1675191 - Enhancing Autonomous Vehicle Safety in Rain: A Deep Learning Approach for Clear Vision

Rain is a ubiquitous meteorological phenomenon that poses a significant challenge for autonomous vehicles navigating roadways.  Due to the adverse impact of rain droplets on camera-based object and feature detection, automotive manufacturers often deactivate autonomous driving features during inclement weather. This research endeavor aims to mitigate this issue by harnessing contemporary deep learning techniques. The primary objective is to develop a deep learning model capable of real-time processing live images captured by vehicle’s cameras in rainy conditions, effectively eliminating the visual hindrance caused by rain droplets. To facilitate this endeavor, a simulation environment is created in CARLA, an open-source simulator designed for autonomous driving research. Within this virtual space, a comprehensive range of clear and rainy images is captured, encompassing diverse environmental and landscape scenarios. These images will serve as the training and testing datasets for developing the deep learning model. The goal of this research is to demonstrate the model’s proficiency in successfully removing rain from images, resulting in visuals closely resembling clear, rain-free images. This work seeks to enhance the performance and safety of autonomous vehicles, with a particular focus on ensuring their reliability during adverse weather conditions.