Deep learning approach to detection of preceding vehicle in advanced driver assistance

Abstract

In paper we propose a detection method for objects in video stream taken in front of a car by means of deep learning. The successful detection of preceding cars is a part of the analysis of current road situation including emergency and sudden braking, unintentional lane change, traffic jam, accident, etc. We include the results of preliminary experiments employing video stream captured by camera installed behind frontal wind screen. The detection and classification are performed using Convolutional Neural Network preceded by road lane detection. We performed several experiments on real-world data in order to check the accuracy of the proposed algorithm.