Integrating Image Segmentation and Deep Learning to Improve Radio Frequency Propagation Models

ABSTRACT This paper proposes a multi‐sensor approach to improve radio frequency (RF) propagation models, which play a key role in the rapidly expanding field of connected vehicle technology. Focusing on the 1‐ to 20‐GHz frequency range, which is critical for both satellite‐to‐vehicle and base station‐to‐vehicle communications, our study introduces a detailed image segmentation algorithm complemented by comprehensive analyses of diverse environmental data. A key aspect of our method is the use of multiple image databases in conjunction with advanced deep learning techniques for accurate Line‐of‐Sight (LOS) classification. This integration of image‐based information aims to enhance the predictive accuracy of RF propagation models, thereby supporting improvements in connected vehicle communications in the domains of safety, entertainment, and navigation. Through an in‐depth examination of the relationships between environmental variables and RF signal propagation, our research provides valuable insights into optimizing connectivity in dynamic vehicular settings. The advancements discussed in this paper contribute to the ongoing development of connected vehicle technology and lay the groundwork for future enhancements, ultimately aiming to achieve consistent vehicle connectivity across varied environmental conditions.