This paper systematically surveys the impact of integrating large-scale language models (LLMs) with computer vision on perceptual tasks such as image segmentation. Focusing specifically on Intelligent Transportation Systems (ITS), we present the applications, challenges, and future directions of LLM-based image segmentation in ITS, where accurate scene understanding is crucial for safety and efficiency. We categorize various LLM-based image segmentation approaches based on their prompting mechanisms and core architectures, and highlight innovations that enhance road scene understanding for autonomous driving, traffic surveillance, and infrastructure maintenance. Finally, we identify key challenges such as real-time performance and safety-critical reliability, and present a perspective on explainable, human-centered AI as essential for the successful deployment of this technology in next-generation transportation systems.
