Camera Pose Estimation Using Obfuscated Image Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image-based localization techniques face challenges in maintaining privacy and security, particularly when transmitting images to remote devices for camera pose estimation, as they may reveal sensitive information such as user faces or confidential data, and existing methods do not adequately protect against unauthorized recreation of original images.
Innovation Solution
The method involves generating an obfuscated image representation by replacing the 2D positions of image features with 2D lines and retaining feature descriptors, which are then transmitted to a remote device for camera pose estimation, making it difficult for the remote device to recreate the original image while still allowing pose calculation based on correspondences with a 3D map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If original unobfuscated images are transmitted to remote devices for camera pose estimation, then pose estimation accuracy is improved, but privacy and security are compromised due to revelation of sensitive information such as user faces or confidential data
Solution Approach 1:
The patent extracts only the essential geometric information (2D lines passing through feature points) from the original image while leaving out the actual feature point coordinates and image content. This extracted obfuscated representation is then transmitted to the remote device, enabling pose estimation without exposing sensitive image data.
Solution Approach 2:
The patent introduces an intermediary obfuscated image representation that acts as a mediator between the original image and the pose estimation process. This intermediate representation contains sufficient geometric information for pose calculation but obscures the actual image content, thereby protecting privacy while enabling the desired function.
2Object-affected harmful factors
If 2D positions of image features are replaced with 2D lines to prevent image recreation, then privacy protection is improved, but the complexity of determining correspondences between obfuscated features and 3D map features increases
Solution Approach 1:
The patent changes the parameter representation from precise 2D point coordinates to 2D lines with associated feature descriptors. This parameter transformation maintains the essential geometric constraints needed for pose estimation while preventing exact image recreation, as the line representation is less informative than point coordinates.
3Reliability
If obfuscated features with 2D lines are used instead of original 2D points, then security against unauthorized image recreation is improved, but the ability to precisely locate image features is reduced
Solution Approach 1:
The patent uses partial information (2D lines passing through feature points rather than exact point coordinates) to achieve the security goal. The line representation provides sufficient constraint for pose estimation but deliberately withholds the precise location information needed for exact image recreation, representing a partial information approach that balances security and functionality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for estimating a camera pose includes recognizing a three-dimensional (3D) map representing a physical environment, the 3D map including 3D map features defined as 3D points. An obfuscated image representation is received, the representation derived from an original unobfuscated image of the physical environment captured by a camera. The representation includes a plurality of obfuscated features, each including (i) a two-dimensional (2D) line that passes through a 2D point in the original unobfuscated image at which an image feature was detected, and (ii) a feature descriptor that describes the image feature associated with the 2D point that the 2D line of the obfuscated feature passes through. Correspondences are determined between the obfuscated features and the 3D map features of the 3D map of the physical environment. Based on the determined correspondences, a six degree of freedom pose of the camera in the physical environment is estimated.