ATM Fraud Detection via Face and Shoulder Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Security guards at ATMs often fail to notice users being swindled due to limited observation capabilities, leading to property loss, especially in crowded environments.
Innovation Solution
A method and device for sending an alarm message that involves acquiring detection images, determining a target detection area using face and head and shoulder detection algorithms, and sending an alarm to a server if a person is detected to be on the phone, with features like reducing or enlarging detection areas based on detection results and using pre-stored training samples to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If security guards manually observe users at ATM, then they can provide human judgment and intervention, but their limited observation ability causes them to miss swindled users in crowded environments
Solution Approach 1:
The patent replaces the mechanical human observation system with an automated image recognition system using deep learning algorithms. The system captures images via cameras, processes them through neural networks to detect face images and determine calling status, and automatically generates alarm messages when fraud is suspected, eliminating the limitations of human observation capacity.
Solution Approach 2:
The patent introduces an intermediate image processing system between the security guard and the user. The system includes image capturing apparatus, processing units for detecting face images and calling status, and a server for generating alarm messages. This intermediary automatically analyzes user behavior and alerts security personnel only when suspicious activity is detected.
2Area of stationary object
If the detection area is enlarged to cover more users, then observation coverage increases, but detection accuracy decreases due to including more non-target areas
Solution Approach 1:
The patent segments the detection process into multiple stages: first capturing a wide-area image, then using deep learning algorithms to identify and extract face images from the crowd, and finally analyzing only the relevant face regions for calling status. This segmentation allows the system to maintain both large coverage area and high detection accuracy by focusing computational resources on detected targets rather than processing entire images.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image. The image capturing apparatus captures the entire scene with uniform quality, but the processing unit applies enhanced local quality analysis only to detected face regions, using specialized algorithms to determine calling status from facial features and hand positions, while other areas receive minimal processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure discloses a method and device for sending an alarm message, and belongs to the field of computer technology. The method includes: acquiring a detection image captured by an image capturing apparatus; detecting a face image area and a head and shoulder image area in the detection image based on a preset face and head and shoulder detection algorithm model; determining a target detection area in the detection image according to the detection results of both the face image area and the head and shoulder image area; and detecting a person's call status information corresponding to an image in the target detection area according to a preset on-the-phone determination algorithm model, and sending a first alarm message to a server if the person's call status information shows that the person is on the phone. By adopting the method and device provided by the present disclosure, a security guard can be informed in time that there may be a swindled user, and to take measures to prevent the user from property.