【作者】 姚宇飞；

【导师】 袁津生；

【作者基本信息】 北京林业大学 ， 计算机应用技术， 2011， 硕士

【摘要】 我国是林业大国,拥有丰富的林业物种资源,随着环境问题的日益突出,濒危植物的数量激增,又由于人工识别具有效率低、主观性强的缺点,因此,基于计算机技术的植物物种识别具有重要的研究意义。叶片是植物形态结构的重要组成部分,不同植物拥有不同的形状和纹理特征,将其作为林种识别的依据具有很高的准确率。本文首先介绍了分形理论的相关知识,包括分形的定义,不同的测定方法以及如何将分形维数作为纹理特征。接着介绍了特征提取、主成分分析等图像识别所需要的技术。接着具体介绍了基于分形理论的叶片图像识别方法,在经过图像预处理后选取了21维形态和纹理特征,并选取前10个主成分来表示原图像的信息,然后采用SVM分类器进行植物叶片分类。最后对26种叶片(390幅图像)的图像库进行实验,实验结果证明分形维数作为纹理特征的方法能够提高叶片识别的准确率,为实际应用提供了新的思路。更多还原

【Abstract】 Our country owns a large number of plant species and is rich in forest. As environment problem rises, the number of endangered plant is increasing greatly. And the low efficiency and high subjectivity of manual recognition are great defect to plant recognition. Because of all of these, the study of plant recognition technique which is based on computer has important significance. Leaf is an essential part of plant. Different plant has leaves with different shape and texture features, according to these features we can recognize plant species with high accuracy.In this paper, the knowledge of fractal theory is introduced firstly, including the definition and different measure method of fractal and how the fractal can be used as texture feature. Next the technique which is necessary in image recognition is presented, such as feature extraction, principal component analysis and so on. And then the method of leaves recognition based on fractal dimension is implemented.21-dimensional shape and texture features is chosen after image pre-process and select the top 10 principal component to represent leaf image. Then SVM classifier is used to classify the leaves images. Finally, the experiment tested on 390 images which belong to 26 kinds of leaves. The result of experiment proves that fractal dimension which is used as texture feature can increase recognition accuracy. This method provides new way to certain utility value.更多还原