Collision recognition from a video

Introduction

Driving is a task that requires attention distribution. One of its many issues is identifying possible collision trajetories by vehicles from behind. Thus, there is a need for a system that automatically recognizes vehicles that are about to collide with the user, and warns him/her.

In this project, we planned a Computer Vision based solution to this problem. Using a video feed from a single simple camera, we propose an algorithm that recognizes moving vehicles in the video, predicts whether they are about to collide with the user, and alerts in real-time before a collision can take place.

We present here Part A of this project. This part focuses on the algorithm itself, without taking into account real-time constraints. This part was implemented using Matlab.

Part B of this project focuses on working under real-time constraints and is implemented on Android. This part is currently work in progress.