多特征与局部线性嵌入融合算法在植物识别中的应用研究

针对植物识别过程中叶片旋转状态下的识别需求, 采用植物多特征提取与局部嵌入融合算法, 应用支持向量机(SVM)建立分类器对植物叶片进行分类辨识。结果表明:基于分块的局部二值模式(LBP)算法可以提取植物叶片的纹理特征; 使用局部线性嵌入(LLE)算法, 对高维的LBP特征进行降维, 减少了分类识别时间, 同时能够达到更好的聚类效果, 有效地提高识别率; 所提出的植物叶片识别方法对旋转状态下的叶片具有良好的实用性。

Application Research of Multi-Feature and Locally Linear Embedding Fusion Algorithm in Plant Recognition

For the demand of leaf recognition under rotation condition, the plant multi-feature extraction and local embedding fusion algorithm was applied to classify plant leaves by Support Vector Machine (SVM).Results showed that the texture feature of the leaf was extracted by Local Binary Pattern (LBP) algorithm based on leaf block.Using Locally Linear Embedding (LLE) algorithm, dimensionality reduction of high dimensional LBP features reduced the classification and recognition time, and at the same time could achieve better clustering effect and effectively improved the recognition rate.The proposed method for identifying plant leaves had good utility for the leaves in the rotated state.