Steel plate surface defect detection method with marks based on convolutional neural network

The invention discloses a steel plate surface defect detection method with marks based on a convolutional neural network, and the method comprises the steps: firstly collecting a steel plate image with surface defects, and constructing an image library; dividing the images in the image library into a training sample set and a test sample set; carrying out steel plate image preprocessing, steel plate image edge detection and steel plate defect marking in the steel plate image, and constructing a convolutional neural network; and finally carrying out steel plate surface defect detection identification and classification. Collected steel plate images are subjected to preprocessing and edge detection, and then the steel plate images are trained. Under the condition of using small samples, theaccuracy and precision of multi-type defect detection on the surface of the steel plate are improved. 本发明公开了一种基于卷积神经网络带标记的钢板表面缺陷检测方法，首先采集带有表面缺陷的钢板图像，构建图像库；将图像库中的图像分为训练样本集和测试样本集；然后进行钢板图像预处理，钢板图像边缘检测，钢板图像中钢板缺陷标记，构建卷积神经网络；最后进行钢板