Carbon fiber defect detection method based on dark field

The invention discloses a carbon fiber defect detection method based on a dark field, and the method comprises the steps: firstly inputting a dark field illumination image into a defect detection system; secondly, filtering image noisy points through a hybrid filtering technology; thirdly, carrying out binarization on the image by using an OSTU algorithm; then, carrying out closed operation to highlight and connect image edges; then, carrying out image region segmentation by using BOLB analysis; and finally, performing image multi-classification defect identification through an SVM algorithm. According to the method, online automatic detection of the carbon fiber quality defects can be realized, and the working efficiency of carbon fiber quality defect detection is remarkably improved; meanwhile, manual detection errors are eliminated, the stability of a detection result is improved, and the detection success rate of small defects of the carbon fiber tow is increased. 本申请公开了一种基于暗视野的碳纤维缺陷检测方法，首先，将暗场照明图像输入缺陷检测系