An Algorithm for Recognition of Foreign Objects in Transmission Lines with Small Samples

Accurate identification of foreign body images in transmission line channels requires a large number of samples for model training, but the actual foreign body image data sets that can be used for model training are seriously insufficient. In order to solve the problems of model failure, over-fitting and low accuracy caused by too few training samples, a new method for image recognition of foreign objects in transmission line channels under small sample conditions is proposed. This method enhances the image Technology and meta-learning technology are combined to train the U-Net image segmentation network, and finally obtain the foreign body image recognition model of the transmission line channel. Experiments were carried out on the foreign body recognition models that use meta-learning method and those that do not use meta-learning. The results show that the proposed method can accurately identify foreign body images of transmission line channels under a small-scale original data set, and the accuracy rate is greatly improved.