Electronic Device Localization via Visual-Inertial Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine vision techniques for locating devices in local environments lack sufficient resolution and efficiency, limiting the utility of location information for applications like augmented reality and simultaneous localization and mapping.
Innovation Solution
The use of both non-image sensor data, such as GPS, and image sensor data from multiple cameras to estimate and refine the device's pose, generating feature descriptors for spatial features and comparing them to stored descriptors to enhance location-based functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional location techniques are used, then device positioning is achieved, but the resolution is insufficient to enhance object identification
Solution Approach 1:
The patent combines multiple data sources including GPS coordinates, accelerometer data, gyroscope data, and camera imagery to create a comprehensive localization system. This multi-sensor fusion approach merges coarse location data with fine-grained visual features to achieve both broad coverage and high precision object identification
Solution Approach 2:
The system segments the localization problem into multiple components: initial coarse positioning using GPS, refinement using accelerometer and gyroscope data, and final precision localization using computer vision feature matching. Each segment addresses a specific resolution level, collectively achieving high-precision location determination
2Loss of time
If conventional location techniques are used, then device position is determined, but the time required limits the utility of location information
Solution Approach 1:
The system performs preliminary coarse positioning using GPS and inertial sensors before initiating computationally intensive computer vision processing. This preliminary action provides an initial location estimate immediately, while more precise visual localization proceeds in parallel, ensuring timely location information is available without sacrificing precision
Solution Approach 2:
The system dynamically adjusts processing priorities based on motion state. When the device is stationary or moving slowly, more computational resources are allocated to precise visual feature matching. When moving rapidly, the system relies more on inertial prediction and reduces visual processing intensity, maintaining responsiveness while preserving accuracy when conditions permit
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method is provided, comprising, in response to a reset event (502) at an electronic device (102), identifying at the electronic device (102) an origin (221) in a first frame of reference (220); identifying a first pose (201) of the electronic device in a second frame of reference (230) corresponding to a geographic frame of reference; concurrent with identifying the first pose, identifying (518) a change in position of the electronic device (102) in the first frame of reference (220) relative to the origin (221); and translating (520) the first pose (201) of the electronic device (102) to a second pose in the second frame of reference (230) based on the change in position in the first frame of reference (220).