Classification of Cotton and Flax Fiber Images Based on Inductive Transfer Learning

Aiming at the existing problems of high labor cost, huge training data and long detection period for Identification technology of cotton flax fiber, which is based on textural feature and convolutional neural network (CNN) method. In this paper, it proposed a cotton and flax fiber detection method based on transfer learning. According to sharing the weight parameters of the convolutional layer and the pooling layer, the model hyperparameters can be adjusted for the new network to achieve high detection accuracy. The experimental results show that the detection accuracy of cotton flax fiber obtained by transfer learning is up to 97.3%, the sensitivity is 96.7%, and the specificity is 98.2%. Compared with traditional machines, transfer learning method have large increase in the three indicators. Furthermore, the transfer learning method has shorter training time and fewer data sets.