The invention relates to the technical field of video scene classification, in particular to a video scene classification method based on multiple features, which comprises the following steps: S1, extracting a key frame set of scene videos; S2, calculating an average key frame of each scene video; S3, dividing each average key frame; S4, extracting scene multi-features, wherein the scene multi-features comprise color features and texture features; S5, performing fusion by using multi-feature fusion to obtain comprehensive features of the scene video; and S6, comparing the comprehensive features with corresponding feature thresholds to perform scene classification, and obtaining a plurality of scene categories by using a threshold judgment classification method. According to the thresholdjudgment classification method, the structural features of the video are fully utilized; the method has high accuracy and strong classification result stability when distinguishing different scenes, is a classification algorithm with high robustness, and has the characteristic of dynamic classification, the classification category number is determined according to a threshold value, category number division in advance and training set training are not needed, and the classification process is easy to realize.