Point Cloud Map Filtering for LIDAR Localization Accuracy

Solution Overview

Problem

Existing methods for merging LIDAR map data with newly acquired scan frame data often introduce localization errors due to difficulties in distinguishing points on perpendicular surfaces and data sparse regions, leading to inaccurate mapping.

Innovation Solution

The method involves filtering point cloud data by attributing selected metrics such as surface normals and data acquisition vectors, and using these attributes to accurately match and localize points, thereby enhancing the accuracy of scan frame data integration with existing maps.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If point cloud data is merged based solely on position information, then the merging process is simple, but localization error increases and mapping accuracy deteriorates

Engineering Contradiction:

Improvesimplicity of merging processVSAvoidlocalization accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing surface normal vectors and data acquisition vectors for each point in the point cloud data before the merging process. This preparation enables the subsequent matching algorithm to efficiently differentiate between points on perpendicular surfaces and resolve ambiguities in data-sparse regions, thereby improving localization accuracy without significantly increasing the complexity of the merging operation itself.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent transitions from using only position information (3D coordinates) to incorporating additional dimensional attributes - specifically surface normal vectors and data acquisition vectors. These additional dimensions provide extra discriminatory power for distinguishing between geometrically similar points, enabling more accurate localization and mapping while maintaining computational feasibility through efficient vector-based comparisons.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If points on perpendicular surfaces are differentiated using additional metrics, then mapping accuracy improves, but computational complexity increases

Engineering Contradiction:

Improvemapping accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing surface normal vectors and data acquisition vectors for each point in the point cloud data before the merging process. This preparation enables the subsequent matching algorithm to efficiently differentiate between points on perpendicular surfaces and resolve ambiguities in data-sparse regions, thereby improving localization accuracy without significantly increasing the complexity of the merging operation itself.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces surface normal vectors and data acquisition vectors as intermediary elements that mediate the matching process between point cloud data and existing maps. These intermediary metrics serve as additional criteria for differentiation, enabling the system to distinguish between points on perpendicular surfaces and resolve ambiguities in data-sparse regions, thereby improving mapping accuracy while maintaining computational feasibility through efficient vector-based comparisons.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20200400442A1Methods and systems for filtering point cloud map data for use with acquired scan frame data

Publication Date: 2020.12.24 CARNEGIE MELLON UNIV

AI summary

A method includes acquiring with a scanning device a scan frame comprising a point cloud comprising a first plurality of points, attributing each of the first plurality of points with a selected metric, filtering an existing map comprised of a second plurality of points based, at least in part, on the selected metric and localizing each of the first plurality of points to the filtered second plurality of points.