Ordered LiDAR Point Cloud Features for Ray-Scan Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional feature extraction methods from LiDAR sensor point clouds do not account for the ray-scanning process, resulting in unordered and irregularly sampled data points that hinder effective pattern recognition tasks in autonomous vehicles.
Innovation Solution
The method organizes point clouds using a 3D spherical grid representation, sorting data points within each volume to generate ordered feature vectors that account for spatial ordering and metadata, enabling coherent feature extraction and improved pattern recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional feature extraction methods are used on LiDAR point clouds, then the processing can be performed on raw data, but the unordered and irregularly sampled data points hinder effective pattern recognition
Solution Approach 1:
The point cloud data is segmented into multiple 3D bins using a spherical coordinate system. Each bin represents a discrete volumetric region in 3D space, and points within each bin are processed together. This segmentation transforms the unordered point cloud into an organized structure where spatial relationships are preserved through the bin assignment, resolving the contradiction between maintaining raw data processing and preserving spatial ordering information for effective pattern recognition
Solution Approach 2:
A spherical coordinate-based 3D binning system is introduced as an intermediary structure between the raw LiDAR point cloud and the feature extraction process. This intermediary organizes points into bins based on their spherical coordinates (range, azimuth, elevation), preserving the spatial ordering information while enabling coherent feature extraction. The binning system acts as a mediator that transforms unordered points into an organized representation without losing the underlying spatial relationships
2Ease of manufacture
If point clouds are processed without accounting for ray-scanning procedure, then processing is simpler, but the spatial ordering custom to the LiDAR sensor is not properly accounted for
Solution Approach 1:
The patent transforms the coordinate system parameters from Cartesian to spherical coordinates to match the LiDAR sensor's native scanning pattern. By expressing point positions in terms of range, azimuth, and elevation angles, the processing aligns with the ray-scanning procedure. This parameter change enables the system to account for spatial ordering accuracy while maintaining processing efficiency, as the spherical coordinate transformation naturally reflects the sensor's scanning geometry
Data Source
AI summary
A method is described which includes receiving a point cloud having a plurality of data points each representing a 3D location in a 3D space, the point cloud being obtained using a detection and ranging (DAR) sensor. For each data point, associating the data point with a 3D volume containing the 3D location of the data point, the 3D volume being defined using a 3D lattice that partitions the 3D space based on spherical coordinates. For at least one 3D volume, the data points are sorted within the 3D volume based on at least one dimension of the 3D lattice; and the sorted data points are stored as a set of ordered data points. The method also includes performing feature extraction on the set of ordered data points to generate a set of ordered feature vectors and providing the set of ordered feature vectors to perform a machine learning inference task.