Search

• DNNCA model considers both LAI and LCC effect on reflectance. • Incremental learning presents opportunity for cross-species LAI and LCC estimation. • Preferred learning strategy further improve performance of DNNCA on specified crop. • NNI LCC derived from LAI-LCC curve is effective for diagnosis of crop N status. Nitrogen is an essential nutrient in crop growth cycle and directly affects the photosynthesis of crops. The leaf chlorophyll content (LCC) and leaf area index (LAI) are important for characterizing photosynthetic capacity of crops and are critical indicators for diagnosing nitrogen status of crops. Hyperspectral remote sensing technology provided a means to achieve crop LCC and LAI estimation. However, the redundancy of spectral data and canopy structure effect can cause poor robustness of the estimating models, further hindering the development and application of estimating models across different crop species. In this study, a method based on incremental learning was proposed for the simultaneous estimation of LCC and LAI, and for nitrogen diagnosis across crops. First, for the spectral dataset generated by the PROSAIL model, a deep neural network was used to construct LCC and LAI estimation models (called DNNCA model). Secondly, for the hyperspectral data collected from field crops (soybean, canola and wheat), incremental learning using regularization (LwF algorithm) was used to update the DNNCA model parameters. Finally, the dilution curve model based on the LCC-LAI anisotropic growth relationship was developed to assess crop nitrogen status. The results showed that: (1) The constrained bi-objective optimizated DNNCA model can consider the interactive effect of LAI and LCC on spectral reflectance, and achieved reliable estimation on PROSAIL simulation data set (LAI:R2 = 0.82, RMSE = 0.77 m2/m2; LCC:R2 = 0.91, RMSE = 6.4 ug/cm2). (2) By incremental learning, DNNCA model has continuous learning capability and stable estimation on cross-crop (canola, soybean and wheat) field-measured data (LAI:R2 = 0.64–0.82, RMSE = 0.58–1.02 m2/m2; LCC:R2 = 0.56–0.82, RMSE = 3.9–10.5 ug/cm2). (3) The relationship between the NNI LCC and the NNI LNC was significant. The NNI LCC derived from the anisotropy relationship between crop LAI and LCC was an effective tool for crop nitrogen status diagnosis. (4) The process of crop LAI, LCC and NNI LCC reflected the effect of water and nitrogen supply on crop growth. The appropriate water-nitrogen treatments contributed to LAI increase and LCC accumulation. The study demonstrated that hyperspectral remote sensing technology combined with incremental learning is an effective method for cross-crop growth monitoring and nitrogen diagnosis. These results provide a reference and basis for filed water-nitrogen supply and management. [ABSTRACT FROM AUTHOR]

