Egomotion Estimation Using Landmark Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current egomotion estimation methods are computationally expensive, rely on prominent landmarks that may not always be available, and increase hardware costs with stereo vision approaches, necessitating a fast, accurate, and inexpensive solution for camera motion estimation in computer vision applications.

Innovation Solution

A system and method that compares differences between image frames using simple arithmetic operations, employing specialized hardware like GPUs and FPGAs, and compensates for camera motion to determine translation and rotation parameters, minimizing computational load and hardware requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If feature tracking and matching is used for egomotion estimation, then measurement precision is improved, but device complexity and computational cost increase

Engineering Contradiction:
Improveegomotion estimation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and utilizes prominent landmarks (such as lane markers, curbs, and road markings) from the image frames as key features for egomotion estimation. By focusing on these specific, easily detectable landmarks rather than performing comprehensive feature tracking across the entire image, the system achieves accurate motion estimation with reduced computational complexity.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the approach from tracking multiple feature points to detecting and matching specific geometric parameters of landmarks (such as line positions, angles, and distances). This parameter transformation simplifies the computational burden while maintaining estimation accuracy, as landmark parameters can be directly related to camera motion through geometric relationships.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If stereo vision is used for egomotion estimation, then measurement precision is improved, but device complexity and hardware costs increase

Engineering Contradiction:
Improveegomotion estimation accuracyVSAvoidhardware complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent makes the single camera system perform multiple functions: it both captures images for landmark detection and provides sufficient information for egomotion estimation through temporal analysis of landmark positions. This eliminates the need for additional stereo camera hardware while achieving comparable estimation accuracy through clever use of monocular visual cues.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Device complexity

If prominent landmarks are used for egomotion estimation, then device complexity is reduced, but reliability decreases when landmarks are not available

Engineering Contradiction:
Improvecomputational simplicityVSAvoidestimation reliability
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent implements a dynamic system that adapts its landmark detection and matching process based on the availability and quality of landmarks in the current image frames. When prominent landmarks are available, the system uses them for accurate estimation; when they are not available or are insufficient, the system can switch to alternative features or adjust its detection parameters, ensuring reliable operation across varying environmental conditions.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2570993B1Egomotion estimation system and method
Publication Date: 2021.06.16 HARMAN INT CHINA HLDG
  • EP2570993B1 patent drawingFigure 1
  • EP2570993B1 patent drawingFigure 2
  • EP2570993B1 patent drawingFigure 3

AI summary

A computer-implemented method for determining an egomotion parameter using an egomotion estimation system (100) is provided. First and second image frames are obtained. A first portion of the first image frame and a second portion of the second image frame are selected to respectively obtain a first sub-image and a second sub-image. A transformation is performed on each of the first sub-image and the second sub-image to respectively obtain a first perspective image and a second perspective image. The second perspective image is iteratively adjusted to obtain multiple adjusted perspective images. Multiple difference values are determined that respectively correspond to the respective difference between the first perspective image and the adjusted perspective images. A translation vector for an ego motion parameter is determined. The translation vector corresponds to one of the multiple difference values.