Archived Background Modeling for Privacy-Protected Live Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for learning a background model in video surveillance require significant computing resources and time, which is inefficient and increases the likelihood of personal identifiable information being present in the background model, leading to privacy concerns.
Innovation Solution
A method to generate a background model using archived video frames from a data repository, allowing for on-demand privacy protection of live video streams by processing these frames in faster-than-real-time to minimize delay and reduce identifiable information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a background model is learned using conventional methods with sufficient image frames, then privacy protection quality is improved, but processing time and computing resources increase significantly
Solution Approach 1:
The system pre-generates background models using archived video frames before live streaming begins. This preliminary action allows the background model to be learned in advance using historical data, so that when live streaming starts, the privacy protection can be applied immediately without waiting for real-time background learning to complete.
Solution Approach 2:
The system segments the background modeling process from the live streaming process by using archived frames separately from live frames. This segmentation allows independent optimization of each process - archived frames can be processed extensively for quality while live frames receive immediate protection.
2Speed
If background model learning is performed in real-time with live video frames, then processing delay is reduced, but the quality and reliability of privacy protection deteriorates due to insufficient data
Solution Approach 1:
The system performs background model learning in advance using archived video frames stored in a database. This preliminary action ensures that a high-quality background model is available before live streaming begins, eliminating the need to wait for real-time learning while maintaining privacy protection quality.
Solution Approach 2:
The system creates a copy of the background model from archived frames and applies it to live streaming. This copying approach allows the system to use pre-processed, high-quality background data without the delays of real-time learning, while still providing effective privacy protection for live content.
3Reliability
If more image frames are processed to create a better background model, then identifiable information is better obscured, but computing resources and processing time increase
Solution Approach 1:
The system processes and learns from archived video frames in advance when computing resources are more abundant and time is not critical. This preliminary processing creates a high-quality background model that effectively obscures identifiable information, while the actual live streaming requires minimal real-time computing resources.
Solution Approach 2:
The system periodically updates background models using archived frames at scheduled intervals rather than continuously processing live frames. This periodic action allows comprehensive processing of multiple frames to improve obscuration quality while reducing overall computing resource consumption compared to continuous real-time processing.
Data Source
AI summary
Methods, systems, and computer-readable media for producing a privacy-protected video stream are described herein. A request to display a live video stream of a camera is received. The live video stream is received in real-time and comprises a plurality of live image frames from the camera. An archived video stream of the camera is accessed in a data repository. A plurality of archived image frames of the archived video stream is processed to generate a background model comprising imagery common to multiple ones of the plurality of archived image frames. The plurality of archived image frames occurring in time prior to the request to display the live video stream. The privacy-protected video stream is processed in real-time. The privacy-protected video stream is output for display.