A tree species recognition method based on transfer learning

Abstract

The invention discloses a tree species identification method based on transfer learning, comprising the following steps: S1, collecting tree images and making tree species image data sets; S2, data enhancement is performed on the original data set image to expand the number of pictures; S3, acquiring more than one pre-training model based on convolution neural network trained on large image dataset, S4, training the pre-training model with tree species image dataset, optimizing one or several full connection layers of the pre-training model in the training process, thereby training a pluralityof classifiers based on convolution neural network; Test the accuracy of each classifier and select the classifier with the highest accuracy. S5 performs tree species identification using the classifier selected in step S4, thereby obtaining an identification result. The invention recognizes tree species based on the transfer learning mode, so that a model with good generalization ability can betrained under the condition of limited sample size of tree species, and the tree species recognition rate under the complex background is greatly improved.

Description

Technical field [0001] The invention relates to a tree species identification method, in particular to a tree species identification method based on migration learning. Background technique [0002] Tree is a kind of plant, generally refers to woody plant. The latest research shows that there are 60065 species of trees in the world, and this number is constantly updated, because about 2000 species of trees are newly named every year. There are many species, including about 8,000 species of trees in China. In the face of a huge amount of tree species information, it is obviously unreasonable to rely on manual identification alone. It is not only laborious, but also inaccurate. And the recognition by the human brain requires researchers to have considerable experience. Therefore, it is necessary to propose a method that can automatically identify the type of tree through the image of the tree. This has certain positive significance for the construction of forestry informatizati...

AI Technical Summary

Problems solved by technology

However, this method still has certain defects. This method trains a classifier with a limited number of samples, and the obtained classifier is directly used for image recognition, which cannot guarantee the reliability of the classifier and the accuracy of recognition.

Images