Detection of Traffic Blackspots Using Deep Learning for Autonomous Vehicles with Street View Imagery

Abstract

Accidents have been a major cause of fatalities and injuries, much of which is attributed to human errors such as over speeding and drunk-driving. The onset of Autonomous Vehicles delivers a promising future with accident rate reducing significantly. But we also cannot deny the fact that these systems struggle in certain cases to avoid accidents. Hence, we propose a method to alert these autonomous vehicles in advance that they are approaching a Traffic Blackspot. A Traffic Blackspot is a dangerous prone-to-accident spot or location. This approach leverages the power of Deep Learning to understand the environmental factors of each location through street view images. This automated system will have access to the GPS location of the autonomous vehicle which it matches with its database of blackspots to determine if the vehicle is in or around such a location and alert the system when it is in the vicinity of a Blackspot. Due to this alert, a timely control can be adopted by the system with respect to speed, position on the road, etc., thus reducing the possibility of an autonomous vehicle potentially meeting with an accident.