The invention puts forward a moving object classification method based on a support vector machine, and mainly solves the problem of complex background, fuzzy moving object and poor classification effect in an existing moving object classification method. The method comprises the following implementation steps of: (1) utilizing a camera to read a video; (2) extracting video frames; (3) reading one frame of video frame image; (4) carrying out preprocessing; (5) extracting the edge, the gradient and the information entropy feature of a processed moving object image; (6) judging whether all video frame images are read or not; (7) collecting moving object features; (8) training the support vector machine; and (9) obtaining a classification result. The method has the advantages that accurate moving objects and moving object features can be extracted in a complex environment and accurate classification is achieved by the support vector machine. The method can be used for moving object classification in a video, and reliable data is provided for subsequent object tracking and identification.