Surveillance Device for Automated Transaction Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surveillance systems for automated transaction devices face challenges in accurately detecting anomalous behavior, leading to false reporting and inconsistent results due to varied human actions on a single operation section.
Innovation Solution
A surveillance device that learns and holds reference scene data sets for normal actions, extracts actual target action data from video images, and compares it to determine anomaly levels, outputting an anomalous occurrence signal for precise detection and response.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a surveillance device monitors various human actions on a single operation section, then the coverage of monitored actions is improved, but the precision of anomalous action detection deteriorates due to inconsistent human behavior patterns
Solution Approach 1:
The operation section is divided into multiple analysis regions (e.g., card insertion region, screen operation region, keyboard region). Each region is monitored independently with its own action recognition model, allowing the system to handle diverse actions while maintaining detection precision through localized analysis rather than treating the entire operation section as a single unit.
Solution Approach 2:
Reference scene data sets are learned and stored in advance for normal actions in each analysis region. The system pre-establishes what constitutes normal behavior patterns for each region, enabling accurate comparison and anomaly detection when actual actions are monitored, thereby maintaining precision despite the variety of monitored actions.
2Adaptability or versatility
If a surveillance control system monitors a large number of people and events, then the monitoring coverage is improved, but false reporting of anomalous action detections increases
Solution Approach 1:
By segmenting the monitoring system into multiple analysis regions with dedicated action recognition models, the system reduces false reporting. Each region's model is specialized for that specific area's normal actions, making it easier to distinguish true anomalies from normal variations in behavior patterns across different regions.
Solution Approach 2:
Different analysis regions have different reference scene data sets tailored to their specific functions. The card insertion region has reference data for card handling actions, while the keyboard region has reference data for typing actions. This local customization of monitoring parameters reduces false reporting by accounting for the unique behavior patterns of each region.
3Adaptability or versatility
If a surveillance device analyzes video images to detect various user actions, then the versatility of detection is improved, but the complexity of the detection system increases
Solution Approach 1:
The detection system is segmented into multiple analysis regions, each with its own action recognition model. This modular approach allows the system to handle diverse actions (inserting card, operating screen, pressing keys) by distributing detection tasks across multiple specialized models rather than using a single complex model, thereby managing system complexity while maintaining versatility.
Solution Approach 2:
The surveillance device employs a universal framework with multiple analysis regions that can handle various types of actions. Each region uses the same basic comparison mechanism (actual scene data vs. reference scene data set), but the framework is designed to be multi-functional, accommodating different action types across different regions without requiring entirely separate systems for each function.
Data Source
AI summary
The present disclosure provides a surveillance device that monitors an operation section of an automated transaction device, the surveillance device including: a learning holding section that learns and holds a reference scene data set in which a reference operation is divided into a sequence of action items; a feature extraction section that extracts actual target action data from actual scene data of the sequence of action items in an operation of a user, the actual scene data obtained from an imaging section that faces and images the operation section; and a detection section that associates actual target action data with a reference scene data set along the sequence of action items, compares for each of the action items, determines an anomaly level of the operation of the user, and outputs an anomalous occurrence signal according to the anomaly level.