AUTHOR=Gu Chen , Ji Shu , Xi Xiaobo , Zhang Zhenghua , Hong Qingqing , Huo Zhongyang , Li Wenxi , Mao Wei , Zhao Haitao , Zhang Ruihong , Li Bin , Tan Changwei TITLE=Rice Yield Estimation Based on Continuous Wavelet Transform With Multiple Growth Periods JOURNAL=Frontiers in Plant Science VOLUME=Volume 13 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.931789 DOI=10.3389/fpls.2022.931789 ISSN=1664-462X ABSTRACT=Yield is an important indicator in evaluating rice planting, and it is the collective result of various factors over multiple growth stages. To achieve a large-scale accurate prediction of rice yield, based on yield estimation models using a single growth stage and conventional spectral transformation methods, this study introduced the continuous wavelet transform (CWT) algorithm and constructed models under the premise of combined multiple growth stages. In this study, the canopy reflectance spectra of four important stages of rice, the elongation stage, heading stage, flowering stage, and milky stage were selected to analyze the correlations between the original spectrum, first derivative transformation, continuum removal, and wavelet transformation with rice yield. Furthermore, the vegetation indices and hyperspectral parameters together were taken as screening objects to establish multivariate stepwise regression models based on first-derivative transformation. Their accuracies were compared with those of the established multivariate stepwise regression (MSR) models based on the first derivative-wavelet transform. The results showed that the sensitivity of the original spectrum to rice yield was substantially improved after the first derivative-wavelet transform, and the yield estimation model established based on the first derivative-wavelet transform presented the best accuracy. In addition, the combination of multiple growth stages significantly improved the model accuracy. The optimal combination of multiple growth stages was elongation-heading-flowering-milky stages, resulting in a coefficient of determination (R2) of 0.81, a root mean square error (RMSE) of 37.6 g.m-2, and a mean absolute percentage error (MAPE) of 4.8% for the training set, an R2 of 0.77, an RMSE of 36.3 g.m-2, and a MAPE of 4.7% for the validation set 1, and an R2 of 0.72, an RMSE of 43.7 g.m-2, and a MAPE of 5.9% for the validation set 2. The research results demonstrated that the established model could accurately predict rice yield, providing technical support and a foundation for large-scale statistical estimating of rice yield.