The invention discloses a computer image processing and mode identification-based method for detecting tail yarns of a ring bobbin. The method comprises the steps of collecting a bobbin image sample; preprocessing a bobbin image to generate a binary image; extracting binary image features; creating a binary image feature library; introducing the bobbin image needed to be processed; preprocessing the bobbin image needed to be processed to generate the binary image; extracting the binary image features needed to be processed; and judging whether the binary image features needed to be processed conform to features of the feature library or not. Firstly, multiple spinning machines in a factory building can be managed by depending on a very few non-professional employees in a modern textile factory, so that the number of employees in an enterprise is reduced and the process efficiency is improved. Secondly, an adopted image collection tool is a general camera, so that hardware equipment does not need to be additionally added to a system and the cost is low. Thirdly, whether the tail yarns exist in the image or not can be automatically identified and fed back without manual intervention, so that the use is simple and efficient.