Since the existence of buried targets influences the surface＇ IR images changing with time, buried targets were detected by multitemporal infrared images. In order to denoise the exposal nioses of partial IR images, a method of kernel minor component analysis (KMCA) was put forward to denoise muhitemporal images and overcome the difficulty in choosing images automatically, thus, the good images could be attained automatically. Then a fuzzy kernel cluster algorithm with spatio and temporal restrictions (STKFCM) was proposed to detect buried targets based on multitemporal IR images, in which the index of temporal information was introduced to modify the temporal weight factor. At last, the position of targets and the number of classes were estimated by the results of classification. After taking the mechanism of buried targets IR images into account, the general characters of targets were obtained. This is an interesting study to detect buried targets by using the technology of IR.