Anonymized Sensor Data Processing for Public Transport Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Collecting realistic training data for adaptable situation recognition algorithms in public transport vehicles is challenging due to data protection concerns, as existing methods like pixelation are ineffective and falsify data, making it difficult to record and store video and audio data without violating privacy laws.
Innovation Solution
A method that records sensor data using on-board units, processes it to remove personal references, and stores the data as training information, allowing for the identification of operating situations without violating data protection laws, using techniques like coarsening and pre-processing algorithms to ensure anonymity while maintaining situational recognition capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If pixelation methods are used to render faces unrecognizable in video data, then data protection is improved, but the training quality of situation recognition algorithms deteriorates because pixelation falsifies the recorded video data
Solution Approach 1:
The patent extracts only the necessary information for situation recognition while leaving out personal identification data. Instead of applying pixelation to the entire video data, the system processes and stores only the processed sensor data after removing personal references, thus maintaining training quality while ensuring data protection.
Solution Approach 2:
The patent applies preliminary data processing to remove personal references before storing the data for training. This preliminary action of data processing and anonymization is performed in advance, allowing the system to store data that is both protective of privacy and useful for training without needing to apply pixelation that would falsify the data.
2Quantity of substance
If video and audio data are recorded without consent to collect sufficient training data, then the quantity of training data is improved, but data protection compliance deteriorates
Solution Approach 1:
The patent introduces processed sensor data as an intermediary between the raw sensor data and the training data storage. The processed sensor data serves as a mediator that removes personal references while preserving situation information, allowing the system to store large quantities of data for training without violating data protection laws.
Solution Approach 2:
The patent changes the parameters of the sensor data through processing to remove personal references while maintaining situation recognition capabilities. By transforming the data parameters through anonymization processes, the system can store data in large quantities for training purposes while maintaining compliance with data protection regulations.
3Object-affected harmful factors
If sensor data is processed to remove personal references, then data protection is improved, but the complexity of data processing increases
Solution Approach 1:
The patent implements self-service through automated data processing that removes personal references as part of the standard recording and storage process. The system automatically processes sensor data to anonymize personal information while preserving situation data, reducing the need for manual intervention and complex manual processing procedures.
Data Source
Figure 1~2
AI summary
The invention relates to a method for providing training data (T) for adaptable situation recognition algorithms used for the automated detection of an operating situation of a vehicle (1) for public passenger transport, in whose passenger compartment (3) persons (P1) are transported and/or on whose route (2) persons (P) are traveling outside the vehicle (1). During vehicle operation, sensor data (V, A) relating to persons are acquired by at least one vehicle-side sensor unit (6, 7). Before data storage, the personal reference in the acquired sensor data (V, A) is weakened by data processing such that the processed sensor data (V', A') no longer allows for the identification of persons, but still enables the detection of an operating situation.The specific operating situation to be recognized is determined through human evaluation of the processed sensor data (V', A') and assigned to the processed sensor data (V', A') as annotation data. Finally, the processed sensor data (V', A') together with the assigned annotation data are stored as training data (T) for the situation recognition algorithms. This provides a method for making a large amount of realistic training data (T) available for adaptable situation recognition algorithms without raising data protection concerns.