摘要
提出一种高效的基于PCA和FisherTrees特征融合的木材识别方法,首先把训练样本分别投影到PCA和FisherTrees空间,得到PCA特征和FisherTrees特征;然后通过算术均值、交换转置均值和加权均值进行特征融合,将融合后的特征应用不同距离函数的分类器进行分类。结果表明:通过交换转置均值融合PCA和FisherTrees特征,然后使用余弦角分类器能获得最好的识别效果。
A new efficient method based on feature fusion of PCA and FisherTrees for wood identification was proposed in this paper. Firstly, the training samples were projected into PCA and FisherTrees space respectively to form the PCA and FisherTrees features, then the two features were fused through three ways, i. e. arithmetic mean, swapping transposition mean and weighting mean. Finally, the feature fusion was applied to classify with different distance functions. The experimental results showed that the new method had a higher recognition rate and was more efficient compared with the tradition subspace methods. The best identification result could be obtained by features fusion of PCA and FisherTrees with swapping transposition mean and by the cosine distance function classifier.
出处
《林业科学》
EI
CAS
CSCD
北大核心
2013年第6期122-128,共7页
Scientia Silvae Sinicae
基金
国家自然科学基金项目(30972361)
浙江省教育厅重大科研攻关项目(ZD2009002)
浙江省自然科学基金项目(Y13C160027)
浙江农林大学研究生科研创新项目(3122013240224)
作者简介
汪杭军为通讯作者。
