基于U-Net卷积神经网络的轮毂缺陷分割

为实现轮毂缺陷检测自动化,该文依据轮毂X射线图像,提出一种基于U-Net卷积神经网络的自动分割的改进方法。将原始U-Net模型的最大池化操作替换为卷积操作,并加入Dropout层对网络进行优化,提升模型可靠性。同时对带有缺陷的轮毂图像做数据预处理,用于训练改进的U-Net模型。结果表明,该网络在复杂轮毂X射线图像的缺陷识别中表现良好,DICE系数为0.8554,SSIM系数为0.9655,识别速度达到3 ms/张;该方法能较好地实现轮毂射线图像缺陷的自动分割,满足无损检测的自动化需要。

Hub Defect Segmentation Based on U-Net Convolutional Neural Network

In order to automate wheel defect detection,this paper proposes an improved method of automatic segmentation based on U-Net convolutional neural network based on wheel X-ray images,replaced the maximum pooling operation of the original U-Net model with a convolution operation. And added the Dropout layer to optimized the network and improved the reliability of the model. At the same time,data preprocessing of the wheel image with defects is used to train the improved U-Net model. The results show that the network performs well in defect recognition of complex wheel X-ray images,with a DICE coefficient of 0.8554,a SSIM coefficient of 0.9655,and a recognition speed of 3 ms/sheet. This method can well realize the automatic segmentation of wheel ray image defects and meet the needs of non-destructive testing automation.