The invention relates to a semi-supervised surgical video process identification method, and belongs to the field of artificial intelligence and medical application. The method comprises the followingsteps: step 1, carrying out spatial feature extraction on a video by using a convolutional auto-encoder; 2, extracting a pair of video frames from the same video context to carry out time sequence sorting task learning, and extracting time sequence characteristics; 3, recognizing an operation video process stage through a multi-classification recognition network; And 4, outputting an operation video process stage identification result. The method serves as an automatic identification and detection tool for the open surgical operation video process, and is a low-cost method for efficiently judging the operation video process stage of an artificial intelligence medical operating room in the future.