Obfuscation Network for Privacy-Safe Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current security systems using CCTV cameras in multi-use facilities face limitations in tracking individuals due to the presence of identification information, requiring complex obfuscation processes to protect privacy, which hinders efficient image analysis for incident investigation.
Innovation Solution
A method and device for tracking objects by obfuscating images using an obfuscation network that generates unidentifiable images to humans but identifiable by a learning network, allowing detection and tracking of target objects while maintaining privacy, and matching obfuscated tracking information with non-obfuscated identification information upon consent.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If original images containing identification information are used for tracking, then tracking accuracy is improved, but privacy protection deteriorates
Solution Approach 1:
The system segments the image processing pipeline into two distinct parts: an obfuscation network that transforms original images into obfuscated versions preserving tracking features while removing identification information, and a detection network that operates on the obfuscated images. This segmentation allows simultaneous achievement of privacy protection and tracking accuracy by separating the privacy-protection function from the tracking function.
Solution Approach 2:
The obfuscated image serves as an intermediary representation between the original image and the tracking analysis. It contains sufficient information for accurate tracking of movement patterns and behaviors while eliminating personally identifiable features. This intermediary form enables privacy-safe tracking without requiring direct access to original identification-containing images.
2Object-affected harmful factors
If obfuscation operations are applied to protect identification information, then privacy protection is improved, but device complexity deteriorates
Solution Approach 1:
The system replaces traditional mechanical or manual obfuscation methods (such as pixelation, blurring, or manual redaction) with deep learning-based neural networks. The obfuscation network and detection network use automated machine learning processes to perform privacy protection and tracking, significantly reducing operational complexity and enabling real-time processing that would be infeasible with manual methods.
Solution Approach 2:
The system changes the fundamental parameter of image representation from original pixel data to obfuscated feature representations. By transforming images through learned parameter spaces that preserve movement-related features while discarding identification features, the system achieves privacy protection through parameter transformation rather than complex processing operations.
3Object-affected harmful factors
If obfuscated images are used for tracking, then privacy protection is improved, but detection capability deteriorates
Solution Approach 1:
The system performs preliminary training of the detection network using paired datasets of original images and their corresponding obfuscated versions. During this preliminary action phase, the network learns to recognize tracking-relevant features in the obfuscated representation. This pre-training ensures that when the system operates, the detection network is already optimized to extract meaningful tracking information from obfuscated images without difficulty.
Solution Approach 2:
The system implements a feedback mechanism where the detection network's performance on obfuscated images continuously informs and refines the obfuscation process. The networks are trained jointly with feedback loops that ensure the obfuscation preserves necessary tracking features while removing identification information, creating a self-optimizing system that maintains detection capability.
Data Source
AI summary
A method for tracking one or more objects in a specific space is provided. The method includes steps of: (a) inputting original images of the specific space taken from camera to an obfuscation network and instructing the obfuscation network to obfuscate the original images to generate obfuscated images such that the obfuscated images are not identifiable as the original images by a human but the obfuscated images are identifiable as the original images by a learning network; (b) inputting the obfuscated images into the learning network, and instructing the learning network to detect obfuscated target objects, corresponding to target objects to be tracked, in the obfuscated images, to thereby output information on the obfuscated target objects; and (c) tracking the obfuscated target objects in the specific space by referring to the information on the obfuscated target objects.