Automated Detection Of Driving Pathway Using Image Processing

Abstract

Image data is one of the most popular real world input data that can be used for variety of applications ranging from robotics and computer vision to security systems. In combination with other methods such as neural network, Artificial neural network and image processing techniques, manipulation of image data can lead to applications such as detection of objects, tracking, identification and vision based robotics and so on. Advanced Driver Assistance System (ADAS) also use image for camera based driver assistance systems.The report covers a hardware model system that tests the software work of detection of traffic signs and path for it own ADAS systems. Different problems were tackled, including the choice of OS, and additional hardware components needed to tackle. The choice of programming languages, equipment, OS and methods were based on simplicity and practicality. Artificial neural network in combination with Open CV libraries were used for stop sign, traffic light and path road detection. The hardware model consisted of RC Car attached to raspberry pi board with a mounted pi camera for video streaming and an arduino controller attached to a radio transmitter for controlling through Open CV running in windows PC.

Key Findings

• Built a functional prototype autonomous vehicle using Raspberry Pi, Arduino, and a custom RC car chassis.
• Implemented real-time computer vision pipelines using OpenCV for path following, stop sign detection, and traffic light recognition.
• Integrated Artificial Neural Networks (ANN) to improve detection accuracy in varying environmental conditions.
• Validated a low-cost, practical hardware-software architecture for educational and experimental ADAS systems.