An Enhanced Integrated Deep Learning Method to Overcome Dehazing Issues on Intelligent Vehicles

Abstract

Intelligent vehicles rely on accurate perception systems for safe navigation under challenging environmental conditions. However, adverse weather phenomena such as haze and fog significantly degrade the quality of visual input captured by onboard cameras. In this study, we propose a novel deep-learning algorithm specifically utilizes an advanced convolutional neural network (CNN) architecture that features parallel networks, encoder-decoder configurations, He Normal initialization, batch normalization, and the Gaussian Error Linear Unit (GELU) activation function. This integration not only improves the dehazing capability, but also the overall image quality, which increases the accuracy of intelligent driving systems. The experimental results demonstrate that our proposed learning-based dehazing algorithm consistently outperforms existing methods in terms of structural similarity (SSIM) and peak signal-to-noise ratio (PSNR). These enhancements demonstrate the potential of our proposed solution to significantly increase the state-of-the-art in image dehazing.