Research on Tree Classification Algorithm Based on Morphology and Leaf

For a long time, forestry workers carry out forestry field investigation work, when they do not know the tree species, they can only search the relevant literature such as “tree annals”. The query procedure is tedious and the accuracy is not high, which seriously affects the efficiency of forestry production. In this paper, after data augmentation of tree images, GAP (Global Average Pooling) was used to fine-tune Inception-V3, ResNet50 and Xception for transfer learning. Feature vectors of the model were extracted and merged, and two feature vectors of tree overall shape and leaves were obtained through training. After feature fusion, SoftMax was used for classification, and the result of tree species determination was obtained. The practice has proved that the tree classification method proposed in this paper has a great advantage and the accuracy is higher.