Description

One-Click Dispatch and Automatic Area Scan for Suspected Terror Incidents

General Description
When a suspected hostile terror incident is reported, for example a possible infiltration or suspects who fled into an open field, orchard or wadi, the police must search the area quickly, often at night. The same need arises when searching for a missing person. Today the standard aerial asset for this is a police helicopter with a thermal camera. In practice it can take around 40 minutes for the helicopter to arrive, which is critical time in a search. A small drone with a thermal camera can be in the air within minutes, but today launching it, planning the flight and watching the video all depend on a skilled operator working by hand. The police drone unit asked for a system that closes this gap for the first minutes of an incident: a command dashboard where an operator selects a predefined area cell (a polygon), presses one button, and a thermal drone is sent to the location and automatically scans the area in a systematic back-and-forth pattern (the -iron- / lawnmower pattern). Onboard or ground-side AI detects people and vehicles in the thermal video and marks them on the map in real time. The project combines path planning, drone control, computer vision on thermal images and a real-time web dashboard, with a real operational client.
Project Scope
The goal of the project is to build a working prototype of the dispatch-and-scan system, developed and validated mainly in simulation, with a clear path to field testing by the police drone unit.
Main objectives:
- Command dashboard with a map showing predefined area cells (polygons), drone status, battery, live video and detections
- One-click dispatch: the operator selects a polygon and the system generates and uploads a full mission (takeoff, transit to the coordinate, scan, return to home)
- Coverage path planning inside an arbitrary polygon: lawnmower pattern with line spacing derived from the thermal camera's field of view, flight altitude and required overlap, including handling of concave polygons and no-fly zones
- Thermal detection of people and vehicles with a deep learning model trained on public thermal drone datasets (for example HIT-UAV), with geolocation of each detection from the drone's position, altitude and camera angle
- Real-time alerts on the map (snapshot, coordinates, confidence) so the operator can direct ground forces or hold the drone over a target
- Full simulation environment (PX4 or ArduPilot SITL with Gazebo) for development and demonstration, and a mission log for after-action review
- Optional: estimate of coverage time per polygon versus helicopter arrival time, and support for splitting a large area between two drones
Student Requirements
Expectations from students:
- Teamwork and fair division of work, tracked in a shared task board
- Full attendance in weekly meetings with the supervisor
- High motivation and independent learning of new material
- Personal responsibility for agreed milestones and deadlines
- Good Python and web development skills- interest in robotics, drones and computer vision
- Willingness to learn drone autopilot concepts (MAVLink, missions, flight modes)
- All development and testing in simulation- any real flight is performed only by the unit's licensed operators and in line with Civil Aviation Authority regulations
- Discretion with any operational information shared by the police unit
Development Tools
The project will use the following tools and technologies:
- Python- PX4 or ArduPilot autopilot with SITL and Gazebo simulation
- MAVSDK / pymavlink for mission upload and telemetry over MAVLink
- Shapely / GeoPandas for polygon handling and coverage path planning
- PyTorch and Ultralytics YOLO for detection on thermal images- public thermal datasets such as HIT-UAV
- OpenCV and GStreamer / WebRTC for video streaming
- FastAPI + WebSocket backend- React with Leaflet or Mapbox for the map dashboard
- Optional edge deployment on NVIDIA Jetson
- Git / GitHub
Deliverables
Required deliverables:
- Specification document (requirements, architecture, data, evaluation plan)
- YouTube video demonstrating the working system
- Poster
- Final presentation
- GitHub link to the code, with README and reproducible experiments

Additional Notes
• Each project team may include up to 5 students.
Contact Details
Supervisor: Idan Tobis
Email: idan.tobis@gmail.com
One-Click Dispatch and Automatic Area Scan for Suspected Terror Incidents

Drones and 3D Sensors

Description

Emphasis in project execution

The project is has cooperation with the industry and combines meeting deadlines while being creative and focused on the task
