ATM Event Detection via Skeletal Similarity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting bank transfer fraud at ATMs require extensive training data and may infringe on user privacy, making them inefficient and costly.
Innovation Solution
An event detection system that calculates the similarity between skeletal information from captured images and registered phone call actions using a database, determining the occurrence of a suspicious event when the similarity threshold is met, without requiring specific pixel information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If learning-based action analysis is used to detect phone call actions, then detection capability is improved, but preparation cost and time increase due to large amounts of training data required
Solution Approach 1:
The patent extracts only the essential skeletal information (positions and movements of key body parts) from the imaging information, separating it from unnecessary pixel data. This extraction allows detection to proceed without requiring large training datasets, thus reducing preparation time while maintaining detection capability.
Solution Approach 2:
The patent creates simplified skeletal models that copy only the essential movement patterns needed for detection. Instead of using complex pixel-level data requiring extensive training, the system uses simplified skeletal representations that can be processed with minimal training data, reducing both time and computational resources.
2Measurement precision
If pixel information is retained for accurate action detection, then detection precision is improved, but user privacy is compromised
Solution Approach 1:
The patent extracts only skeletal information (positions and movements of key body parts) from the imaging data, deliberately excluding detailed pixel information. This extraction maintains sufficient precision for detecting phone call actions while removing personally identifiable visual features, thus protecting user privacy.
Solution Approach 2:
The skeletal information serves as an intermediary representation between the original imaging data and the detection process. It preserves the essential movement patterns needed for accurate detection while acting as a privacy-protecting layer that prevents direct analysis of personal appearance and identity features.
3Measurement precision
If traditional action analysis methods are used, then detection accuracy is improved, but system complexity and data processing requirements increase
Solution Approach 1:
The patent segments the imaging information processing into distinct stages: extracting skeletal information first, then analyzing only that skeletal data for action detection. This segmentation simplifies the overall system by breaking down the complex task of full image analysis into manageable steps, reducing computational requirements while maintaining accuracy.
Solution Approach 2:
The patent extracts and isolates only the relevant skeletal information needed for detection, removing unnecessary pixel data processing from the system. This extraction simplifies the data processing pipeline and reduces system complexity while preserving the essential information needed for accurate action detection.
Data Source
AI summary
An event detection system (10) includes: a calculation unit (16) that calculates a degree of similarity between at least a part of skeletal information extracted from a captured image in which a user who is visiting an ATM is captured and at least a part of registration skeletal information, the registration skeletal information being extracted from a registration image showing a phone call action of a person and being registered in an action database; and a determination unit (17) that determines that an event related to the ATM has been detected when the degree of similarity is equal to or greater than a predetermined threshold. Thus, a problematic event at an ATM can be easily detected while protecting privacy.