AI-Based Smart ATM Security System with Face Recognition
ESP32 + AI Agent + n8n Automation + Telegram Voice Alerts + Google Sheets + ThingSpeak Cloud Dashboard
AI-Based Smart ATM Security System with Face Recognition
ESP32 + AI Agent + n8n Automation + Telegram Voice Alerts + Google Sheets + ThingSpeak Cloud Dashboard
1. Project Overview
Project Title
AI-Powered Smart ATM Security Monitoring System using ESP32-CAM, Face Recognition, n8n Automation, Telegram Voice Alerts, Google Sheets Logging, and ThingSpeak Cloud Analytics
Objective
Develop an intelligent ATM security system that:
Detects unauthorized persons near ATM.
Performs face recognition.
Captures and uploads images.
Sends instant Telegram alerts.
Generates AI voice notifications.
Logs events to Google Sheets.
Stores sensor data in ThingSpeak Cloud.
Uses AI to predict suspicious activity and power consumption.
Provides real-time monitoring dashboard.
2. System Architecture
+------------------+
| ESP32-CAM |
| Face Recognition |
+---------+--------+
|
|
WiFi / Internet Connection
|
v
+---------------------------------------+
| n8n Server |
| AI Agent + Automation Engine |
+---------------------------------------+
| | |
| | |
v v v
Telegram Bot Google Sheets ThingSpeak
Voice Alert Logging Dashboard
|
v
Security Personnel
3. Features
Security Features
Face Recognition
Recognizes:
Authorized ATM staff
Security guards
Maintenance personnel
Detects:
Unknown visitors
Suspicious loitering
Intrusion Detection
Using PIR sensor:
Human motion detection
ATM door opening detection
Night-time monitoring
Real-Time Alerts
Telegram notifications:
⚠ ATM ALERT
Unknown person detected
Location:
ATM Branch 03
Time:
11:42 PM
Confidence:
91%
Image Captured
Voice Alerts
AI-generated voice:
Warning.
Unauthorized person detected near ATM.
Please verify immediately.
Cloud Monitoring
ThingSpeak dashboard displays:
Motion events
Face recognition confidence
Temperature
Humidity
Power usage
Security score
4. Hardware Components
Component Quantity
ESP32-CAM AI Thinker 1
FTDI Programmer 1
PIR Motion Sensor HC-SR501 1
Magnetic Door Sensor 1
Buzzer 1
Relay Module 1
DHT22 Sensor 1
OLED Display 0.96" 1
LEDs 2
Jumper Wires Several
Breadboard 1
5V Adapter 1
5. Pin Connections
PIR Sensor
PIR ESP32
VCC 5V
GND GND
OUT GPIO13
Door Sensor
Sensor ESP32
One End GPIO12
Other End GND
Buzzer
Buzzer ESP32
Positive GPIO15
Negative GND
DHT22
DHT22 ESP32
VCC 3.3V
GND GND
DATA GPIO14
6. Circuit Schematic
+----------------+
| ESP32 CAM |
+----------------+
GPIO13 ------ PIR OUT
GPIO12 ------ Door Sensor
GPIO15 ------ Buzzer
GPIO14 ------ DHT22 DATA
5V ---------- Sensors VCC
GND --------- Sensors GND
7. System Workflow
Start
|
v
Initialize ESP32
|
Connect WiFi
|
Motion Detected?
|
Yes
|
Capture Image
|
Face Recognition
|
Known Face?
/ \
Yes No
| |
Log Trigger Alert
| |
Store Event
| |
Send Telegram
| |
Voice Alert
| |
Update Sheets
| |
ThingSpeak Update
|
Loop
8. Detailed Flowchart
+--------------------+
| Power ON |
+---------+----------+
|
v
+--------------------+
| Connect WiFi |
+---------+----------+
|
v
+--------------------+
| Motion Detection |
+---------+----------+
|
v
+--------------------+
| Capture Face |
+---------+----------+
|
v
+--------------------+
| Face Recognition |
+---------+----------+
|
+----+----+
| |
v v
Known Unknown
| |
v v
Log Telegram Alert
|
v
Voice Message
|
v
Google Sheets
|
v
ThingSpeak
9. ESP32 Source Code (Core Example)
#include
#include
const char* ssid = "YOUR_WIFI";
const char* password = "YOUR_PASSWORD";
#define PIR_PIN 13
#define BUZZER 15
void setup()
{
Serial.begin(115200);
pinMode(PIR_PIN, INPUT);
pinMode(BUZZER, OUTPUT);
WiFi.begin(ssid,password);
while(WiFi.status()!=WL_CONNECTED)
{
delay(500);
}
}
void loop()
{
if(digitalRead(PIR_PIN))
{
digitalWrite(BUZZER,HIGH);
HTTPClient http;
http.begin("YOUR_N8N_WEBHOOK");
http.addHeader("Content-Type",
"application/json");
http.POST("{\"event\":\"motion\"}");
http.end();
delay(5000);
digitalWrite(BUZZER,LOW);
}
}
10. Face Recognition Logic
Face Enrollment
Store authorized faces:
Security Guard
ATM Manager
Maintenance Engineer
Cash Loading Staff
Recognition Process
Camera Capture
↓
Face Detection
↓
Feature Extraction
↓
Embedding Generation
↓
Database Comparison
↓
Known / Unknown
11. n8n Automation Workflow
Workflow Modules
Webhook Trigger
↓
Receive ESP32 Event
↓
AI Agent
↓
Decision Engine
↓
Telegram Alert
↓
Google Sheets
↓
ThingSpeak Update
↓
Voice Notification
Sample n8n JSON Structure
{
"nodes": [
{
"name": "Webhook",
"type": "n8n-nodes-base.webhook"
},
{
"name": "Telegram",
"type": "n8n-nodes-base.telegram"
},
{
"name": "Google Sheets",
"type": "n8n-nodes-base.googleSheets"
}
]
}
12. Telegram Bot Setup
Step 1
Open Telegram.
