RGB-D Camera SLAM System Pose Accuracy and Drift Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current SLAM systems for RGB-D cameras face challenges in accurately estimating camera motion and mapping due to noise, outliers, and quantization errors, which affect pose accuracy and drift reduction.

Innovation Solution

The proposed method employs a dense visual odometry approach using a probabilistic sensor model for keyframe generation and loop constraint detection, combined with entropy-based similarity measures and occlusion filtering, to optimize the keyframe graph and reduce the impact of erroneous loop constraints, while utilizing a σ-DVO algorithm to minimize photometric and geometric errors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If dense visual odometry is used to exploit all pixel information, then pose accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improvepose accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the visual data processing by separating dense pixel-level odometry computation from sparse feature-based loop closure detection. The frontend performs dense visual odometry on all pixels to achieve high pose accuracy, while the backend selectively processes keyframes for loop constraints, dividing the computational workload to manage complexity while maintaining precision.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system applies dense visual odometry to all pixels (excessive action) for odometry estimation, but only processes a subset of keyframes (partial action) for loop closure detection. This selective processing reduces overall computational complexity while maintaining the accuracy benefits of dense methods where most critical.

Inventive Principle:
Principle #16Partial or excessive action

2Reliability

If loop constraint detection is performed frequently to reduce drift, then trajectory accuracy is improved, but false constraints increase due to noise and outliers

Engineering Contradiction:
Improvetrajectory accuracyVSAvoidloop constraint accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The system introduces an entropy-based similarity measure as an intermediary filter before adding loop constraints. This intermediary assessment evaluates the quality of potential loop closures by measuring frame similarity, allowing the system to distinguish true loops from false matches caused by noise and outliers, thereby maintaining trajectory accuracy while avoiding erroneous constraints.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements feedback mechanisms where loop constraint quality is continuously evaluated using entropy-based similarity measures. Poor quality constraints are rejected or down-weighted, and the system adapts keyframe selection based on accumulated drift estimates, creating a feedback loop that maintains constraint accuracy while reducing the impact of false positives.

Inventive Principle:
Principle #23Feedback

3Loss of information

If all keyframes are processed to ensure complete mapping coverage, then mapping completeness is improved, but computational requirements increase

Engineering Contradiction:
Improvemapping completenessVSAvoidcomputational requirements
Core Design Contradiction:
Loss of informationVSUse of energy by moving object

Solution Approach 1:

The system extracts and processes only the most informative keyframes for loop closure detection rather than all frames. By taking out and selectively processing a subset of keyframes based on entropy-based similarity measures and drift estimates, the system maintains mapping completeness while significantly reducing computational requirements compared to processing every frame.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP3475917B1RGB-d camera based slam system and method thereof
Publication Date: 2022.01.26 ROBERT BOSCH GMBH
  • EP3475917B1 patent drawingFigure 1
  • EP3475917B1 patent drawingFigure 2
  • EP3475917B1 patent drawingFigure 3

AI summary

A visual SLAM system comprises a plurality of keyframes including a keyframe, a current keyframe, and a previous keyframe, a dual dense visual odometry configured to provide a pairwise transformation estimate between two of the plurality of keyframes, a frame generator configured to create keyframe graph, a loop constraint evaluator adds a constraint to the receiving keyframe graph, and a graph optimizer configured to produce a map with trajectory.