NDT Map Compression Using Mean and Covariance Tree Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The large data size of 3D point cloud maps used in autonomous driving systems limits the scalability and real-time efficiency of autonomous driving systems, making it difficult to expand the driving area effectively.
Innovation Solution
The method involves processing 3D point cloud data as a normal distribution set to generate compressed NDT map data by transforming mean vector and covariance matrix data into tree data structures and setting reference ranges for offset vectors and standard deviations, allowing for efficient compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D point cloud map data is used for autonomous driving systems, then the system can recognize the surrounding environment and understand road structures, but the data size becomes too large to support wide-area scalability and real-time operation
Solution Approach 1:
The patent transforms the representation parameters of map data from raw 3D point cloud coordinates to statistical parameters (mean vectors and covariance matrices) of normal distributions. This parameter transformation compresses the data while preserving the essential geometric and spatial characteristics needed for environment recognition and road structure understanding.
Solution Approach 2:
The patent extracts only the essential statistical characteristics (mean and covariance) from the complete 3D point cloud data, discarding redundant information. By taking out only the critical parameters that define the spatial distribution of points, the system achieves compression while maintaining functionality for autonomous driving tasks.
2Productivity
If NDT map data is compressed to reduce data size, then scalability and real-time operation efficiency improve, but registration performance may be compromised
Solution Approach 1:
The patent changes the parameter representation from raw point cloud data to normal distribution statistical parameters, which maintains the essential spatial information needed for registration while enabling compression. The mean vectors preserve position information and covariance matrices preserve shape and orientation information, both critical for accurate map registration.
Solution Approach 2:
The patent creates a statistical copy (normal distribution model) of the original point cloud data that captures the essential characteristics without requiring the complete original data. This copied representation is sufficient for registration operations while being much more compact in size.
Data Source
AI summary
Provided are a method, apparatus, and a computer program for compressing NDT map data for an autonomous driving system. Various embodiments of the present invention provide a method of compressing normal distribution transform (NDT) map data generated by modeling a three-dimensional (3D) point cloud for a predetermined area as a normal distribution set for an autonomous driving system, which is performed by a computing device, the method including: processing mean vector data included in the NDT map data; processing covariance matrix data included in the NDT map data; and generating compressed NDT map data using the processed mean vector data and the processed covariance matrix data.