Contextual Privacy Filters for Hash-Based Moving-Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workplace monitoring systems struggle to track moving objects while maintaining privacy by anonymizing data effectively, leading to potential breaches of personal information.
Innovation Solution
A method involving sensor blocks and a computer system that utilize cryptographic hash functions to transform object and human features into hash values, storing these in containers, and managing collisions to ensure anonymization, thereby generating anonymized trajectories and metrics without revealing identifying information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If workplace monitoring systems track moving objects to derive metrics such as occupancy counts and motion statistics, then productivity and monitoring capability are improved, but privacy is compromised and identifying information may be exposed
Solution Approach 1:
The patent extracts only the necessary tracking information (location, motion轨迹) from the original visual data, separating the useful monitoring metrics from the identifying information. By taking out only the essential movement data and discarding or anonymizing identifying features, the system achieves monitoring capability while preventing privacy breaches.
Solution Approach 2:
The patent creates a copy of the tracking data in an anonymized format, where the trajectory information is preserved but linked to anonymous identifiers rather than real identities. This copy allows metric derivation without exposing identifying information, as the copy contains only the necessary movement data stripped of personal identifiers.
2Object-affected harmful factors
If cryptographic hash functions are used to transform object and human features into hash values for anonymization, then privacy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the computational task by implementing hash functions at the sensor block level rather than requiring centralized processing. Each sensor block independently computes hashes for detected objects and humans, dividing the overall computational burden into smaller, distributed operations. This segmentation reduces the complexity burden on any single system component while maintaining privacy protection.
Solution Approach 2:
The patent changes the parameter of data representation from original feature vectors to compressed hash values. By transforming the data into a different parameter space (hash domain), the system achieves privacy protection with reduced data size and computational requirements for subsequent processing, as hash values are more compact than original feature sets.
3Measurement precision
If features are transformed into hash values and stored in containers to enable tracking, then measurement precision for anonymized trajectories is improved, but data management complexity increases
Solution Approach 1:
The patent introduces hash values as an intermediary between the original identifying features and the anonymized trajectory data. This intermediary layer allows precise tracking through consistent hash identification while maintaining anonymity. The hash values serve as mediators that link trajectory segments across different sensor blocks without exposing the underlying identifying information, thus achieving measurement precision without direct exposure of personal data.
Solution Approach 2:
The patent adds a new dimension to data storage by organizing information in a multi-dimensional structure: sensor blocks generate local trajectories, which are then aggregated into a global trajectory dimension. By storing data across multiple dimensions (local sensor data, hashed identifiers, global trajectory context), the system achieves precise anonymized tracking while distributing the data management complexity across hierarchical levels rather than concentrating it in a single complex database.
Data Source
AI summary
A method includes: accessing a first frame; detecting a first object in the first frame; extracting a first location and a first set of features of the first object; transforming the first set of features into a first set of hash values; storing the first set of hash values in a first object container in a set of object containers; accessing a second frame; detecting a second object in the second frame; extracting a second location and a second set of features of the second object; transforming the second set of features into a second set of hash values; querying the set of object containers for the second set of hash values; and in response to a second object container containing a threshold quantity of hash values in the second set of hash values, identifying the second object as previously detected and represented by the second object container.