Smart Aerial Drone based Surveillance System

Project Title

Smart Aerial Drone based Surveillance System

Project Area of Specialization Artificial IntelligenceProject Summary

There has been a major increase in crimes and terrorist activities in the country. Additionally, riots and protests culture is also rampant in our society with not enough tools for crowd monitoring and control. From the current pool of solutions available today, the most widely used is the deployment of vehicles and security guards for patrolling. In some places static CCTVs cameras are also deployed for surveillance. Although these solutions are effective to some extent, they also have many down sides to their application as well, such as, security personnel are prone to fatigue, distractions and human errors and are not suited to cumbersome repetitive tasks such as surveillance, be it physical or through CCTV cameras. Cost is another important factor to consider here which is directly related to the number of guards on the payroll or the number of vehicles in use. Keeping in mind these issues and to further improve and automate the current surveillance and security apparatus, we are proposing an innovative, autonomous, UAV based approach which will not only help make the system more robust but will also bring down the overall cost as well.

Project Objectives

To design a fully autonomous, affordable, sustainable, and automated mobile aerial monitoring solution which we believe can eliminate the human and technical limitations inherent to current systems deployed. Our drone will guide itself through a GPS geo-fenced, user designated area, running pre-trained AI models for the objects of interest against the live feed from its camera. If, for example it detects guns or recognizes any face which has been marked as suspicious, the drone will raise an alarm in the control room, alerting the authorities and having the situation investigated promptly, while also recording the incident to be later used as evidence for forensic or legal uses. The drone will also be able to detect and count people in a crowd. All this will be done using on-board processing with only control and feedback connection to the base station making it highly useful for remote areas with bad network coverage or infrastructure.

Project Implementation Method

The project will employ image processing and smart pathfinding. Furthermore, navigation would be comprised of three things: GPS, obstacle avoidance, and dynamic pathfinding. Of course, manual override will be a feature just in case the drone starts straying off-course. For pathfinding, and obstacle avoidance, we will be employing a Convolution Neural Network (CNN), a reinforced learning algorithm. The drone would periodically shift locations and charge at the charging station whenever it deems its batteries to be running low. Facial and object recognition would allow the drone to understand who is doing what.

The running of AI algorithms will be done on-board on Nvidia Jetson module with its own dedicated battery. There will be no reliance on any cloud computing resource making the solution robust and feasible for virtually any place. A second battery pack will power the flight control, Pixhawk, and the drone components such as motors.

To start the surveillance, GPS coordinated would be fed to the flight controllers, and a geo-fencing of an area would be done where the drone will fly around. It will communicate with the base station, to ping an alert and send pictures, using RF signals.

Benefits of the Project

The main benefit of our project would be to ensure safer public and private spaces. At the same time, we would be providing enhanced security at a reduced overall cost, all while having minimal carbon footprint. The most obvious problem in the current system is that the guards as humans are more prone to distractions and fatigue. Keeping a vigilant watch on all things, which includes every person leaving and entering the premises or all the different streams of CCTV footage, throughout their shift can be a challenging task. Increasing the number of guards can help but it is not feasible economically having too many guards on the payroll.

Secondly, vehicular patrolling creates a huge overhead of procuring cars, their maintenance and their daily fuel consumption. Not to mention the noise and environmental pollution they can cause along with being extra traffic on residential roads which usually are quite narrow. For all this expense the benefit they give is pretty limited as their area of surveillance is small, usually on the street they are currently on whereas an aerial drone can cover a much larger area from a single point at only a fraction of the cost. Moreover, the static CCTV cameras planted at different locations can at times be easy to bypass as they have a limited angle of coverage. Also, the camera feed is usually monitored by a single guard on a small to medium sized screen which makes it hard to keep track of the minute things happening on all the feeds simultaneously and at the same time has the added risk of the guard falling asleep and rendering the whole system useless.

Additionally, aerial drones for surveillance purposes are hardly used anywhere because firstly, the commercial drones available in the market right now which are capable of such surveillance are too costly(usually to the tune of $1500 - $3000), and secondly they have to be manually remote controlled which doesn’t bring much value to the overall solution as guards would have to be dedicated for the purpose of flying the drone, day and night, which at best, may only reduce the effort by a small bit, and would not justify the expense on the drone. To further add to the expense, the guards would need to undergo special trainings and proper network infrastructure would need to be deployed to be able to deliver live feed to the base station.  And since guards would’ve to stay up all night to operate it, it carries with it the risk of guards falling asleep on the job or missing something because of the lack of focus. Overall, this solution is not refined enough and does not cut the cost enough or give substantial advantages for it to be considered as a viable solution. A solution targeted for this specific use case is needed. The proposed solution caters to all these shortcomings in a highly cost-effective and eco-friendly manner.

Technical Details of Final Deliverable

Our Final Deliverable will be a smart autonomous drone which will be employing computer vision along with smart path finding and obstacle avoidance for aerial manoeuvring. Furthermore, it will be using edge AI to analyze the environment for objects and people to identify criminal activity such as any trespassing, or illegal gatherings, and also for crowd counting and control.

Final Deliverable of the Project HW/SW integrated systemCore Industry SecurityOther Industries IT Core Technology Artificial Intelligence(AI)Other Technologies RoboticsSustainable Development Goals Good Health and Well-Being for People, Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Responsible Consumption and Production, Climate Action, Peace and Justice Strong Institutions, Partnerships to achieve the GoalRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
NVIDIA Jetson Nano Developer 2GB	Equipment	1	14000	14000
Complete Pixhawk Flight Controller Set (incl. GPS module, Cables, etc)	Equipment	1	32000	32000
Batteries	Equipment	2	3250	6500
DIY Drone Kit (incl. frame, motors, ESCs, propellors, etc)	Equipment	1	11000	11000
Prinitng and Documentations	Miscellaneous	1	7000	7000
Logistics	Miscellaneous	1	3000	3000
Raspberry Pi NoIR Infrared Camera Board v2.1 (8MP, 1080p)	Equipment	1	5000	5000
Drone Landing Skid/Stand	Equipment	1	1500	1500