基于农田实景图像的玉米覆盖度估算模型

为实时、准确地获取玉米覆盖度,为玉米长势监测提供科学依据,以农田小气候实景站采集的玉米图像为数据源,利用图像处理技术对玉米图像进行分析处理,结合9种图像颜色指数(NDI、ExG、ExR、ExG-ExR、R-G、G-B、VEG、CIVE、(G-B)/|R-G|),构建了基于图像颜色指数的玉米覆盖度估算模型。结果表明：利用图像颜色指数可以有效估算玉米覆盖度,除颜色指数(G-B)/|R-G|外,其他8种颜色指数估算玉米覆盖度的精度均较高,其中,以VEG为变量的玉米覆盖度估算模型精度最高,预测结果的R²为0.993,RMSE和MAE均最小,分别为0.0270、0.0223。该方法可以有效、快速的提取玉米覆盖度。

The Estimation Model of Fractional Vegetation Cover of Maize Based on the Real Image of Farmland

To acquire the fractional vegetation cover (FVC) of maize on real-time, which could provide a scientific basis for maize growth monitoring, we took the image shot by the digital camera installed in automatic weather observation station as data sources, built FVC estimation models using 9 color vegetation indices of images (NDI, ExG, ExR, ExG-ExR, R-G, G-B, VEG, CIVE and (G-B)/|R-G|) extracted by employing image processing technologies. The results showed that these indices could estimate the FVC of maize effectively with high accuracy except the (G-B)/|R-G|, the vegetative (VEG) was the optimal color vegetation index for estimation of maize FVC. The model with VEG as variable had the highest estimation accuracy for maize FVC, with the highest coefficient of determination (R²), and lowest root mean square error (RMSE) and mean absolute error (MAE) of 0.993, 0.0270 and 0.0223, respectively. The method could be applied to the extraction of crop FVC effectively.