Vehicle Detection using UAV

Project Title

Vehicle Detection using UAV

Project Area of Specialization Artificial IntelligenceProject Summary

Project Summary:

In recent years, Unmanned Aerial Vehicles (UAV) have gained popularity in traffic monitoring system, military surveillance and object detection etc. UAVs are now chosen widely because of low cost, high resolution, good flexibility and wide coverage over the whole area.

UAVs with cameras are considered low-price platform which can offer efficient facts and figures for intelligent transport System.

As the use of vehicles has increased in today's world, the needs of traffic management has also been increased. Conventional method of traffic data collection are fixed only to the local region and thus it is far expensive and labor intensive to monitor and display traffic activities across huge areas. However UAV has its advantages in traffic management and monitoring over a large distance. UAV is capable of carrying various type of sensors to collect better surrounding knowledge and can utilize it for traffic management.

The purpose of UAV for this project is to detect a vehicle. The vehicle detection goes through five stages. In the first stage, the dataset of images of vehicles will be provided to the UAV. In the second and third stage, the UAV will detect and track the vehicle of interest using computer vision algorithms. In the last two stages, the vehicle tracker will produce vehicle localization coordinates per image, and finally a transform will convert those image coordinates into a new Global Positioning Coordinates (GPS).

Project Objectives

Objectives:

• To research and study about such equipment (camera) which can detect an object (vehicle) from specific height with great resolution that can be used for training.
• To tradeoff between camera module and high resolution cameras.
• To assemble such a device (UAV) which can hold camera at specific height and act accordingly.
• To gather and manage datasets for different vehicles at different angles.
• To train our model, classification of objects which means the model has to identify if the image or video of the dataset is a vehicle, people, tree or any other thing.
• To select the best object detection technique which is fast, efficient and accurate for the system in real-time.
• To develop such a system which can generate GPS coordinate from image captured by UAV through image processing by measuring the distance between UAV’s current position and object position in real-time.

Project Implementation Method

Methodology:
Manual control will be initiated from ground to deal with any emergency situation. The project working is divided into the following stages:

1. Vehicle detection:
   In this step, the vehicle detection module obtains aerial images from the data acquisition model. The pixel position is sent as initial pixel position to the visual tracking module in the next stage.
    
2. Vehicle Tracking and Localization:
   The data acquisition model provides image and flight motion parameters to the visual tracking and geolocation module. The tracking module then calculates the future direction of the vehicle by predicting its speed and direction in the image obtained in step 1 of vehicle detection. The flight motion provides UAV’s current location.
    
3. Image Coordinate Transformation into GPS coordinates:
   The vehicle tracking module along with geolocation module produces vehicle localization coordinates per each image. Then, a transform is used to convert those localized coordinates into an estimated GPS coordinates of the tracked vehicle.

The below diagram shows the block diagram of the whole process going on:

Figure 1: Block diagram of vehicle detection, tracking, and coordinate transform
Also the flow chart given below would help in understanding the work flow in a better way.

Figure 2: Flow-chart of vehicle tracking

Benefits of the Project

Benefits:
There are a number of benefits attached to the project which are listed below

1. First of all it would help the agencies in successful spying of the suspects which could help them protecting the country from some severe dangers. Also the UAV tracking algorithm would be made to be fully automatic so only a single man can view the real time tracking of 3 to 4 suspects which are followed by different UAVs otherwise at least a single man with a car or bike would be required for the surveillance of a single suspect which involves mental as well as physical workout of the man to track and follow the target vehicle also there are chances that the suspect know that they are being followed. Whereas using this project only a single man can easily view the real time tracking of the target without any physical workout.
2. Film making industry can also use this project for motion picture filmmaking.
3. Because of its high altitude positioning, it can be used for spy footages as well.
4. It can be used by armed forces or military for enemy reconnaissance.

Technical Details of Final Deliverable

Deliverables:
There are 4 crucial milestones that must be accomplished for the completion of our project. All of them are of utmost importance and each requires a good deal of time to achieve.
These are listed below:

1. Assembling Hardware:
   Hardware Assembling involves procurement of equipment like camera, UAV, controller etc. and then assembling them together so that we can look that if the UAV is capable for flying for the certain time frame
   and if it can bear the weight of camera and controller easily without effecting the time of flight that much. So we have to do some research in selecting the camera and controller such that they provide a good value 
   without effecting the quality of video for camera and range for the controller.
2. Procuring and Managing Dataset:
   Secondly, we have to work on gathering the dataset from which our neural network would get trained so it will definitely consume time since we have to first collect the images for the dataset and then the annotation of dataset and after that it would be ready for the neural network.
3. Training Neural Network:
   After our dataset is ready we would be testing different algorithms and would use the best in our case to be implemented on our project so this area would be consuming more time than the other 2 areas mentioned above.
    
4. Testing Hardware:
   Finally we would be testing our hardware and would try our best to find its shortcomings and would try to fix if we found some since it is a totally new design it would certainly have some shortcomings which would get highlighted in this phase and would be fixed. 

Final Deliverable of the Project HW/SW integrated systemCore Industry SecurityOther Industries Transportation , Media Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Industry, Innovation and Infrastructure, Peace and Justice Strong InstitutionsRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60179
GoPro Camera HERO 7 Silver	Equipment	1	50679	50679
Pixhawk Flight controller	Equipment	1	9500	9500