Pipeline weld defect detection method based on improved yolk network

The invention provides an improved yolk-based pipeline weld defect detection method, which comprises the following steps of: preprocessing an acquired pipeline weld defect image, and constructing an improved yolk network model, and specifically comprises the following steps of: performing feature extraction on an up-sampling feature extraction module by adopting two improved residual network structures; the method comprises the following steps of: designing a data training set to realize double growth of detection precision and detection speed of a network model, designing a regularization cross entropy loss function to improve the problem of sample imbalance, finally performing parameter training on the model by using the data training set, and testing the trained model by using a test set. According to the invention, the detection precision and the detection efficiency of the pipeline weld defect are effectively improved. 本发明提出一种基于改进yolo的管道焊缝缺陷检测方法，该方法先对采集到的管道焊缝缺陷图像进行预处理，构建改进yolo网络模型，具体实施方法为：在上采样特征提取模块采用两个改