摘要
由于竹子本身的特殊性,竹子分类繁琐复杂,在计算机领域的机器视觉中探究和拓展竹子分类方法,能辅助并促进竹子分类学研究和资源利用等,对相关研究具有重要意义。尝试选择11种竹类叶片,设计拍摄光箱获取共计440幅叶片图像,对其进行相机标定、伽马矫正、灰度化、二值化等预处理操作,再提取出纹理、几何、颜色3类特征共27维数据;采用MATLAB工具箱中的Classification Learner进行分类,分类器采用工具箱中所包含的集成学习(Ensemble)分类方法。经过多次试验证实,基于纹理、几何、颜色3种特征组合的分类方法优于其它特征组合的分类方法,平均分类准确率为94.02%,正确分类率最高可达到94.5%。研究表明,此种竹子叶片分类方法具有可行性,为今后进一步探究竹子分类方法研究提供了一种可靠且有力的方法。
Because of the particularity of bamboo,the classification is complicated.It is of great significance to explore and expand bamboo classification methods in machine vision under the computer field,which can promote scientific research such as bamboo taxonomy research and resource utilization.This paper selected 11 kinds of bamboo leaves,designed a shooting light box to get 440 leaf images,and carried out camera calibration,gamma correction,gray scale,binarization and other preprocessing operations,and then extracted 27-dimensional data of texture,geometry and color features.The classification learner in the matlab toolbox was used to classify.The classifier adopted the Ensemble classification method contained in the toolbox.The results of many experiments show that theclassification method based on the feature combination of texture,geometry and color is superior to the classification methods based on other feature combinations with an average classification accuracy of 93.64%and a maximum correct classification rate of 95%。It shows that the classification method of bamboo leaves is feasible,which provides a reliable and favorable method for further research on bamboo classification methods in the future.
作者
邓莹
金爱武
朱强根
方婷
DENG Ying;JIN Ai-wu;ZHU Qiang-gen;FANG Ting(State Key Laboratory of Subtropical Silviculture,Zhejiang A&F University,Hangzhou 311300,China;Lishui University College of Ecology,Lishui 323000,China)
出处
《软件导刊》
2021年第6期33-38,共6页
Software Guide
基金
中央财政林业科技推广示范资金项目([2019]TS16)
丽水市科技局高层次人才培养项目(2017R13)。
关键词
竹子分类
特征组合
几何特征
纹理特征
颜色特征
bamboo classification
feature combination
geometric characteristics
texture features
color features
