Background Subtraction Re-initialization for Sensor Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for background subtraction in intelligent physical spaces face challenges in accurately updating background representations when there are changes in the set of active sensors, leading to potential inaccuracies in object detection and space mapping.
Innovation Solution
A computing system determines a background representation of a space using sensor data from active sensors, divides it into subspaces, and adjusts this representation by subtracting data points corresponding to moving objects after detecting changes in the sensor set, ensuring accurate reflection of the space's background.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the background representation is updated continuously to reflect changes in sensor configuration, then the accuracy of object detection is improved, but the computational complexity and processing time increase
Solution Approach 1:
The background representation is divided into multiple subspaces, each associated with specific sensor data points. When sensor configuration changes, only the affected subspaces need re-initialization rather than the entire background representation, significantly reducing computational complexity while maintaining detection accuracy.
Solution Approach 2:
The system dynamically adjusts the background representation by re-initializing only the necessary subspaces when sensor configuration changes are detected. This dynamic approach allows the system to adapt to changing conditions without performing unnecessary computational operations across the entire background model.
2Reliability
If the background representation is re-initialized after sensor changes, then the reliability of space mapping is improved, but the system response time increases
Solution Approach 1:
By segmenting the background into subspaces linked to specific sensors, the system can quickly identify and re-initialize only the affected subspaces rather than the entire background model, reducing the time required to maintain reliable space mapping after sensor changes.
Solution Approach 2:
The system maintains a mapping between subspaces and sensor data points in advance, enabling rapid identification of which subspaces need re-initialization when sensor changes occur, thus reducing response time while ensuring reliability.
3Quantity of substance
If all sensor data points are used to depict the background, then the completeness of space representation is improved, but the accuracy of moving object detection deteriorates due to interference from static background elements
Solution Approach 1:
The background representation is segmented into subspaces, allowing the system to distinguish between data points representing static background elements and those representing moving objects. This segmentation enables accurate object detection by analyzing changes in specific subspaces while maintaining complete spatial representation.
Solution Approach 2:
The system extracts and separates data points corresponding to moving objects from the static background representation by analyzing subspace changes over time. This extraction allows the system to maintain complete space representation while accurately detecting moving objects by focusing on temporal variations in subspace data points.
Data Source
AI summary
Example methods and systems for background subtraction re-initialization are disclosed. In an example method, computing system may use a set of active sensors to determine a background representation of a space that can be divided into subspaces with each subspace depicted by a quantity of data points that depends on the active sensors. The system may determine locations of objects moving the space using the active sensors and the background representation. In some instances, the system may receive an indication of a change in the active sensors and responsively determine new quantities of data points that depict each subspace after the change. The system may further adjust the background representation of the space by subtracting data points corresponding to the determined locations of the objects moving in the space from the new quantities of data points that depict each subspace of the space after the change in the active sensors.