A Plant Recognition Method Based on Local Linear Embedding and Multi-Feature Fusion

Abstract

In order to acquire the accurate and practical feature of leaves to improve the accuracy of leaves image recognition, a new method based on local linear embedding and multi-feature fusion is proposed. Firstly, the LBP algorithm is used to extract texture features of leaves. Considering the high dimension of LBP feature, local Linear embedding (LLE) algorithm is used to reduce the dimensions of leaf texture features. At the same time, the shape features of the leaves are considered. The LBP rotation invariant features are combined with the shape features. Then using the K nearest neighbor (KNN) method to recognize plant leaves. Experimental results based on the leaf image database prove that the proposed method has high recognition accuracy and stability.