Smart Classification System of Solid Wood Board Based on Convolutional Neural Networks

Solid wood board classification is the most important detection process in the production of solid wood integrated materials. At present, most of the detection links still use artificial methods. In this paper, the network model of CSPDarkNet is lightweight and introduced into the efficient channel attention (ECA) mechanism, the ECA-CSPDarkNet algorithm model is built. It is used for the preferred classification of the solid wood board, which can improve the automation of the production process of solid wood integrated materials, reduce the labor of workers and reduce waste of wood. According to the actual needs of solid wood plate classification, we collect 9 kinds of surface feature images of solid wood boards to construct a data set of 9,000 images. Using the proposed ECA-CSPDarkNet model to train, test, and compare with the detection accuracy and detection speed of other convolutional neural networks models. Finally, the average detection accuracy of 9 kinds of samples reached 98.44%, and the detection speed reached 5. 88ms/sheet.