3D Point-Cloud Localization with Adaptive LSH Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for simultaneous localization and mapping (SLAM) in 3D point clouds face inefficiencies in real-time relocalization due to the vast amount of data processing, particularly in feature matching, which is not adequately addressed by current Locality Sensitive Hashing (LSH) techniques.

Innovation Solution

A novel method utilizing online learning to select hash keys that minimize a cost function, aiming for uniform bucket sizes and high collision rates in Locality Sensitive Hashing (LSH) to improve the efficiency of approximate nearest neighbor (ANN) search during SLAM, by redistributing bit selection within the SLAM process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional Locality Sensitive Hashing (LSH) is used for feature matching in SLAM relocalization, then the method can handle large 3D point clouds, but the processing time is too long to achieve real-time performance

Engineering Contradiction:
Improvereal-time relocalization speedVSAvoidfeature matching processing time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-selecting and storing hash keys during the mapping phase that are likely to be useful for future relocalization. The system proactively builds a hash key table with precomputed binary descriptors and hash values, so that during relocalization, the system only needs to query and compare against these pre-prepared keys rather than processing all 3D points from scratch. This advance preparation significantly reduces the time penalty during actual relocalization operations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent extracts only the essential features needed for relocalization by selecting specific hash keys and their corresponding binary descriptors from the full 3D point cloud data. Instead of using all available 3D points and features, the system extracts and stores a curated subset of hash keys with their binary representations in a hash key table. This extraction process isolates the most discriminative features while discarding redundant information, thereby reducing processing time during relocalization while maintaining matching accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

2Productivity

If binary feature descriptors are used to reduce data amount, then processing speed improves, but matching accuracy may be compromised

Engineering Contradiction:
Improvefeature matching efficiencyVSAvoidfeature matching accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies parameter changes by dynamically adjusting the hash key selection criteria and binary descriptor parameters based on the specific relocalization scenario. The system modifies parameters such as hash table size, number of hash functions, and binary descriptor bit length to optimize the balance between processing speed and matching accuracy. By changing these parameters adaptively rather than using fixed values, the system can achieve real-time performance while maintaining sufficient matching precision for the application requirements.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3251051B1Localization and mapping method
Publication Date: 2025.10.01 NOKIA TECHNOLOGIES OY
  • EP3251051B1 patent drawingFigure 1~2
  • EP3251051B1 patent drawingFigure 3~4
  • EP3251051B1 patent drawingFigure 5a~5c

AI summary

A method comprising: obtaining a three-dimensional (3D) point cloud about an object; obtaining binary feature descriptors for feature points in a 2D image about the object; assigning a plurality of index values for each feature point as multiple bits of the corresponding binary feature descriptor; storing the binary feature descriptor in a table entry of a plurality of hash key tables of a database image; obtaining query binary feature descriptors for feature points in a query image; matching the query binary feature descriptors to the binary feature descriptors of the database image; reselecting one bit of the hash key of the matched database image; and re-indexing the feature points in the table entries of the hash key table of the database image.