Abstract: Leakage represents one of the most critical failure modes in earth-rock dam engineering, and timely detection of leakage hazards is essential for ensuring dam safety. Based on the correlation between leakage and temperature distribution in earth-rock dams, this study systematically elucidates the principle of applying infrared thermal imaging technology for non-contact leakage detection. A boustrophedon scanning path model based on dual-loop PID control is proposed. Combined with an adaptive region-growing algorithm and IHS (Intensity-Hue-Saturation) image fusion technique, an intelligent leakage identification method is developed using thermal infrared remote sensing images, with the Xiangshan Reservoir dam project in Heihe City as a case study. The results demonstrate that thermal infrared imagery can effectively visualize temperature anomalies caused by leakage, characterized by a distinct temperature gradient that decreases inward and increases outward. The adaptive region-growing method achieves automatic and accurate identification of leakage areas, significantly improving detection efficiency and adaptability. Furthermore, the IHS-based multi-source image fusion generates high-contrast, multi-dimensional fused images integrating infrared and visible spectra, providing a technically viable solution for dam leakage inspection and identification.