基于无人机多光谱遥感和机器学习的苎麻理化性状估测

苎麻生理生化性状是其遗传基础和环境条件综合影响的结果，能够反映特定胁迫环境下苎麻的生长发育状况。无人机遥感技术为大规模田间作物长势监测提供了有效手段，利用无人机搭载多光谱相机对苎麻理化性状进行综合评价具有实际意义。因此，以苎麻种质资源为研究对象，采用无人机多光谱遥感获取苎麻冠层的光谱参数和纹理参数，运用相关性分析法(Pearson correlation analysis, PCA)、递归特征消除法(Recursive feature elimination, RFE)2种最优特征筛选方法和线性回归(Linear regression, LR)、决策树(Decision tree, DT)、随机森林回归(Random forest, RF)、支持向量机(Support vector machines, SVM)、偏最小二乘回归分析(Partial least squares regression analysis, PLSR)5种机器学习算法分别构建了苎麻叶绿素相对含量(SPAD值)、叶面积指数(Leaf area index, LAI)和叶片相对含水量(Relative water ...

Estimation of Ramie Physicochemical Property Based on UAV Multi-spectral Remote Sensing and Machine Learning

The physiological and biochemical properties of ramie are the result of comprehensive influence of genetic basis and environmental conditions, which can reflect ramie growth under specific stress environment. Therefore, a fast, accurate and inexpensive method is needed to monitor the dynamic changes of ramie physicochemical property during the whole growth cycle. Unmanned aerial vehicle (UAV) remote sensing technology provides an effective means for monitoring crop growth in large field, which has been widely concerned and applied by virtue of its advantages of fast, non-destructive, timely and accurate. However, at present, there are few researches on the comprehensive evaluation of ramie physicochemical property by using UAV multi-spectral images. The UAV was equipped with a multi-spectral camera to acquire the multi-temporal canopy images of ramie. Then, the canopy orthophoto image was obtained by DJI terra, and the spectral and texture characteristic values of ramie plants were further extracted. Pearson correlation analysis (PCA) and recursive feature elimination (RFE) were used to screen the sensitive eigenvalues. Finally, based on multi-temporal remote sensing data, linear regression (LR), random forest regression (RF), support vector machines (SVM), partial least squares regression analysis (PLSR) and decision tree (DT) were used to estimate ramie physicochemical property, respectively. The results showed that there was a significant correlation between the ramie physicochemical property and spectral skewness parameters. Both PCA and RFE can improve the accuracy of the estimation model, but RFE had better performance. The accuracy of the LR-SAPD estimation model was 0.662. The R2 and RMSE of LR-RWC estimation model were 0.793 and 2.213%, respectively. The SVR-LAI model could better estimate ramie LAI (R2=0.737, RMSE was 0.630). In conclusion, an accurate, efficient, cost-effective and universal dynamic monitoring method for physicochemical property of field ramie was proposed.