Project Title
Edge AI-Powered Surveillance: Intelligent Threat Detection and Crisis Response

Why this Project?
Our project draws inspiration from regular terrorist/criminal activities observed all across the world. A large part of defence systems are based utilizing sensed data from cameras, radar modules, and other sensing devices. The data retrieved from these sensors is then processed using modern day computer vision and machine learning techniques. However, these computation-intensive tasks are typically performed on high-end computers or processors, which can result in delayed responses during critical situations.
We believe that AI and Edge-computing operations on microcontrollers itself present as a viable solution to this widely encountered issue worldwide.

Objectives
1) Camera based detection of hand-held weapons such as ammunitions, knifes etc.
2) Develop a suitable efficient detection deep learning algorithm.
3) Deployment of algorithm on microcontroller evaluation board.
4) Development of a wireless network gateway such as NODE-RED or either transmission of data using LoRaWAN module.
5) If time permits we may also want to utilise the MEMS microphone present on the board for audio based crisis detection.

Components Required-
1) 1 x Maixduino Kit (including camera module and TFT Display)
2) 2 x LoRa module RFM95W-915S2R
3) Battery (For future independent power supply)
4) 1 x TFT Card

Data Collection
1) Cameras: Continuously monitor live video footage.
2) Analysis: TensorFlow Lite model analyses the footage for weapon-based threats.
Data Processing
1) Evaluation Board: Processes the camera feed and uses the deep learning model to analyse the situation.
2) Trigger Generation: Generates a trigger to the LoRa module to transmit if any threat is detected.
Data Transmission
1) LoRa Module: Sends the node ID details where the threat is detected to the central gateway device or Node-Red dashboard.
2) Database: Node ID details of different nodes are pre-registered in the database.
Alerts and Data Storage
1) Alerts: Include exact date and time of the illicit activity, along with node ID and smart recommendations for future action.
2) Image Capture: An image of the criminal activity can be captured and stored on an on-board Micro-SD card for future reference.
Microphone Integration (Additional Feature)
1) MEMS Microphone: Identifies certain phrases based on an ML model during crises to provide emergency help.

Conclusion:

On a microcontroller board, this AI-dependent security system is an advanced response to safety and defense in real time.This gives a threefold protection by combining such technologies as TensorFlow Lite for video analysis, LoRa modules for adequate data transmission, and MEMS microphones for audio detection.
It also aids in incident handling and decision making. This is why it is an important device that provides protection in cases of possible attack by giving peace of mind through modern technology.