• Fractional-order derivative algorithm was used to process soil VIS-NIR spectroscopy. • The correlation coefficient with most soil salt ions were best at 0.80–0.90 order. • Three-dimensional indices were used to magnify the soil salinity information. • Iteratively retaining informative variables algorithm had the best effect. • Mg2+ showed the highest estimation accuracy, with an RPD and RPIQ up to 3.81 and 3.71 respectively. This study aimed to improve the potential of visible-near infrared (VIS-NIR) spectroscopy in predicting the contents of salt and major soluble ions in the topsoil at Hetao Irrigation District in Inner Mongolia, China. A total of 120 soil samples (sampling depth is 0–20 cm) were collected from the field. Fractional-order derivatives (FODs) (with intervals of 0.05 and range of 0–2) were used for soil spectral pretreatment. Coefficient of determination (R2) was used to determine the optimal FOD spectrums of the soil salt and major soluble ions. Extreme learning machine (ELM) models were calibrated with such spectral parameters as band reflectance, newly constructed two-band indices, and three-band indices from the optimal FOD spectrum, and the models were subsequently applied to estimate the contents of soil salt and soluble ions. Different algorithms including Monte Carlo uninformative variable elimination (MCUVE), iteratively retaining informative variables (IRIV), and bootstrapping soft shrinkage (BOSS) were applied to variable selection. The results indicated that the correlation between the contents of soil salt and soluble ions and FOD spectra increased as the derivative order increased. The optimal FOD spectrum of soil salt content (SSC), K+, Na+ and SO 4 2− were the 0.9-order derivative spectrum while the optimal FOD spectrum of Ca2+, Mg2+, CO 3 2–, HCO 3 – and Cl− were 1.65, 0.85, 1.90, 0.80 and 1.55 order derivative spectrum, respectively. The newly proposed three-band index (TBI4) showed a better correlation with ions content (e.g., TBI4 of Mg2+ had the best effect, and its R2 was 0.92). In most cases, the IRIV-ELM models outperformed MCUVE-ELM and BOSS-ELM models in estimating the contents of soil salt and soluble ions. Difference existed among the optimal prediction accuracy of the contents of soil salt and different ions: the prediction accuracy of SSC, Ca2+, Na+, Mg2+, Cl− and SO 4 2− was very high while that of CO 3 2− was high and K+ and HCO 3 2− was relatively lower. However, these models have reached and even exceeded the quantitative prediction level. Therefore, we concluded that the contents of soil salt and soluble ions could be estimated using FOD algorithm and the optimal combination of spectral indices. [ABSTRACT FROM AUTHOR]

• Multi-angle spectral-texture observation presents new opportunity for diagnosis. • Non-nadir spectral-texture observation cover more crop growth information. • Stacking ensemble can effectively integrate multiple spectral-texture features. The leaf chlorophyll content (LCC) and leaf area index (LAI) play a crucial role in assessing crop growth status and optimizing field water-fertilizer management. Compared to labor-intensive traditional measurement methods, low-cost unmanned aerial vehicle (UAV) remote sensing technology provides a unique opportunity for monitoring small-scale farmland crop growth information. Currently, the estimation method combining machine learning with spectral or texture information from UAV images has attracted much attention. However, most studies commonly used spectral or texture information from vertical observation, and the potential of multi-angle spectral-texture information has not been fully explored. Meanwhile, the simultaneous usage of multiple characteristic information increases the computational complexity of the model, presenting challenges for crop growth parameter estimation. To address this, this study proposes a multi-feature fusion framework for canola growth parameter estimation using stacking-based ensemble learning algorithm. Firstly, 17 spectral features (VI s and Band) and 136 texture features (TF s) are extracted from UAV spectral images with seven view zenith angle (Nadir, ±20◦, ±40◦ and ± 60◦), respectively. Secondly, important feature variables are selected through correlation analysis and then feature datasets are constructed. Finally, a base learners-meta learner stacking structure was employed to ensemble four machine learning models (SVM,PLSR,RF and GBDT) to predict canola growth parameters. The result shows that: (1) Compared with nadir observation, off-nadir observations (especially for −20◦ and −40◦) can provide more spectral-texture information related to canola growth; (2) The stacking-based ensemble learning can achieve accurate estimation of canola LAI (R2 = 0.72; RMSE = 0.74) and LCC (R2 = 0.78; RMSE = 6.4 ug/cm 2); (3) Compared with single machine learning model (LAI: R2 = 0.56 ∼ 0.67, RMSE = 0.87 ∼ 1.08; LCC: R2 = 0.56 ∼ 0.75, RMSE = 8.1 ∼ 10.6 ug/cm 2), the stacking-based ensemble learning has advantage in growth parameters estimation. The corresponding spatiotemporal mapping reflects the impact of field treatment on the canola growth. Overall, These above results demonstrate the application value of multi-angle spectral-textural information and stacking-based ensemble learning in accessing crop growth, providing new insights into remotely quantitative diagnosis of crop growth. [ABSTRACT FROM AUTHOR]

