In this study, we propose a methodology for leak detection and monitoring of steam pipes installed in ship engine rooms. Generally, steam pipes in engine rooms are wrapped with insulation materials, making it difficult to visually identify steam leaks and potentially delaying initial response. Therefore, we propose hardware and software design methods for a sensor system that provides complementary information using RGB and thermal cameras. Specifically, the proposed system consists of a camera server module, a camera calibration module, an image alignment module, a thermal-map training module, and an inference and visualization module. Considering that the steam pipe leak results in high temperatures, we define the concept of a thermal-map and propose algorithms for effective learning of the thermal map, detection of abnormally high temperatures based on the thermal map, and visualization of the detected abnormal regions. The effectiveness of the proposed method was evaluated through various experiments using an experimental setup that simulates the steam piping system of a ship.