基于图像分析的叶片特征识别相关度研究

Research of Leaf Feature Relevance for Classification Rate Based on Image Analysis Algorithm

为了有效解决叶片图像的分类识别中有效特征的选择问题，本文提出一种基于改进型Sobel算子的新型叶片特征提取算法，提取多种叶片识别特征构建特征矩阵，采用模糊半监督加权聚类算法对不同特征矩阵进行聚类分析，通过对不同种类的植物叶片图像识别率进行对比分析，获取叶片分类识别过程中的关键特征。自测数据集的相关实验结果表明，文中数据集下的叶片边缘特征拥有最高的分类识别率。

In order to solve the problem of how to efficiently select the effective feature for leaf image recognition, a novel leaf feature extraction algorithm, based on the improved Sobel operator, was used to extract multiple features from input images to construct feature matrixes for classification. And a semi - supervised fuzzy cluster algorithm was adopted to take a cluster analysis of different feature matrix. Through the comparison and analysis of different kinds of plant leaf image recognition rate, the key features of the classification and recognition process of leaves were obtained. The experimental results of self -testing datasets indicated that the leaf margin feature in dataset has the highest recognition rate.