Visibility Enhancement in Poor Weather-Tracking of Vehicles

摘要

Poor weather conditions are considered a major cause of vehicle accidents. According to the US Department of Transportation, there are over 22% of crashes each year are caused by poor weather conditions, something that can be reduce if we can enhance the visibility conditions. In this paper, we present an approach that can improve driver's visibility during poor weather conditions based on the Retinex algorithm and augment this by object detection and a distance quantification within the view of the driver. The Retinex algorithm uses the principle of the human visual system as governed by the retina and the cortex. It relies on light levels and color perception of the human eye and how both the retina and cortex ensure the capturing of a suitable dynamic range of colors. Gaussian Mixture Model and blob analysis are used to detect and track objects in front of the driver. Then, we used depth estimation to estimate objects distances within the view of the driver which has a proven accuracy of over 96% to calculate distances. We will present simulation results of different low visibility conditions and show a warning system of distance from objects that is accurate.