论文部分内容阅读
田间全面积均匀喷施除草剂不经济,还污染环境,精准喷施除草剂意义重大,其关键是正确识别杂草。用便携式野外光谱仪,在田间测量了玉米、马唐和稗草植株冠层在350~2 500 nm波长范围内的光谱数据,经过数据预处理,数据分析波长选为350~1 300和1 400~1 800 nm。数据处理采用支持向量机(SVM)模式识别方法。SVM具有可实现对小样本建模结构风险最小化、结果最优化、泛化能力强的优点。用线性、多项式、径向基和多层感知核函数对玉米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80%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,建立了玉米、马唐和稗草的一对一多分类SVM模型,正确识别率达80%。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一种研究思路和应用基础。
Uniform spraying of herbicides in a uniform area in the field is not economical, but also pollutes the environment. It is of great significance to precisely spray herbicides. The key point is to correctly identify weeds. Spectral data of corn, crabgrass and barnyardgrass canopy in the wavelength range of 350-2 500 nm were measured in the field with a portable field spectrometer. After data preprocessing, the wavelength of data analysis was selected as 350-1 300 and 1400-400 nm. 1 800 nm. Data processing using support vector machine (SVM) pattern recognition method. SVM has the advantages of minimizing the structural risk of modeling small samples, optimizing the result and generalizing ability. The binary classification model of maize and weeds was established by linear, polynomial, radial basis and multi-layer perceived kernel function. The results show that the correct recognition rate of the SVM classification model with the third-order polynomial kernel is the highest, reaching more than 80%, and the ratio of support vector Smaller. Based on the dichotomous model, a one-to-one multi-classification SVM model of maize, crabgrass and barnyardgrass was established using the voting mechanism. The correct recognition rate was 80%. Field spectroscopy has a great influence on illumination, background and instrument measurement accuracy, but the results still show that SVM combined with spectroscopy has great potential in the field of weed recognition. This study provided field weed recognition and sensor establishment A research idea and application basis.