Urban Point Cloud Segmentation via Block-Based Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for segmenting point cloud data from urban environments are inefficient and prone to over-segmentation, particularly in complex scenes with multiple objects and overhanging structures, as they require prior knowledge of object locations or assume specific scene geometries.
Innovation Solution
A method that employs ground filtering to remove terrain points, followed by clustering non-ground points based on cell continuity, and uses block-based segmentation to handle large scenes, allowing for adaptive measurement adjustments and parallel processing to prevent over-segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If graph cut method is used for segmentation, then segmentation can be performed on 3D point cloud data, but the method requires prior knowledge of object locations which reduces adaptability
Solution Approach 1:
The patent segments the point cloud data into multiple blocks first, then performs segmentation on each block independently. This divide-and-conquer approach reduces the complexity of processing large datasets while maintaining the ability to handle unknown object locations through local analysis. Each block is processed separately using continuity measures, eliminating the need for global prior knowledge of object positions.
Solution Approach 2:
The patent transforms the segmentation problem from a global 3D point cloud analysis to a localized block-based analysis. By dividing the continuous point cloud into discrete blocks and applying continuity measures within each block, the method changes the dimensional scope of analysis, enabling segmentation without requiring prior knowledge of overall object locations.
2Measurement precision
If traditional segmentation methods are applied to large urban scenes, then segmentation can be performed, but processing time and memory requirements increase significantly
Solution Approach 1:
The patent divides the large point cloud dataset into multiple smaller blocks, enabling parallel processing and reducing memory requirements. Each block can be processed independently using continuity measures, significantly improving processing efficiency while maintaining segmentation accuracy. This block-based approach allows the system to handle large urban scenes that would otherwise be computationally infeasible.
Solution Approach 2:
The patent applies continuity measures selectively within each block rather than performing exhaustive analysis on the entire point cloud. By focusing computational resources on localized regions and using adaptive thresholds, the method achieves sufficient segmentation accuracy without the excessive processing time and memory consumption associated with global analysis methods.
3Device complexity
If fixed continuity measures are used for segmentation, then processing is simplified, but over-segmentation occurs in complex scenes with overhanging structures
Solution Approach 1:
The patent employs adaptive continuity measures that adjust dynamically based on the local characteristics of each block. Instead of using fixed thresholds, the method calculates continuity measures relative to the specific point cloud distribution within each block, allowing the segmentation algorithm to adapt to complex geometries such as overhanging structures. This dynamic adjustment prevents over-segmentation while maintaining relatively simple processing logic.
Solution Approach 2:
The patent applies different continuity measures and thresholds to different blocks based on their local characteristics. Each block is analyzed independently, and the continuity measure is adjusted according to the specific geometric properties of that region. This local quality approach allows the system to handle diverse urban environments without requiring complex global algorithms, preventing over-segmentation in complex scenes while keeping the overall method simple.
Data Source
AI summary
A method, apparatus, system, and article of manufacture provide object descriptors for objects in point cloud data for an urban environment by segmenting the point cloud data. Point cloud data for an urban environment is obtained using a ground-based laser scanner. Terrain points are filtered out from the point cloud data using ground filtering. The point cloud data is then segmented into two or more blocks. Objects that lie on neighboring adjacent blocks are combined. Object descriptors for the combined objects are then provided (e.g., to the user or a program used by the user).