Device Pose Estimation via Surface Normal Frequency Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies face challenges in accurately estimating the pose of electronic devices using image data, particularly in environments with complex surfaces, as they often rely on external sensors or fail in scenarios where simultaneous localization and mapping (SLAM) is ineffective.
Innovation Solution
The method involves determining surface normal frequency data from image data using image sensors, analyzing this data to identify relative maximums, and employing spherical coordinate transformations to create histograms that allow for the estimation of device orientation without external sensory data, such as inertial measurement units (IMUs).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external sensors (such as IMUs) are used for pose estimation, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts the pose estimation function from external sensors and implements it purely through image processing algorithms. By analyzing surface normal vectors and frequency data from image data alone, the system eliminates the need for external sensors like IMUs, thereby reducing device complexity while maintaining pose estimation capability
Solution Approach 2:
The patent replaces the mechanical sensor-based pose estimation system with an optical/image processing-based system. Instead of using physical sensors to detect orientation, the system uses computational analysis of image data and surface normal vectors to derive pose information, substituting a mechanical approach with a computational one
2Adaptability or versatility
If SLAM algorithms are used for pose estimation, then adaptability is improved, but reliability deteriorates in complex surface environments
Solution Approach 1:
The patent changes the fundamental parameters used for pose estimation from global mapping approaches (SLAM) to local surface normal vector analysis. By transforming image data into surface normal frequency space and identifying dominant orientations through spectral analysis, the system achieves reliable pose estimation in complex environments where traditional SLAM fails
Solution Approach 2:
The patent applies spherical coordinate transformations and spherical harmonic analysis to handle the curvature and complexity of surface normals in three-dimensional space. This mathematical transformation allows the system to effectively analyze oriented surfaces in complex environments by mapping them into a spherical frequency domain, improving reliability in scenarios with curved or irregular surfaces
3Device complexity
If image processing algorithms are used without external sensors, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent transitions from analyzing image data in spatial domain to analyzing it in frequency domain through spherical harmonic transforms. By adding this dimensional transformation, the system extracts more discriminative features from the image data, enabling accurate pose estimation without external sensors. The frequency domain representation reveals dominant surface orientations that are not obvious in the spatial domain
Data Source
AI summary
A method includes obtaining image data corresponding to a physical environment from an image sensor in an electronic device. The electronic device may determine surface normal frequency data based on the image data. The electronic device may determine an orientation of the electronic device in the physical environment based on the surface normal frequency data.