Design and analysis of computer vision techniques for object detection and recognition in ADAS

Abstract

The rising number of road accidents has been an increasing concern in recent years in the world. To ensure overall road safety, there is a need for a reliable system that could assist drivers to avoid road accidents. This study aims to find, analyze and develop reliable techniques and algorithms that can ensure the safety of the drivers by minimizing the factor of human error in most road accidents. By focusing on the above concerns, we proposed a system in which an electronic device is used in the vehicles to assist drivers by providing accurate and reliable data about the road environment with the help of different sensors, ensuring overall road safety. The system uses a camera module for visual feed to the system and has an ultrasonic sensor to sense different obstacles on the rear side of the model car. Two convolutional neural networks (CNNs) have been compared and the best one has been selected for object detection and recognition in this system. The system has the feature of alerting the driver if the vehicle departs from its circumscribed lane. The system is effective in applications like blind-spot monitoring, forward collision warning, and so on.