Pick Pocket Detection - Implementation

System Logic & Decision Making

"The core of this system is a Multi-Sensor Fusion Algorithm. To minimize false positives (like accidental pocket movement), the Raspberry Pi processes real-time data from four distinct sensors. An alert is only triggered if the system detects a Sum of Anomalies >= 2."

Anomaly Detection Criteria:

LDR: Light levels drop below the calibrated dark threshold (Pocket Removal).
Capacitive Touch: Loss of physical contact with the user's leg/fabric.
Accelerometer: Sudden G-force spike exceeding 2.5G (The Snatch).
Gyroscope: Rotational velocity exceeding 300°/sec (The Twist).

Software Implementation (Rough Code)

Part A: Smartphone Module (Python)

import RPi.GPIO as GPIO

import time, serial, threading, smbus2, json, pynmea2

import paho.mqtt.client as mqtt

# GPIO & Pin Setup

LDR_PIN, TOUCH_PIN, BUZZ_PIN = 18, 17, 27

GPIO.setmode(GPIO.BCM)

GPIO.setup(LDR_PIN, GPIO.IN)

GPIO.setup(TOUCH_PIN, GPIO.IN, GPIO.PUD_UP)

GPIO.setup(BUZZ_PIN, GPIO.OUT)

buzzer = GPIO.PWM(BUZZ_PIN, 1000)

buzzer.start(0)

# MPU6050 & MQTT Config

bus = smbus2.SMBus(1)

MQTT_TOPIC = "iot-pickpocket/final/gps"

client = mqtt.Client()

client.connect("broker.hivemq.com", 1883, 60)

client.loop_start()

def get_accel_mag():

try:

# Reading raw Accel data (X, Y, Z)

ax = (bus.read_byte_data(0x68, 0x3B) << 8 | bus.read_byte_data(0x68, 0x3C)) / 16384.0

ay = (bus.read_byte_data(0x68, 0x3D) << 8 | bus.read_byte_data(0x68, 0x3E)) / 16384.0

az = (bus.read_byte_data(0x68, 0x3F) << 8 | bus.read_byte_data(0x68, 0x40)) / 16384.0

return (ax**2 + ay**2 + az**2)**0.5

except: return 1.0

def send_bt_alert():

try:

with serial.Serial('/dev/rfcomm0', 9600, timeout=1) as bt:

bt.write(b'A') # Send Alert signal to Wristband

except: pass

# Main Monitoring Loop

alert_mode = False

while True:

mag = get_accel_mag()

light_anom = GPIO.input(LDR_PIN) == 0 # Light detected

touch_anom = GPIO.input(TOUCH_PIN) == 1 # Loss of contact

motion_anom = mag > 2.5 # Sudden snatch

if (light_anom + touch_anom + motion_anom) >= 2:

print("THEFT DETECTED!")

send_bt_alert()

alert_mode = True

time.sleep(5) # Cooldown

time.sleep(0.1)

Part B: Haptic Wristband (C++ / Arduino Nano)

#include <SoftwareSerial.h>

#include <Wire.h>

#include <Adafruit_GFX.h>

#include <Adafruit_SSD1306.h>

SoftwareSerial BT(2, 3); // RX, TX

Adafruit_SSD1306 display(128, 64, &Wire, -1);

void setup() {

BT.begin(9600);

pinMode(9, OUTPUT); // Vibration Motor

pinMode(4, INPUT_PULLUP); // Find Phone Switch

display.begin(SSD1306_SWITCHCAPVCC, 0x3C);

showStatus("READY");

}

void loop() {

// 1. Listen for Alert from Phone

if (BT.available()) {

if (BT.read() == 'A') {

triggerAlert();

}

// 2. Handle 'Find Phone' Switch

if (digitalRead(4) == LOW) {

BT.write('F'); // Send Find signal back to Pi

showStatus("FINDING...");

delay(1000);

showStatus("READY");

}

void triggerAlert() {

display.clearDisplay();

display.setTextSize(2);

display.setCursor(10, 20);

display.print("!!THEFT!!");

display.display();

for(int i=0; i<5; i++) {

digitalWrite(9, HIGH); delay(300);

digitalWrite(9, LOW); delay(200);

}

showStatus("READY");

}

void showStatus(String txt) {

display.clearDisplay();

display.setTextColor(WHITE);

display.setCursor(30, 25);

display.print(txt);

display.display();

}

Part C: Cloud Connectivity (MQTT & GPS)

"The tracking system publishes real-time coordinates using the NMEA 0183 standard. The data is parsed on the Raspberry Pi and sent as a JSON payload to the cloud dashboard."

Broker: broker.hivemq.com
Topic: iot-pickpocket/final/gps
Visualization: Integration with Leaflet.js map for real-time pin tracking.

Github: Link