The invention discloses an SAR (synthetic aperture radar) image change detection method based on a high-order neighborhood triplet Markov random field model, and mainly solves problems of high false detection amount and low overall precision of an existing method. The SAR image change detection method comprises the following steps: 1, inputting two time phase SAR images and generating a differential image; 2, initializing a labeling field X; 3, initializing a likelihood parameter; 4, performing definition on the initialized labeling field X by adopting a 3*3 neighborhood and initializing an auxiliary field U; 5, constructing a priori potential energy function comprising a homogeneous region, a heterogeneous region and a U field part by adopting a 5*5 high-order neighborhood; 6, updating the labeling field X and the auxiliary field U; 7, updating the likelihood parameter according to the updated labeling field X; 8, performing iterative updating on the labeling field X and the auxiliary field U and obtaining a final change detection result. Compared with the prior art, the SAR image change detection method reduces the false detection amount, improves the overall detection precision, enhances the noise robustness, and can be used for SAR image recognition.