Abstract:

Objective Sizing is one of the key processes before weaving in the textile industry, and it requires the use of a sizing machine. During the process of sizing, the yarn on the warp beam of the sizing machine often breaks, leading to yarn accumulation and wastage. Currently, inspections for broken yarn on the warp beam zone of sizing machines predominantly rely on manual patrols, which suffer from poor timeliness and low efficiency. To address this issue, an algorithm was proposed in this research for detecting broken yarn faults in the warp beam zone of sizing machines based on machine vision.

Method By analyzing the image features captured by the cameras, an algorithm for determining the parameters of image recognition was investigated based on the theory of vanishing points, comprising an automatic vanishing point extraction algorithm, a warp roller axis calibration algorithm, and an automatic region of interest (ROI) selection algorithm. With the determined parameters of image recognition, the original images captured by the cameras was preprocessed, followed by edge detection and ellipse fitting. Subsequently, by assessing the positional relationship between the fitted ellipse and the top and bottom edges of the yarn roller, the recognition of broken yarn faults in the warp beam region was achieved.

Results The results demonstrated the significant effectiveness of the image recognition algorithm based on the theory of the vanishing point. Firstly, the automatic vanishing point extraction algorithm, using exhaustive search, successfully and consistently identified the vanishing point within a coordinate variation range of 5 pixels, ensuring the accuracy of subsequent analysis. Secondly, the vanishing point-centerline calibration method improved the accuracy of the axle-centerline identification by addressing the deviation issue of the dual-ellipse centerline method, providing a crucial geometric reference for broken yarn detection. Additionally, by utilizing the vanishing point information, the automatic selection algorithm for the ROI could accurately partition the ROI region that varies with the roller even in cases of smaller roller radius and increased background interference, enhancing the robustness of the algorithm. Based on the comprehensive algorithm mentioned above, the broken yarn detection algorithm performed edge detection on the preprocessed image and used ellipse fitting to fit the contour of the formed protruding circular white yarn loop. The relationship between the fitted ellipse and the roller edge lines was used to determine broken yarn faults, and the algorithm accurately identified the intersection points between the fitted ellipse and the upper and lower roller edge lines, showing good performance in detecting broken yarn faults in the middle and end sections. The detection algorithm took approximately 700 to 800 ms to process a single image. By concurrently processing images from multiple cameras, the time required to handle multiple images was significantly reduced. Experimental results indicated a 0.91 accuracy of the detection algorithm.

Conclusion Practical application shows that under most normal conditions in the sizing production process, the broken yarn fault identification algorithm for the warp beam zone can accurately and promptly detect broken yarn issues and trigger the corresponding alarms, meeting the needs of the project and actual industrial production. This detection system has been applied in relevant textile enterprises, confirming its feasibility.

Key words: warp beam zone of sizing machine, broken yarn detection, machine vision, image vanishing point, ellipse fitting