With the development of the computer and the digital image processing technology, application of machine vision technology to make online detection research has been gradually developed, so the machine vision technology has been introduced into the registration detection of rotary screen printing, in order to enhance the accuracy and speed of the registration error’s detection, that is a very significant research topic. Firstly, general planning about the rotary screen printing machine registration detection system has been designed. In addition, texture noise of printing picture has been analyzed, which has been collected by smart camera. Finally, based on mean shift arithmetic for color image segmentation has been put forward. Experiments show that the algorithm based on mean shift segmentation method for color image segmentation textile printing can achieve good results, moreover that can accurately carry out the fabric chromatic region segmentation, which for the registration error detection system design on rotary screen printing machine lays the foundation.