摘要
为了解决塔式小波分解中丢失高频信息的问题,提出了将塔式小波分解和灰度共生矩阵融合的方法,生成小波灰度共生矩阵特征来描述植物叶子纹理,结合具有尺度、平移和旋转不变性的形状特征,生成一组有效的分类特征向量来对植物种类进行分类预测。用支持向量机(support vector machine,SVM)等分类器对两组实验数据进行分类测试,分类准确率分别达到了97.2426%和96.7972%。实验结果表明,小波灰度共生矩阵特征能够有效地描述植物叶子纹理特征,具有很强的分类能力。
To solve the problem of high-frequency information loss in tower wavelet decompositions, a new method is proposed to generate an effective feature vector that has texture features integrating tower wavelet decomposition with gray-level co-occur- rence matrix and shape features of scale, translation and rotation invariance combined to classify plant leaves. Finally, experi- ments are conducted using SVM （support vector machine） and other classifiers to classify two datasets, which achieve accuracies of 97. 2426% and 96. 7972% respectively. Experimental results show that the proposed method is effective in describing the plant leaf textures and powerful to classify plant species.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第12期4774-4778,共5页
Computer Engineering and Design
