Circular Grid Map for LiDAR Point Cloud Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional point cloud clustering techniques using 2D square grid maps face challenges in optimally clustering points indicating both long-range and short-range objects, as large cell sizes merge nearby objects and small cell sizes separate distant points of the same object.
Innovation Solution
The method projects a point cloud from a 3D LiDAR sensor onto a 2D circular grid map with varying circle sizes, using a controller to detect representative points and cluster them based on a reference frame corresponding to each cell's size and shape, allowing for efficient clustering of both long-range and short-range objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a large cell size is used in the square grid map, then clustering computation is simplified and faster, but points indicating different objects located in a short range are clustered as one
Solution Approach 1:
The patent applies local quality by using a circular grid map where cell sizes and shapes vary depending on the distance from the vehicle. Cells closer to the vehicle are smaller to distinguish nearby objects, while cells farther away are larger to reduce computational complexity. This spatially adaptive cell configuration allows the system to maintain both high object distinction accuracy in the near field and efficient computation in the far field.
2Measurement precision
If a small cell size is used in the square grid map, then object distinction accuracy is improved, but points indicating one object located in a long range are separately clustered
Solution Approach 1:
The circular grid map implements local quality by dynamically adjusting cell characteristics based on spatial location. In the near field, small cells provide high resolution for accurate object distinction. In the far field, larger cells maintain computational efficiency while still grouping points belonging to the same distant object, preventing the fragmentation problem that occurs with uniform small cells across the entire map.
3Device complexity
If a uniform cell size is used in the square grid map, then the clustering method is simple, but it cannot optimally cluster both long-range and short-range objects
Solution Approach 1:
The patent applies dynamics by transitioning from a static uniform cell structure to a dynamic cell structure where the circular grid map automatically adjusts cell sizes and shapes based on the distance from the vehicle. This dynamic adaptation allows the same clustering method to handle both near and far objects effectively, with cells becoming smaller and more numerous near the vehicle for detailed object distinction, and larger and fewer in the far field for efficient processing.
Data Source
AI summary
An apparatus for clustering a point cloud can include: a three-dimensional (3D) light detection and ranging (LiDAR) sensor configured to generate a point cloud around a vehicle and a controller configured to project the point cloud generated by the 3D LiDAR sensor onto a circular grid map to be converted into two-dimensional (2D) points, the circular grid map including a plurality of cells, and to cluster the 2D points on the circular grid map.