3D Point Cloud Label Assignment Through 2D Mapping and Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for assigning labels to virtual voxels in three-dimensional point cloud data are inefficient and require manual identification of objects, making it difficult to generate label data for object recognition processes.
Innovation Solution
An image processing apparatus that segments a three-dimensional point cloud using clustering parameters, maps clusters to a two-dimensional image, assigns labels based on two-dimensional image information, and determines labels through voting to automate the labeling process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the same label value is assigned to connected virtual voxels manually, then the labeling process can be completed, but the process requires manual identification and is inefficient for generating large amounts of label data
Solution Approach 1:
The patent projects three-dimensional point cloud data onto a two-dimensional image plane, creating a two-dimensional representation that copies the spatial relationships of the 3D data. This allows existing two-dimensional labeling techniques to be applied to the 3D point cloud, automatically generating labels without manual identification of each virtual voxel.
Solution Approach 2:
The patent introduces a two-dimensional image as an intermediary between the three-dimensional point cloud and the labeling process. By mapping 3D points to 2D image coordinates and using 2D labeling algorithms, the system automatically generates labels for 3D virtual voxels, eliminating the need for manual 3D labeling while preserving spatial relationships.
2Measurement precision
If manual identification is used to identify objects corresponding to label values, then accurate labeling can be achieved, but the process is time-consuming and cannot easily generate large amounts of label data
Solution Approach 1:
The patent creates a two-dimensional copy or projection of the three-dimensional point cloud data. This 2D representation maintains the spatial relationships and object structures, allowing automated 2D labeling algorithms to accurately identify objects and assign labels without manual intervention, thus achieving both accuracy and efficiency.
3Loss of information
If three-dimensional point cloud data is processed directly without mapping to two-dimensional image, then the full 3D information is preserved, but the complexity of the labeling process increases significantly
Solution Approach 1:
The patent transforms the three-dimensional point cloud data into a two-dimensional image projection, changing the dimensional representation. This dimensionality reduction simplifies the labeling process by allowing the use of成熟的2D labeling algorithms, while the projection process preserves the essential spatial relationships and object information from the 3D data.
Data Source
AI summary
An image processing apparatus includes: a three-dimensional point cloud segmentation unit for segmenting a three-dimensional point cloud of an object using clustering parameters so that the smaller the distance between points in the three-dimensional point cloud is, the more likely the points are to be clustered into the same cluster; a cluster mapping unit for generating a mapped point cloud by mapping clusters of the three-dimensional point cloud to a two-dimensional image based on the correspondence between the coordinates of the three-dimensional point cloud and the coordinates of pixels of the two-dimensional image; a label matching unit for assigning labels to points in the mapped point cloud based on a target label and position information in the two-dimensional image; a matching result voting unit for voting for labels of the points; and a label determination unit for determining the label for the points based on voting results.