The research based on the computer vision of intelligent mobile robot serving NC machine feeding system, puts forward a novel method which names the semantic intermittent boundary theory about Partly Sheltered cylindrical work piece. For the cylindrical work piece being irregularly placed, a part of its appearance would be sheltered by other objects. It is very difficult to recognize these partly sheltered cylindrical work piece. The intermittent boundary segments theory will successfully solve this problem: firstly, the image preprocessing about the partly sheltered cylindrical work piece is going on, after this, the intermittent edges are achieved by the Canny edge extraction algorithm; secondly, The intermittent boundary are set up by connecting the edges got on last step. The intermittent boundary include many local shape characteristics embodied by the integrated object boundary semantic; thirdly, structuring the top semantic guidance for pattern recognition; Finally, based on the local feature of scale P in variance, The intermittent boundary a' recognition is completed by SVM referring to the top semantics guidance. The experiment shows that the accurate identification rate is 80%, and the speed of recognition with this semantic intermittent boundary theory increase greatly. The theory satisfies NC machine feeding system in real-time. © 2013 American Scientific Publishers All rights reserved.