Adaptive Event Definition Generation for Video Surveillance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for detecting complex events, such as human behaviors interacting with inanimate objects, face challenges in achieving real-time accurate detection due to the exponential growth of digital data and the need for predefined object detection aspects, leading to outdated systems and high false positives, especially in environments with vast video feeds and sensor data.
Innovation Solution
A system that utilizes video signals to generate complex definitions of human behavior and interconnecting relationships, employing a rating system to log and refine event definitions through expert review, enabling real-time monitoring and notification of specific activities like card counting, with a plug-in architecture for dynamic event detection and automation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection systems are used with predefined detection aspects, then the system structure is simple and easy to deploy, but the detection accuracy decreases and the system becomes outdated quickly as events and behaviors change rapidly
Solution Approach 1:
The system dynamically generates event definitions from video data rather than using static predefined rules. The event definitions are continuously updated based on learned patterns from historical video data, allowing the system to adapt to changing behaviors and events without manual reconfiguration, thus maintaining high detection accuracy in dynamic environments
Solution Approach 2:
The system performs self-learning by automatically analyzing video data to generate and refine event definitions without requiring external expert intervention for each new event type. The automated generation of event definitions from video data allows the system to independently adapt to new patterns, reducing the need for manual system updates while maintaining high detection accuracy
2Measurement precision
If manual review of all video feeds is performed by security staff, then detection accuracy can be high, but the processing time increases and real-time detection becomes impossible due to the exponential growth of digital data
Solution Approach 1:
The system replaces manual mechanical review of video feeds by security staff with automated computer-based event definition generation and matching. The automated system processes video data to generate event definitions and performs real-time matching, achieving both high accuracy and high processing speed that cannot be achieved through manual review of exponential data growth
Solution Approach 2:
The system introduces event definitions as an intermediary layer between raw video data and detection results. By generating compact event definitions from video data and using these as matching criteria, the system efficiently filters and processes large volumes of video data in real-time without requiring manual review, thus maintaining both accuracy and processing speed
3Adaptability or versatility
If predefined event detection rules are used, then the system is easy to implement initially, but the adaptability decreases when new types of threats or actions are identified
Solution Approach 1:
The system performs preliminary analysis of video data to generate event definitions before deployment. By pre-processing video data to extract patterns and generate event definitions in advance, the system is ready to detect new event types immediately upon deployment without requiring complex manual configuration, thus maintaining both ease of implementation and high adaptability
Solution Approach 2:
The system uses a universal event definition generation mechanism that can handle multiple types of events and behaviors through the same automated process. The event definition generation system is not limited to specific event types but can adapt to detect various threats and actions uniformly, providing high versatility without increasing implementation complexity
Data Source
AI summary
Data from one or more sensors are used to detect an event corresponding to an activity by a person. The detection of the event is then used to loosely model the event to produce an initial event definition. The initial event definition is used in searching historical sensor data to detect other prospective occurrences of the event. Based on the detection of these additional occurrences, the event definition is refined to produce a more accurate event definition. The resulting refined event definition can then be used with current sensor data to accurately detect when the activity is being carried out. For example, the activity may be the use of a video imaging device for cheating, and a casino can use the refined event definition with video imaging to detect when people are using the video imaging device for such purposes.