Search:
@BotFather
Step 2
Create Bot
/newbot
Step 3
Provide:
ATM Security Bot
Step 4
Receive:
BOT TOKEN
Example:
123456:ABCDEFxxxxx
Save it.
Step 5
Get Chat ID
Use:
@getidsbot
Save Chat ID.
13. Telegram Alert Automation
Message:
🚨 ATM SECURITY ALERT
Unknown Person Detected
Location:
ATM Branch Hyderabad
Confidence:
93%
Timestamp:
2026-06-11 22:05:00
Image Attached
14. Telegram Voice Alert Automation
n8n Flow
Alert Generated
↓
AI Text
↓
Text-To-Speech API
↓
Generate MP3
↓
Telegram Send Audio
Voice:
Attention.
Suspicious activity detected near ATM.
Immediate verification required.
15. Google Sheets Integration
Columns
Timestamp Face Status Confidence Motion Power
Example:
|2026-06-11|Unknown|94%|Detected|12W|
n8n Node
Google Sheets Append Row
Stores every event.
16. ThingSpeak Setup
Create Account
Use the official platform:
ThingSpeak
Create Channel
Fields:
Field1 Motion
Field2 Face Confidence
Field3 Temperature
Field4 Humidity
Field5 Power Usage
Field6 Security Score
API Write URL
https://api.thingspeak.com/update
17. ThingSpeak Dashboard
Widgets:
Gauge
Security Score
Line Chart
Power Usage
Bar Chart
Motion Events
Heat Map
Intrusion Frequency
18. AI Agent Logic
The AI Agent inside n8n evaluates:
Motion Frequency
Face Confidence
Time of Day
Door Status
Power Consumption
Risk Score Formula
Risk Score=0.4M+0.3F+0.2D+0.1T
Where:
M = Motion Risk
F = Face Risk
D = Door Status Risk
T = Time Risk
Classification
Score Status
0-30 Safe
31-60 Warning
61-100 High Risk
19. AI Power Consumption Prediction
Inputs
Temperature
Camera ON Time
WiFi Usage
Sensor Activity
Alert Frequency
Linear Prediction Model
P=aT+bC+cW+dS+eA
Where:
P = Predicted Power
T = Temperature
C = Camera Usage
W = WiFi Activity
S = Sensor Usage
A = Alert Count
Output
Predicted Daily Consumption:
2.8 kWh
Expected Monthly:
84 kWh
20. AI Decision Engine
Example:
Unknown Face
+
Motion Detected
+
After Midnight
=
Critical Alert
Response:
Activate Buzzer
Capture 5 Images
Send Voice Alert
Notify Security Team
21. Security Enhancements
Multi-Factor Verification
Face Recognition
Motion Detection
Door Sensor
Anti-Spoofing
Detect:
Mobile screen attacks
Printed photographs
Video replay attacks
Techniques:
Blink detection
Head movement tracking
Depth estimation
22. Deployment Architecture
ATM Site
|
ESP32-CAM
|
Internet
|
n8n Cloud Server
|
+-------------+
| AI Agent |
+-------------+
|
Telegram
Sheets
ThingSpeak
23. Future Enhancements
AI Enhancements
YOLO person detection
Weapon detection
Crowd analysis
Behavior analytics
Anomaly detection
Cloud Enhancements
AWS IoT Core integration
Azure IoT Hub integration
MQTT broker support
Edge AI inference
ATM Security Upgrades
SMS backup alerts
Email escalation
Police notification workflow
Geofencing
Multi-ATM centralized dashboard
24. Expected Project Outcome
The completed system will:
✅ Detect suspicious ATM activity in real time
✅ Recognize authorized and unauthorized individuals
✅ Send instant Telegram text and voice alerts
✅ Log every event to Google Sheets
✅ Visualize security metrics on ThingSpeak
✅ Use AI-based risk analysis and power prediction
✅ Enable centralized monitoring through n8n automation
✅ Provide a scalable smart ATM security solution suitable for academic projects, research prototypes, and pilot deployments in banking environments.
Comments
Post a Comment