Worldwide, the area of seasonally frozen soil is extensive, accounting for approximately 56 % of the Earth's land surface. Road infrastructure development in frozen soil regions is poised to experience rapid growth in the coming decades. Nonetheless, within these diverse and challenging climatic environments, subgrade frost heave remains a substantial hurdle for infrastructure construction and maintenance in frozen soil regions. Subgrade coarse-grained fillers are typically unsaturated, featuring larger and interconnected pores. While this configuration is not conducive to the capillary rise of liquid water during freezing, it creates favorable conditions for the migration of gaseous water. In recent years, the mechanism of moisture accumulation triggered by gaseous water migration has garnered significant attention and emerged as a prominent topic in frozen soil science. Gaseous water migration is a crucial factor in the development of frost heaving in coarse-grained fillers. The aim is to study the gaseous water migration and frost heaving characteristics of coarse-grained fillers. Based on the concept of On-Site Visualization (OSV) and the law of light reflection and refraction, Polymer optical fiber (POF) sensors are proposed to monitor the light intensity in the fillers during the freezing process. Therefore, the intrinsic correlation between the light intensity and the phase transition and migration of water can be analyzed. This paper introduces the unidirectional freeze tests of coarse-grained fillers by employing a self-developed gaseous water migration apparatus and POF sensors. The experimental results show that POF sensors effectively monitored distinct stages of light intensity changes. This enables the prompt acquisition of dynamic data of the freezing depth within the seasonally frozen region and facilitates understanding of the water migration and water-ice phase transition progress. It was of great significance for predicting soil stability and understanding the water movement in the soil. POF sensors can capture the formation of freezing front and the distribution of ice. Through mathematical analysis of the relationship between changes in light intensity and water inflow, a significant correlation between the two is apparent and has an excellent fitting effect. POF sensor can be used to monitor the total amount of gaseous water migration in frozen soil in real-time.