3D Point Assignment Using Depth Classes in Overlapping Polygons
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D object recognition methods based on machine learning are computationally intensive and require large amounts of annotated data, which is time-consuming and expensive to produce, particularly in the context of large datasets for supervised learning.
Innovation Solution
A method for assigning data points from three-dimensional data to objects using two-dimensional surrounding polygons and depth classes, where overlapping polygons are merged to form groups, utilizing a projection matrix and the Hungarian method to optimize assignments, reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning-based 3D object recognition methods are used, then object recognition accuracy is improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The method segments the 3D point cloud data into multiple depth classes (e.g., near, mid, far ranges) and processes each depth class separately. This segmentation reduces the computational complexity of the overall task by breaking down the large-scale 3D recognition problem into smaller, more manageable sub-problems that can be processed more efficiently
Solution Approach 2:
The invention projects 3D point cloud data onto 2D image space using a projection matrix, transforming the problem from three-dimensional to two-dimensional. This dimensionality reduction allows the system to leverage 2D image processing techniques and surrounding polygon information, significantly reducing computational requirements while maintaining recognition accuracy
2Measurement precision
If machine learning-based 3D object recognition methods are used, then object recognition capability is improved, but processing speed decreases due to computational intensity
Solution Approach 1:
By dividing the point cloud into depth classes and processing each class separately with corresponding surrounding polygons, the method enables parallel processing of multiple depth segments, thereby improving overall processing speed while maintaining recognition capability
Solution Approach 2:
The system performs preliminary projection of 3D points to 2D space and pre-computes surrounding polygons before the main recognition process. This preliminary action prepares the data in advance, allowing the subsequent recognition step to operate more efficiently on pre-processed information
3Measurement precision
If large datasets are annotated using manual methods, then data quality for training is improved, but time consumption and costs increase significantly
Solution Approach 1:
The method enables automated annotation by allowing the system to generate surrounding polygons and depth class assignments automatically from the point cloud data itself, without requiring manual human intervention. This self-service approach maintains data quality while dramatically reducing annotation time and costs
Solution Approach 2:
The system creates simplified 2D projections and surrounding polygon representations that copy the essential spatial relationships from the 3D point cloud. These simplified representations can be used as training data, providing high-quality annotated information without the time investment required for full manual 3D annotation
Data Source
Figure 1~2
Figure 3~4
Figure 5a~5b
AI summary
The invention relates to a computer-implemented method for assigning data points from first data (26) to each of several objects (14, 14'), wherein the first data (26) comprise at least three spatial dimensions, and the method comprises the following step: - forming groups of projected data points (18, 18') assigned to the respective objects (14, 14') taking into account received information data from two-dimensional surrounding polygons (10, 10') and from a depth class (20, 20') of the data point by means of an assignment method such that, in the case of overlapping polygons (10, 10') where projected data points (18, 18') are located within more than one two-dimensional surrounding polygon (10, 10'), the groups are formed such that no depth class (20, 20') is simultaneously assigned to several of the overlapping polygons (10, ... 10') is assigned.Furthermore, the invention relates to a data processing device comprising means for carrying out the above method, as well as a computer program product comprising instructions that, when executed by a computer, cause the computer to carry out the above method. The invention also relates to a computer-readable data carrier on which the above computer program product is stored.