3D Point Cloud Object Recognition via Hybrid Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D object recognition systems face challenges in accurately identifying and classifying objects in 3D street scenes due to variations in texture, color, and illumination, and are hindered by the large volume of 3D point cloud data and limitations in computer hardware, particularly in distinguishing ground, building, and vertical objects.
Innovation Solution
A hybrid two-stage approach using unsupervised and supervised segmentation methods to detect ground and building objects from 3D point cloud data, followed by voxel-based segmentation to classify remaining vertical objects, employing boosted decision trees for semantic labeling, which reduces data processing and manual labeling requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised segmentation method is used for all objects in 3D point cloud data, then classification accuracy is improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent divides the 3D point cloud data processing into two distinct stages: first applying unsupervised segmentation to ground and building objects, then applying supervised segmentation only to remaining vertical objects. This hierarchical segmentation strategy reduces the overall computational complexity while maintaining classification accuracy for all object types.
Solution Approach 2:
The patent applies different segmentation methods (unsupervised vs. supervised) to different types of objects based on their specific characteristics. Ground and building objects use unsupervised methods, while vertical objects use supervised methods, optimizing the balance between accuracy and computational efficiency for each local region.
2Measurement precision
If supervised segmentation method is applied to all 3D point cloud data, then object classification accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the processing task into two phases: unsupervised segmentation for ground/building objects (faster processing) and supervised segmentation for vertical objects (higher accuracy). This time-efficient segmentation strategy reduces overall processing time while maintaining high classification accuracy.
Solution Approach 2:
Instead of applying computationally intensive supervised segmentation to all objects, the patent applies it partially only to vertical objects where higher accuracy is most beneficial, while using faster unsupervised methods for ground and building objects.
3Measurement precision
If manual training data is collected for all object categories, then classification accuracy is improved, but the need for manual labeling and data preparation increases
Solution Approach 1:
The patent segments the data preparation process by using unsupervised methods for ground and building objects (no manual labeling needed) and supervised methods only for vertical objects (limited manual labeling). This improves productivity by reducing overall manual data preparation while maintaining accuracy where needed.
Solution Approach 2:
The unsupervised segmentation component enables the system to automatically classify ground and building objects without manual training data, making the system self-sufficient for these object types and reducing the burden of manual data collection and labeling.
Data Source
AI summary
A method comprising: obtaining a three-dimensional (3D) point cloud about at least one object of interest; detecting ground and/or building objects from 3D point cloud data using an unsupervised segmentation method; removing the ground and/or building objects from the 3D point cloud data; and detecting one or more vertical objects from the remaining 3D point cloud data using a supervised segmentation method.