基于图像分析的叶片特征识别相关度

Leaf feature relevance for recognition rate based on image analysis algorithm

为有效解决叶片图像的分类识别中有效特征的选择问题,探索提高原始叶片图像识别准确度的有效途径,提出一种基于改进型sobel算子的叶片特征提取算法。抽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,数据集下的叶片边缘特征拥有最高的分类识别率。

To solve the problem that how to select the effective feature for leaf image recognition efficiently and to explore the optimal ways to improve the recognition accuracy of the original leaf images,a leaf feature extraction algorithm based on improved sobel operator was proposed and the multiple features were extracted from input images to construct feature matrixes for classification.A semi-supervised fuzzy clustering algorithm with feature discrimination（SFFD）with obtained matrixes was adopted to analyze the relevance for leaf classification.As a result,the key features were acquired during the process of leaf classification.Experimental results with self-testing datasets indicate that the margin feature wins the highest recognition rate with the datasets proposed.