Real-time detection of yarn evenness based on image features

Abstract

In order to obtain the yarn quality change information in the spinning process in time and optimize the yarn process parameters, a real-time measurement method for detecting the yarn evenness is proposed. In order to improve the measurement efficiency, firstly, the minimum bounding rectangle was used acquire the region of interest (ROI) of the real-time continuous yarn image, and automatically cut the acquired yarn image according to the yarn direction. Then the yarn evenness binary image was obtained based on OTSU threshold segmentation and morphological operation. Finally, the SIFT algorithm was used to detect feature points and remove overlapping data, ensuring the integrity and continuity of yarn data，proposing a method for calculating yarn diameter by adding yarn inclination angle. Based on this method, we calculated the yarn diameter of three types of linear density compact spun pure cotton yarn and detected yarn dryness and defects. The average error of the experimental measurement diameter is 2.38%. The CV value of the variation coefficient is basically consistent with the test results of the Uster Classimat5 yarn evenness tester, indicating that this method can meet the real-time detection of yarn evenness with a yarn motion speed of 5~30 m/min.

Keywords

• yarn diameter
• yarn evenness
• real-time detection
• minimum circumscribed rectangle
• morphological operation
• scale-invariant feature transform（sift）algorithm