Incremental Point Cloud Segmentation for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional point cloud segmentation methods are inefficient and resource-intensive, as they process each frame independently, failing to utilize correlated information across consecutive frames, which leads to unnecessary calculations and potential loss of correlated data.
Innovation Solution
The method involves segmenting a first frame of point cloud using a segmentation algorithm, mapping the resulting clusters to a subsequent frame, identifying and segmenting only the non-mapped points, and combining the results to generate the segmentation for the second frame, thereby leveraging correlated information and reducing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire point cloud is segmented for each frame using a segmentation algorithm, then the segmentation result is complete and accurate, but the computational resources and processing time are excessively consumed
Solution Approach 1:
The patent divides the point cloud processing into two segments: (1) mapping and reusing point clusters from the previous frame, and (2) segmenting only the new points in the current frame. This segmentation of the processing task allows the system to maintain accurate segmentation results while significantly reducing computational resources by avoiding redundant segmentation of unchanged points.
Solution Approach 2:
The patent performs preliminary mapping of point clusters from the previous frame to the current frame before segmentation. By predicting the positions of existing point clusters and marking them in advance, the system avoids the need to re-segment these points, thus improving processing efficiency while maintaining segmentation accuracy.
2Ease of manufacture
If the entire point cloud is segmented for each frame independently, then the segmentation is computationally straightforward, but correlated information among consecutive frames is lost
Solution Approach 1:
The patent implements a feedback mechanism where the segmentation results from the previous frame are mapped to the current frame. This feedback loop preserves correlated information across frames by carrying forward the segmentation knowledge, allowing the system to maintain simplicity while utilizing temporal correlations for improved efficiency.
Solution Approach 2:
The patent creates a copy of the previous frame's point clusters and maps them to the current frame's coordinate system. This copying approach preserves the segmentation information from the previous frame, maintaining the simplicity of the segmentation process while preventing information loss through frame-by-frame independent processing.
3Manufacturing precision
If all points in the new frame are segmented, then the segmentation result is complete, but unnecessary calculations are performed on points that have already been segmented in previous frames
Solution Approach 1:
The patent extracts and separates the new points from the current frame that have not been present in previous frames. By identifying and isolating only these new points for segmentation, the system achieves complete segmentation results without performing unnecessary calculations on points that have already been processed, thus reducing computational energy consumption.
Solution Approach 2:
The patent applies partial action by segmenting only the necessary portion of the point cloud (the new points) rather than the entire point cloud. This selective approach ensures segmentation completeness for all points while avoiding excessive computational resources spent on re-segmenting points that have already been processed.
Data Source
AI summary
A method for segmentation of a point cloud includes receiving a first frame of point cloud from a sensor; segmenting the first frame of point cloud to obtain a first set of point clusters representing a segmentation result for the first frame of point cloud; receiving a second frame of point cloud from the sensor; mapping the first set of point clusters to the second frame of point cloud; determining points within the second frame of point cloud which do not belong to the mapped first set of point clusters; segmenting the points within the second frame of point cloud which do not belong to the mapped first set of point clusters to obtain a second set of point clusters; and generating a segmentation result for the second frame of point cloud by combining the first set of point clusters and the second set of point clusters.