基于LSP与GLCM融合的禾本科牧草种子特征提取算法

Seed feature extraction algorithm of Gramineous grass based on the fusion of LSP and GLCM

关键词:  种子识别  纹理特征  局部相似模式  灰度共生矩阵

Key words:  seed identification  textural feature  LSP  GLCM

摘要:

针对禾本科牧草种子相似性较高、识别困难的问题，采用局部相似模式（LSP）和灰度共生矩阵（GLCM）的方法，对禾本科牧草种子的分类识别进行研究。结果表明：1）局部相似模式与灰度共生矩阵融合的方法可以有效的提取禾本科牧草种子的纹理特征，能够识别颜色、形状、大小等特征都十分相似的牧草种子，且其识别率优于传统的LSP特征算子和GLCM特征算子。2）与传统LSP算法相比，结合灰度共生矩阵算法后，得到的特征受到相似种类种子图像的影响较小，具有更广泛的适应性。因此，基于LSP和GLCM的融合算法可以有效地提取相似禾本科种子图像的纹理统计特征，采用线性判别分析分类器（LDA）进行分类，识别率最高达到98.64%。

Abstract:

Aiming at the problem of high similarity and identification difficulties of Gramineous grass seeds,local similarity pattern (LSP) and gray level co-occurrence matrix (GLCM) are combined to study the classification and identification of Gramineous grass seeds.The results show that:1) The fusion of LSP and GLCM can effectively extract the texture characteristics of grass seeds to identify the seeds whose color,shape,size and other characteristics are very similar,and the accuracy is superior to that of the traditional LSP or GLCM operators.2) Compared with the traditional LSP algorithm,the extraction algorithm of LSP fused with GLCM can extract features,which are less affected by the similar species of seed images,representing a wider adaptability in Gramineous grass seed identification.Therefore,the extraction algorithm of the fusion of LSP and GLCM can effectively extract the statistical texture features of similar Gramineous grass seeds,which could reach the highest recognition accuracy of 98.64% when combined with LDA classifier.