Video object segmentation method based on semantic label guidance

The invention provides a video object segmentation method based on semantic label guidance. The method comprises steps: according to a semantic class label to which an object belongs, an object bounding box detector and an object contour detector are sequentially used for detecting each frame of an inputted video, and a candidate object bounding box set and a candidate object contour set for each frame of the inputted video are obtained; a joint allocation model containing the candidate object bounding box set and the candidate object contour set is built, an initial segmentation sequence corresponding to the object in the inputted video is solved, the initial segmentation sequence is processed, and the shape probability distribution for the object is estimated; and in combination with the shape probability distribution, graph cut is used sequentially for carrying out optimal processing on each sequence containing the object to obtain the best segmentation sequence corresponding to the object. The technical scheme of the invention solves the problems that the existing video object segmentation method is inaccurate and can not applied to semantic class object segmentation for a single inputted video.