Debris chain formation has been one of the major obstacles in the feature extraction of on-line visual ferrography. In this paper, a debris chain segmentation method based on YOLOV7-tiny is proposed. YOLOV7-tiny is used to track the growth process of the debris chain, the model output information is converted into the coordinate information of the detected wear debris in the image, which is used as the foreground markers, and the image to be segmented is thresholded binarized as the background markers, which is combined with the foreground markers, and the segmentation of the input ferrogram is performed by the mark-watershed algorithm. The method proposed in this paper is characterized by high efficiency and high accuracy, and provides a feasible segmentation method for real-time segmentation of online oil monitoring.