Obstacle detection and distance estimation for autonomous electric vehicle using stereo vision and DNN

Abstract

Automation—replacement of humans with technology—is everywhere. It is going to become far more widespread, as industries are continuing to adapt to new technologies and are trying to find novel ways to save time, money, and effort. Automation in automobiles aims at replacing human intervention during the run time of vehicle by perceiving the environment around automobile in real time. This can be achieved in multitude of ways such as using passive sensors like camera and applying vision algorithms on their data or using active sensors like RADAR, LIDAR, time of flight (TOF). Active sensors are costly and not suitable for use in academic and research purposes. Since we have advanced computational platforms and optimized vision algorithms, we can make use of low-cost vision sensors to capture images in real time and map the surroundings of an automobile. In this paper, we tried to implement stereo vision on autonomous electric vehicle for obstacle detection and distance estimation.