precision
If manual point-by-point labeling is used, then each point can be labeled individually with high precision, but the overall labeling process is time-consuming and labor-intensive
Solution Approach 1:
The patent segments the labeling task by first classifying points into different target objects based on their spatial coordinates and visual characteristics in the 3D scene. This segmentation allows operators to label groups of points belonging to the same object simultaneously rather than processing each point individually, maintaining high accuracy while significantly reducing labeling time and effort
Solution Approach 2:
The system performs preliminary organization of laser points into a structured 3D scene with established spatial relationships and visual characteristics before the actual labeling process. This preliminary action pre-sorts and pre-positions points for efficient batch labeling, enabling operators to quickly identify and label multiple points at once while maintaining precision through the pre-established spatial framework
3Ease of operation
If traditional laser point cloud labeling is used, then the process can be completed with basic software, but the ability to visually judge and identify target types is limited
Solution Approach 1:
The patent creates a three-dimensional scene representation that organizes laser points according to their spatial coordinates and visual characteristics. This 3D scene provides intuitive spatial relationships and visual context, enabling operators to easily judge and identify target types through visual inspection rather than processing raw point cloud data, thereby enhancing ease of operation with moderate software complexity
Data Source
AI summary
The present application discloses a method and device of labeling laser point cloud. The method comprises: receiving data of a laser point cloud; constructing a 3D scene and establishing a 3D coordinate system corresponding to the 3D scene; converting a coordinate of each laser point in the laser point cloud into a 3D coordinate in the 3D coordinate system; mapping laser points included in the laser point cloud into the 3D scene respectively according to the 3D coordinate of the laser points; labeling the laser points in the 3D scene.