论文部分内容阅读
精确施药的关键是快速正确识别杂草。为此,利用ASD野外便携式光谱仪,在田间测量了大豆、马唐和稗草植株冠层在350～2 500nm波长内的光谱数据,经过数据预处理,数据分析波长选为350～1 300nm和1 400～1 800 nm。数据处理采用支持向量机(SVM)模式识别方法 ,用线性、多项式、径向基和多层感知核函数对大豆和杂草建立二分类模型。结果表明:三阶多项式核函数SVM分类模型的正确识别率最高,达到85%以上,且支持向量比例较小;以二分类模型为基础,利用投票机制建立了大豆、马唐和稗草的一对一多分类SVM模型,正确识别率达83%;田间光谱测量受光照、背景和仪器测量精度等条件的影响较大,但结果仍表明SVM结合光谱技术在田间杂草识别中的应用潜力很大。此研究为田间杂草识别及传感器的建立提供了研究思路和应用基础。
The key to precise application is to quickly and correctly identify weeds. Therefore, spectral data of soybean canopy, Crabgrass and barnyardgrass canopy were measured in the field with 350 ~ 2500 nm wavelength using ASD field portable spectrometer. After data preprocessing, the wavelength of data analysis was selected as 350 ~ 1300nm and 1 400 ~ 1 800 nm. Data processing uses support vector machine (SVM) pattern recognition method to establish a binary classification model of soybean and weeds with linear, polynomial, radial basis and multi-layer perceived kernel functions. The results show that the correct recognition rate of the SVM classification model with the third-order polynomial kernel function is the highest, reaching more than 85% and the support vector proportion is relatively small. Based on the dichotomous model, the voting mechanism is used to establish the SVM classification model of soybean, crabgrass and barnyardgrass For a multi-classification SVM model, the correct recognition rate is 83%. Field spectroscopy has a great influence on illumination, background and instrument measurement accuracy, but the results still show that the potential of SVM combined with spectral techniques in field weed recognition is very high Big. This study provides a research idea and application basis for field weed identification and sensor establishment.