Video Surveillance Motion Filtering via Spatial Temporal Bounding Box Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera-equipped security systems are costly and require extensive computational resources and human intervention to accurately detect moving objects of interest, often producing many false detections due to environmental noise and lighting changes.
Innovation Solution
A video monitoring system that uses spatial and temporal filtering of bounding boxes to eliminate invalid detections, employing preprocessing techniques like frame subtraction and morphological processing, followed by machine learning for object classification, which can be deployed on stationary or autonomous cameras.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing camera-equipped security systems use comprehensive object detection and classification procedures, then detection accuracy is improved, but computational cost and system complexity increase significantly
Solution Approach 1:
The patent segments the object detection process into distinct stages: motion detection identifies potential objects, spatial filtering eliminates invalid detections based on location, temporal filtering removes false positives through time-based analysis, and finally classification is performed only on remaining candidates. This segmentation reduces overall system complexity while maintaining detection accuracy.
Solution Approach 2:
The patent applies preliminary filtering actions before the expensive classification step. Spatial filtering pre-processes bounding boxes by evaluating their location and size characteristics, and temporal filtering further pre-processes by analyzing motion patterns over time. These preliminary actions eliminate invalid detections before classification, reducing computational burden while preserving accuracy.
2Measurement precision
If existing systems perform comprehensive object classification on all detected patches, then detection accuracy is improved, but computational cost increases
Solution Approach 1:
The patent applies partial action by performing expensive classification operations only on a subset of detected objects rather than all patches. The filtering stages selectively identify and eliminate invalid detections, so classification is applied only to promising candidates. This partial application of classification maintains accuracy for valid objects while significantly reducing overall computational cost.
3Productivity
If existing systems classify all detected image patches, then object identification completeness is improved, but processing time increases
Solution Approach 1:
The patent segments the processing pipeline to handle different object types efficiently. Motion detection and filtering stages quickly process all detected patches to identify valid candidates, then classification is applied only to these candidates. This segmentation maintains identification completeness for valid objects while reducing overall processing time by avoiding unnecessary classification of invalid detections.
Solution Approach 2:
The patent performs preliminary filtering actions (spatial and temporal filtering) before classification to pre-identify valid candidates. This preliminary action reduces the number of objects requiring time-consuming classification, thereby reducing total processing time while maintaining completeness of identification for valid objects.
4Measurement precision
If existing systems use motion detection to find potential objects, then detection sensitivity is improved, but false detection rate increases
Solution Approach 1:
The patent segments the validation process into spatial filtering and temporal filtering stages. Spatial filtering immediately eliminates bounding boxes that fail location or size criteria, and temporal filtering further eliminates candidates that fail motion consistency tests over time. This segmentation maintains detection sensitivity by keeping all potential candidates initially, then systematically reduces false detections through multiple filtering passes.
Solution Approach 2:
The patent introduces intermediary filtering stages between motion detection and final classification. These intermediary filters (spatial and temporal) act as mediators that refine the set of potential objects by eliminating false positives. The filters preserve sensitivity to valid objects while reducing false detection rates through progressive validation.
Data Source
AI summary
A video monitoring method that involves determining motion changes in a set of video frames to find potential objects is described. One or more bounding boxes are defined around the potential objects. These bounding boxes are spatially and temporally filtered to eliminate potential object candidates, with only potential objects in the bounding boxes remaining after filtering being classified or identified.