【作者】 邓巍； 赵春江； 何雄奎； 张瑞瑞； 许刚； 陈天恩；

【Author】 DENG Wei~1,ZHAO Chun-jiang~1,HE Xiong-kui~2,ZHANG Rui-rui~1,XU Gang~1,CHEN Tian-en~1 1.National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China 2.College of Science,China Agricultural University,Beijing 100193,China

【机构】 国家农业信息技术研究中心； 中国农业大学理学院；

【摘要】 分别用径向基神经网络(RBF-NN)、支持向量机(SVM)和决策树(DT)三种模式识别方法对相同数据建模,以比较此三种算法应用于田间杂草光谱识别的效能和潜力。光谱数据用美国ASD公司生产的便携式野外光谱仪在田间测得,测量对象为玉米、马唐和稗草幼苗植株冠层,波长范围为350～2 500 nm,训练和检验样本分别为55和59个。研究结果表明,SVM、RBF-NN和DT的3分类正确识别率分别为80%,78%和63%,SVM的分类正确率最高,且精度较高,此方法更适合应用于田间小样本数据的分类和分析。更多还原

【Abstract】 One of the key issues for weed spectral discrimination in fields is to select an appropriate pattern-recognition method in order to improve the correct recognition rate of analyzing the small-size samples.Three kinds of pattern-recognition patterns, namely,Radial Basis Function Neural Network(RBF-NN),Support Vector Machine(SVM),and Decision Tree(DT),were used to model the same data set,so as to compare the performance and potential of the three methods when they were used in the weed spectral discrimination in fields.The spectral data were measured in fields by using a handheld FieldSpec~R 3 Spectroradiometer manufactured by ASD Incorporated Company in USA.The measuring objects were canopies of seedling corns,Dchinochloa crasgalli,and Echinochloa crusgalli weeds,and the wavelength range was 350-2 500 nm.Training and testing samples were 55 and 59,respectively.The predicting results of testing samples showed that the correct rates of 3-classification for SVM,RBFNN, and DT were 80%,78%,63%,respectively.The correct classification rate of SVM was the highest which also had relative high accuracy.It indicated that SVM were optimal to classification and analysis of small-size samples measured in fields.更多还原