Dynamic Privacy Masking Using Calibrated Camera Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems face challenges in reliably protecting privacy, as object detectors may lack reliability in determining what and when objects should be masked, and conventional solutions often require manual intervention, increasing labor and costs.
Innovation Solution
The system applies a mask to objects in images or video using privacy thresholds determined by the object's geometry, including its orientation and depth of view, allowing for different privacy levels at various points on the object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If object detectors are used to detect sensitive image areas for masking, then privacy protection is achieved, but reliability in determining what and when objects should be masked deteriorates
Solution Approach 1:
The patent changes the parameter basis for masking from object detection confidence scores to geometric parameters (depth, orientation, position) that are more reliable and objective. By using calibrated camera geometry to determine mask application, the system achieves more consistent and reliable privacy protection without losing important contextual information.
2Reliability
If manual intervention is used to select objects for redaction, then privacy protection accuracy is improved, but labor and costs increase
Solution Approach 1:
The system enables automatic privacy protection by having the camera system self-calibrate and self-determine masking parameters based on geometric analysis. The calibrated camera automatically identifies when and what to mask using depth and orientation calculations, eliminating the need for manual intervention while maintaining high accuracy and enabling parallel processing of multiple cameras.
3Reliability
If background modeling is used to detect unusual areas for masking, then privacy protection is achieved, but computing resources and processing time increase
Solution Approach 1:
The patent extracts the essential geometric parameters (depth, orientation, position) directly from calibrated camera data without performing continuous background modeling or learning. By taking only the necessary geometric information needed for mask determination, the system achieves reliable privacy protection with minimal computational overhead, avoiding the resource-intensive iterative background learning process.
4Adaptability or versatility
If a fixed mask size is applied to objects, then privacy protection is simplified, but adaptability to different situations deteriorates
Solution Approach 1:
The patent makes the mask parameters dynamic by calculating them based on real-time geometric parameters of the object and camera position. The mask size, shape, and position are dynamically adjusted according to the object's depth, orientation, and distance from the camera, allowing the system to adapt to different situations automatically without complex manual configuration.
Data Source
AI summary
A system and method for applying a mask to an object in a set of images. A computing system identifies a geometry of the object in an image from the set of images. Privacy thresholds are determined for each of a plurality of points on the object in the image using the geometry of the object. A mask is applied to the object in the image using the privacy thresholds, such that the mask provides different privacy at different points on the object. The geometry includes an orientation of the object and/or a depth of view of the object.