The invention discloses a multi-feature fusion based infrared forest fire judgment method. The method comprises the following steps: A, inputting an infrared image sequence; B, performing binarization processing; C, performing morphological processing; D, extracting flame features; E, performing SVM classification judgment on the flame features; F, performing classification decision-making by utilizing a Bayes classifier; and G, judging whether a fire occurs or not through data in the step F. The flame features in the step D include an infrared flame height characteristic, an infrared flame gradient characteristic, an infrared flame moving characteristic and a flame spectrum characteristic. According to the multi-feature fusion based infrared forest fire judgment method provided by the invention, the fire can be detected quickly, accurately and safely in the early stage of the forest fire, so that the judgment precision of forest fire prevention is greatly improved and the working efficiency is improved.