Thermal Alert Discriminator Intensity Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Thermal video monitoring systems generate false alerts when snow falls from vehicles, which are often mistaken for dropped cargo or stopped vehicles, leading to increased operational costs and safety risks due to the overwhelming number of false alarms.
Innovation Solution
A method that evaluates the intensity of stationary foreground objects in thermal images by comparing them to background intensity, converting candidate alerts to actual alerts only when the object's intensity is higher than the background, and discarding alerts with high disorder thresholds, thereby distinguishing between relevant and irrelevant events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If thermal video monitoring systems trigger alerts for all detected stationary foreground objects, then the system achieves high sensitivity in detecting potential events, but the number of false alerts increases significantly
Solution Approach 1:
The patent changes the parameter used for alert conversion from simple detection to intensity comparison. By evaluating whether the intensity of a stationary foreground object is above the background intensity, the system distinguishes between relevant objects (stopped vehicles, dropped cargo) and irrelevant objects (snow, rain) that typically have lower or equal intensity to the background, thereby reducing false alerts while maintaining detection sensitivity
Solution Approach 2:
The patent introduces an intermediary evaluation step between detection and alert generation. The intensity comparison acts as a mediator that filters candidate alerts before they become actual alerts, providing a additional criterion that helps differentiate between true events and false events without requiring direct modification of the detection algorithm
2Reliability
If operators inspect all candidate alerts manually, then the system maintains high detection coverage, but operator efficiency decreases and actual alerts become less noticeable
Solution Approach 1:
The patent performs preliminary filtering of candidate alerts through automatic intensity evaluation before they reach the operator. By pre-converting obviously valid alerts (those with intensity above background) and pre-discarding obviously invalid alerts (those with intensity equal to or below background), the system reduces the burden on operators to inspect only the ambiguous cases, thereby maintaining detection coverage while improving operator efficiency
3Measurement precision
If the system converts all candidate alerts with intensity above background to actual alerts, then the system achieves high precision in identifying relevant events, but the number of remaining false alerts from other sources increases
Solution Approach 1:
The patent applies different conversion rules to different intensity conditions, creating a localized quality approach. Instead of a uniform conversion rule, the system evaluates each candidate alert's intensity relative to background and applies appropriate conversion (convert, discard, or require manual inspection) based on the specific intensity relationship, thereby achieving high precision in differentiation while minimizing remaining false alerts from sources with equal or lower intensity
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces false alarms by accurately differentiating between relevant alerts, such as stopped vehicles, and irrelevant events like falling snow, thereby improving operator efficiency and reducing costs associated with false alarms.
Implementation Method 1
thermal cameras can be used for automatic detection and recognition in video monitoring
Implementation Method 2
The wording 'thermal image' should be construed as an image representing amounts of infrared energy emitted by an object
Data Source
Figure 1A~1B
Figure 1C~1D
Figure 2
AI summary
A method (600) and an alert discriminator (730) for converting alerts in a video monitoring operation is provided. The method (600) comprises: triggering (S602) a candidate alert, in response to detecting, in a stream of thermal images of a scene, a stationary foreground object in the scene; evaluating (S604) an intensity of the detected stationary foreground object (220, 320, 420) in relation to a background intensity; and in response to the intensity of the detected stationary foreground object (220, 320, 420) being above the background intensity, converting (S606) the candidate alert to an actual alert.