Video Fire Detection with Dynamic Region Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video-based fire detection systems face challenges in reducing false alarms and missed detections due to the dynamic nature of smoke and flames, as fixed regions of interest or masks do not adequately account for these variables, often requiring large areas to be masked, which can lead to unnecessary alarm suppression or missed detections.
Innovation Solution
The system defines user-defined acceptable regions within the field of view and associates rules with each region, allowing for the analysis of all video data to determine whether detected fire regions overlap or correlate with these areas, thereby deciding whether to trigger or suppress alarms based on predefined rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If large areas of the field of view are masked to avoid false alarms from smokestack emissions, then false alarms are reduced, but missed detections occur in the large masked areas
Solution Approach 1:
The patent segments the field of view into multiple regions, each with its own set of acceptable regions and associated rules. This allows different parts of the video feed to be processed with different sensitivity levels, enabling false alarm reduction in specific areas while maintaining detection capability in other areas.
Solution Approach 2:
The patent introduces dynamic acceptable regions that can change over time based on learned patterns. Instead of fixed masks, the system adapts the acceptable regions dynamically, allowing the detector to distinguish between transient smoke patterns (acceptable) and sustained fire conditions (not acceptable), thereby reducing false alarms without compromising detection accuracy.
2Reliability
If fixed ROIs or masks are applied to prevent false alarms, then false alarms are reduced, but the system cannot account for the dynamic nature of smoke and flames
Solution Approach 1:
The patent implements dynamic acceptable regions that evolve over time based on observed smoke behavior patterns. The system learns from temporal patterns to distinguish between transient emissions (acceptable) and sustained fire conditions, enabling adaptation to dynamic smoke behavior while maintaining false alarm reduction.
Solution Approach 2:
The system incorporates feedback mechanisms where detected smoke patterns are analyzed over time to update acceptable region definitions. This feedback loop allows the system to learn from actual fire conditions and adjust its acceptable regions accordingly, improving adaptability to dynamic smoke behavior.
3Measurement precision
If the entire field of view is processed without masking, then detection coverage is maintained, but false alarms from expected smoke emissions occur
Solution Approach 1:
The patent applies local quality by defining acceptable regions with specific rules for different areas of the field of view. Each acceptable region has associated rules that determine whether detected smoke patterns should trigger alarms, allowing the system to maintain detection coverage everywhere while suppressing false alarms in areas where smoke is expected.
Data Source
AI summary
A video recognition system detects the presence of fire based on video data provided by one or more video detectors, but suppresses the triggering of an alarm in situations based on the selection of acceptable regions and application of rules associated with each acceptable region. A user defines acceptable regions within the field of view of the video detector and associates with each acceptable region a rule. During processing of video data associated with the field of view, video metrics are calculated and analyzed to detect the presence of fire (e.g., flame or smoke). Prior to triggering an alarm, regions identified as indicative of fire are compared with the user-defined acceptable regions. If there is overlap between the two regions, the rule associated with the acceptable region is applied to determine whether the alarm should be suppressed or triggered.