Online, Real-time and Robust Detection and Localization of Foreign Objects on Paper Surface using Machine Vision and Clustering

To serve the needs of surface inspection during the high-speed paper production process, this paper focuses on the design of the machine vision system and the development of associated algorithms for computationally efficient robust detection and localization of unwanted foreign objects. We address the practical challenges during the design of the architecture of an overall system that consists of cameras, illumination arrangement, and computing platforms. For the processing of images, we use machine vision for feature extraction followed by clustering for the quantification and localization of defects. We also conduct simulations and analyse detection error and localization error to evaluate the performance of the proposed system and algorithms. The proposed system has been deployed in the industry for surface inspection.