Abstract:

In order to avoid the fraction caused by mechanical remaining yarn inspecting devices in combined spinning and winding frames and preserve the quality of surface yarn around bobbins, a non-contact algorithm based on computer vision designed for bobbin inspection was introduced. Each bobbin tube was detected and segmented from a single frame. The interest region(ROI) of a single bobbin was then selected from the background and adjusted to vertical state. An optimized odd Gabor filter group was implemented on the ROI to enhance boundaries between the tube and yarn. Finally, the yarn region was examined and confirmed by connectivity checking. In the next section, the influence of ROI extraction threshold on detection accuracy was analyzed by experiments. The optimization of determining rule of filter parameters was elaborated based on maximum output princeple. The reaults show that the odd  Gabor filter with the ratio of wavelength to standard deviation of 4.808, the ratio of standard deviation to diameter of 0.81 and the direction parallel to the yarn has the best effect on remaining yarn inspection.

Key words: combined spinning and winding frame, bobbin sorting, Gabor filter, maximum response, edge detection