Research On The Classification Method Of Plant Leaves

In the field of computer vision,the automatic classification of plant leaves is an important application direction and an inevitable product of the combination of biological knowledge and artificial intelligence technology.Using leaf images as the basis for classification to identify plant species is gradually becoming the most efficient and accurate general method.Therefore,studying leaf image feature extraction methods and different types of leaf classification methods are of great significance for improving the accuracy of leaf classification.On this basis,this article mainly studies the effective methods of plant leaf classification and improves the accuracy of classification through digital image processing and the calculation of feature descriptors.The main work of this paper has the following three points:1.A low-dimensional feature extraction method for plant leaves is proposed.The method is mainly to extract the shape,moment invariant and texture features of the leaf image.Four basic shape features and two kinds of moment invariant features are calculated on the binary image of the leaf,and seven inferred shape features are proposed on the basis of the basic shape feature.The texture feature of the leaf is described by the gray-level co-occurrence matrix parameters.The experimental results of feature extraction prove that the above three features can completely describe the overall feature information of the leaves,and can maintain good stability and invariance.2.Propose the extraction method of Gabor feature and LBP feature of plant leaves.The gray leaf image is equally divided into blocks,the Gabor feature values of the image blocks are calculated on each sub-image block,the LBP feature value of the gray image is calculated and the pixel distribution of the image is calculated,and the above two features are fused in series to form the feature vector of the image.The features are processed by normalization and feature dimensionality reduction.The experimental results prove that the features extracted by this method have strong classification ability.3.Perform plant leaf classification experiments on the Flavia dataset,Swedish dataset and ICL dataset for different feature extraction methods and analyze the experimental results.The experiment proves that the two leaf feature extraction methods proposed in this paper are used in the classification experiment.It has high classification accuracy and universal applicability.