Intelligent Traffic Analysis

Abstract

An accurate and efficient traffic flow density estimation gives important insights for construction of new road infrastructure and for maintenance of the existing one. In Pakistan there is no automated system available to perform this crucial estimation. The manual approach to handling this problem is full of flaws due to the non-feasibility of performing this task on hundreds of hours of video streams. There is a need for an intelligent system that can automate this process and provide precise estimates about the number and types of vehicles that uses a particular route for commute. Vehicle detection and recognition from aerial imagery have become possible due to latest advances in computer vision and the increasing use of Unmanned Aerial Vehicles (UAVs) for road traffic monitoring. We have proposed an intelligent system that will automate the detection and recognition of vehicles from aerial imagery for this purpose. This system will use state of the art deep learning frameworks for the detection and recognition from aerial imagery. Our system will detect and recognize vehicles from aerial video captured from drone and will be robust to the variations in altitudes, scale, orientations and angles of the vehicles. In Chapter 1, we will establish the need and motivation for this project followed by the tools and technologies that we have used in the project. Chapter 2 will include a birds eye view of the architecture and modules for our proposed system. In Chapter 3 and 4 we will elaborate the detection and recognition modules and the frameworks used in them. Chapter 5 will conclude by elaborating on the possible future direction of this project.