Abstract: An air-leak diagnosis and localization method based on infrared thermography is described in this paper, which is developed in an attempt to overcome the disadvantages of low efficiency and poor anti-jamming ability associated with the traditional methods for localization of leaks from a vessel. First the thermal images of the tested objects were captured before and after air charging. The image processing of enhancement and weighted difference were done. Then a temperature difference factor <em<θ</em< was introduced as a parameter changed with temperature range of the target region to improve the robustness of image segmentation. Furthermore, the position and boundary of the leak could be extracted from processed images through a factor <em<θ</em< based kernelized fuzzy clustering segmentation. Experiment results indicate that the method proposed is effective and sensitive. The purpose of air-leak detection and localization has been achieved. The leak detection ability could be 10 mL/min(@0.1 MPa).