Computer Vision Based Self Driving Vehicle Model With Several Autonomous Controlling Features

Project Title

Computer Vision Based Self Driving Vehicle Model With Several Autonomous Controlling Features

Project Area of Specialization Artificial IntelligenceProject Summary

For a decade the autonomous car has been the headline of the news and still continuously dominate the auto headlines. Therefor the autonomous car has attracted the researcher to the robotic companies and automobiles industries. That is why many technological companies have develop a keen interest in autonomous vehicles or self-driving cars productions. Self-driving car (driver less car, AI action based car and autonomous car) is a vehicle that rely with a heavy sensor like radars cameras and GPS to travel from a place to another place without any slight causality because self-driving cars are capable of performing object detection, detecting the motion of other vehicles, lane detection, pedestrian detection and traffic signs detection. The heavy sensor system makes the car completely autonomous. These detections can be done through image processing by a camera vision. This autonomous car is also capable to calculate the distance through camera vision.

Self-driving car perform actions through a complete process. Firstly, images are taken by the cameras which can be of either pedestrians, vehicles or traffic signs, then these images are fed to a processor which uses artificial intelligence (AI) to  process the input data. These processors are trained AI which are capable of taking different difficult decisions with micro seconds of precision. Then the AI controls the working of the controlling module of the car in accordance with the input images of the environment, i.e. either to go, stop, turn right or turn left. The better decision is taken by the AI which has the abilities by the program burned in the machine through processor. It makes the machine enable to think like a human but in a more efficient manner and take correct decisions.

Project Objectives

The primary objective is to implement the image processing techniques on a Prototype of a Self-Drive Car and controlling the steering, brakes and acceleration of the car.

The image processing includes Pedestrian Detection, Lane Detection and Path Condition Detection. We also aim to make the car completely autonomous and safe to drive and it could also be used as a driver assistance in existing cars.

Project Implementation Method

First of all the input which is the surrounding image is captured with the help of cameras mounted on the front of the Car. Camera is used for this purpose to get a wide angle image of the surroundings. The camera is connected with the onboard computer or processor, which is the AI and controlling part of the Car. AI is made using Deep Learning with the help of different image processing techniques which includes lane detection, pedestrian detection, vehicle detection and path condition detection. The software part is burned on the computer which consist of different algorithms and data set to get the desired information from the images of the camera. Lane Detection, Pedestrian Detection and Path Condition Detection algorithms are the techniques used for processing the image. The computer is equipped with the power supply for continuous power during its processing. The Computer is then connected with the controlling body of the Car which are wheels, steering, acceleration and brakes. The Cameras provide the image of the surrounding and the Computer process them using different Image Processing Techniques and then control the motion of the car according to the environment. The car will track its path with the help of lane detection and path planning algorithm.

Benefits of the Project

Here are few of the ways a driverless future might benefit us all:

Green Machines:

Autonomous vehicles are, to all intents and purposes, software on wheels. The technology involved in a driverless car of the future will be such that each vehicle can be optimized to ensure fuel consumption is as efficient as possible. So much so that new-age cars are expected to help reduce emissions by 60%.

Safer Streets:

With the potential for human error removed, self-driving cars will reduce instances of accidents caused by driver error, drunk driving or distracted drivers. Once driverless cars become commonplace on our streets, it is expected that accidents are likely to fall by a whopping 90%.

Time is Money:

Average commuter times in metropolitan areas in the US are currently estimated to be around 27 minutes each way. With humans no longer involved in driving, commuters are likely to save up to an hour every day, time that will undoubtedly have many spin-off benefits from wellbeing to boosting the economy.

Tailoring Traffic:

Every year, people living in American urban areas spend almost 7 billion hours in traffic, waste 3.1 billion gallons of fuel and lose around $160 billion due to traffic congestion. With driverless cars able to access up-to-the-minute data to help monitor traffic, as well as digital maps and other tools, they will be able to determine the fastest, most efficient route possible. All of this will result in less traffic, less congestion and less time and fuel waste.

Technical Details of Final Deliverable

As mentioned in the title of the project, we will design a prototype of a self-driving vehicle based on computer vision. Thus, our main aim is to implement deep learning-based image recognition which will then be processed to let the car take its own decisions at any given instant. In case if any obstacle or pedestrian comes n the way of the car, it could intelligently take decision to slow down the car or even stop at that particular moment, thereby, controlling the motors to control the speed, acceleration and brakes. Also, to control the steering to properly let the car follow a defined track. This would led to a safer drive which will avoid accidents that commonly occur in our country. The project would be implemented on a ride-on remote control toy car. A couple of cameras to fetch input images, an on-board computer to process the algorithms of deep learning and then control the car either in ways of its steering, acceleration or brakes.

Final Deliverable of the Project HW/SW integrated systemCore Industry TransportationOther Industries Manufacturing , Health , Security Core Technology Artificial Intelligence(AI)Other Technologies RoboticsSustainable Development Goals Good Health and Well-Being for People, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	72600
RC Car	Equipment	1	13000	13000
Camera	Equipment	1	10000	10000
Raspberry Pi Kit	Equipment	1	10000	10000
Arduino Mega	Equipment	2	1500	3000
Graphics Card	Equipment	1	30000	30000
Proximity Sensors	Equipment	2	800	1600
Resistors	Equipment	10	10	100
Capacitors	Equipment	10	20	200
Power Bank	Equipment	1	1200	1200
PCBs	Equipment	2	200	400
LEDs	Equipment	10	10	100
Vero Boards	Equipment	2	150	300
Card Board	Miscellaneous	2	500	1000
Tapes	Miscellaneous	4	50	200
Printing	Miscellaneous	1	1000	1000
Glues	Miscellaneous	2	250	500