Point Cloud Segmentation via Line Segment Fitting and Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing point cloud segmentation methods in autonomous driving are inefficient, requiring multiple scans of point cloud data and failing to effectively segment ground and non-ground objects in real-time, leading to low efficiency and high calculation costs.
Innovation Solution
A method that clusters target point cloud data into first datasets with feature points fitted on segmented line segments and combines these datasets based on distances between line segments, allowing for efficient segmentation by traversing the data only once, thereby reducing over-segmentation and under-segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If point cloud data is scanned multiple times using traditional segmentation algorithms, then segmentation completeness is improved, but processing efficiency deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the point cloud data processing into two distinct phases: line segment fitting (grouping collinear points) and object clustering (grouping connected line segments). This two-stage segmentation approach allows the algorithm to process data in a structured manner, achieving complete segmentation in a single pass through the point cloud data, thereby resolving the contradiction between segmentation completeness and processing efficiency.
2Measurement precision
If traditional point cloud segmentation algorithms are used, then segmentation accuracy is maintained, but calculation complexity increases
Solution Approach 1:
The algorithm segments the complex segmentation task into two simpler sub-tasks: line segment fitting and object clustering. By breaking down the problem this way, each sub-task can be solved with simpler, more efficient algorithms while maintaining overall segmentation accuracy. The line segment fitting step reduces dimensionality by grouping collinear points, and the clustering step operates on these fitted segments rather than raw points, reducing computational complexity.
Solution Approach 2:
The patent applies preliminary action by performing line segment fitting before object clustering. This preliminary organization of points into line segments creates a structured intermediate representation that simplifies the subsequent clustering operation. By preparing the data in advance through line fitting, the algorithm reduces the complexity of the main clustering task while preserving segmentation accuracy.
3Reliability
If multiple scans are performed for point cloud segmentation, then segmentation reliability is improved, but processing time increases
Solution Approach 1:
The patent implements continuity of useful action by designing an algorithm that processes point cloud data in a single continuous pass, eliminating the need for multiple scans. The two-stage approach (line fitting followed by clustering) operates sequentially on the same data stream, maintaining continuous processing without restarting or re-scanning, thereby achieving reliable segmentation results in real-time.
Data Source
AI summary
This application discloses a method and apparatus for segmenting point cloud data, a storage medium and an electronic device. The method includes: obtaining target point cloud data by scanning target objects around a vehicle with laser lines; clustering the target point cloud data to obtain a plurality of first datasets, wherein feature points represented by point cloud data within each of the plurality of first datasets are fitted on one segmented line segment, each feature point being a point on a respective target object; and combining the plurality of first datasets according to distances between the corresponding plurality of segmented line segments to obtain a plurality of second datasets, wherein each second dataset includes at least one of the plurality of first datasets. This application resolves a technical problem of relatively low efficiency of point cloud segmentation in the related art.