Machine Vision-Based V-Notch Detection Method for Conical Yarn Paper Tubes

Taper yarn paper tube is an indispensable and important product in the spinning industry, for the problems of low precision and low efficiency of traditional V-notch detection means for taper yarn paper tube, this paper proposes a V-notch detection method based on machine vision. Firstly, the corresponding features are extracted using edge detection and ORB corner point detection, and then the PCA technique is used for dimensional reduction, and finally fed into the SVM model for classification prediction. The V-notch dataset of conical yarn paper tube is constructed and enhanced, experiments are carried out on this dataset, the training strategy of cross-validation is adopted and the parameters of SVM are adjusted with the corresponding evaluation indexes, and the optimal model is finally obtained, and the accuracy rate of the proposed method reaches 91.1% after experimental validation and the recall rate reaches 100%, and the algorithm's running time is 0.15 seconds, which can be used in the real-time conical yarn V-notch detection system.