带先验信息的SVM决策树的鸟类图像识别研究

Research on birds Image Recognition with SVM decision tree with Priori information

本文在认真研究分析鸟类图像基础上,有效提取了鸟类图像的分类特征,并提出一种带先验信息的SVM决策树解决方案,该方案充分发挥了SVM和决策树的优点,简化了分类器的设计;并利用先验信息分类器分类效率高的优点,在一定程度上弥补了决策树错误累积的缺陷。实验表明,提取的特征简单有效,分类方案合理,获得了较好的分类识别率,实现了对闽江口15种鸟类的有效分类。

Based on the careful analysis of birds images, the paper has extracted classification features of birds images and proposed a classification scheme of SVM decision tree with priori information, which gives full play to the advantages of SVM decision tree and sim- plified classifier design. Besides, it takes advantage of the high efficiency of priori informationclassifier and, to a certain extent, makes uo for the deficiencies of the decision tree in error accumulation. Experiments show that the extracted features are simple and effective, the classification scheme is reasonable which obtains a better classification recognition rate, and it has accomplished an effective classification of 15 kinds of birds in Minjiang River mouth.