[Display omitted] • Different window sizes will affect the monitoring accuracy of soil moisture; • It is feasible to select the optimal window size with the local variogram; • The window size of 13 * 13 is the optimal window size applicable in the study area. Soil moisture content monitoring with UAV remote sensing always involves the selection of an appropriate window size for spectral information extraction, but research on the effect of window size on the accuracy of soil moisture content monitoring models and the selection of the optimal window size has been rarely reported. To solve these problems, an experiment was conducted on three typical bare plots in Shahaoqu Experimental Station at Hetao Irrigation District, Inner Mongolia, China. First, remote sensing images were obtained from the three bare plots from April 15 through 17, 2019, with a six-rotor UAV equipped with a six-channel multispectral camera. Synchronously, the moisture content at 0–10 cm of the surface soil was measured using the drying method. Then, the spectral information was extracted through windows of 16 different sizes (ranging from 1 * 1 to 31 * 31). Followed was the construction of thirty spectral indices using the ratio and normalized ratio methods, and the processing of the constructed indices using principal component analysis. The principal components accounting for 95% of the cumulative contribution rate were selected as the input variables for the construction of the monitoring models based on BP neural network. Finally, the model accuracy was tested using ANOVA, and the local variogram of the spectrum was used to explore the optimal window size selection. The results demonstrated: (1) There are differences in the spectral information extracted from different sizes of windows, which affects the accuracy of soil moisture monitoring model; (2) The spatial autocorrelation threshold of the plots at the local variogram was 13 * 13, resembling the window size with the highest accuracy, so it is feasible to select the optimal window size with the local variogram; (3) As the window size of spectral information increased, R2 first increased and then decreased, reaching the maximum value of 0.261 at the size of 13 * 13, and RMSE first decreased and then increased, reaching the minimum value of 0.017 when at the size of 7 * 7. These results can provide some reference for window size selection in spectral information extraction to monitor soil moisture content. [ABSTRACT FROM AUTHOR]

• New texture indices were constructed by texture metrics and effectively used to estimate LAI of maize. • The potential of UAV-based thermal infrared sensor for estimating LAI was proved. • The combination of UAV-based multispectral and thermal infrared data improved the estimation accuracy of LAI. • The RFR model had the strong robustness and achieved the best performance on LAI estimation. Crop leaf area index (LAI) is one of the important indicators to evaluate crop growth and guide field management, and can be used to predict crop yield. Spectral and thermal information extracted by multispectral (MS) and thermal infrared (TIR) sensors mounted on an unmanned aerial vehicle (UAV) can be used for LAI estimation. Image texture is sensitive to the changes in crop surface grayscale or color characteristics, and can be combined with spectral and thermal information to estimate the LAI. But single texture metric has limitations in LAI estimation. Therefore, the purpose of this study is to construct new texture indices based on texture metrics extracted from MS and TIR images, and combined spectral and thermal information to enhance the estimation accuracy of maize LAI. Three replicates of maize experiments under different irrigation treatments were conducted in 2020. The MS and TIR sensors were mounted on a UAV to acquire maize canopy images during critical growth stages and acquire field LAI value of samples synchronously. The LAI estimation models were established using MS data, TIR data, as well as their combination. These models were constructed by Back Propagation Neural Network (BPNN), Partial least squares regression (PLSR), and Random Forest Regression (RFR). Finally, the performance of LAI estimation models was evaluated by the coefficient of determination (R2), root mean square error (RMSE) and relative root mean square error (rRMSE). Results shown that: (i) Among the eight kinds of texture metrics extracted from MS and TIR images, the texture metric mean (MEA) has the best performance. Compared with single texture metrics, texture indices constructed by different metrics has stronger correlation with LAI. (ii) Adding texture indices to estimation models significantly improved model accuracy, especially multispectral three-texture index(MS-TTI) has higher LAI estimation potential than thermal infrared three-texture index(TIR-TTI). (iii) Compared with the use of MS or TIR data alone, the estimation model constructed by combining MS data and TIR data have better performance. The best estimation model obtained by the RFR method (R2 = 0.862, RMSE = 0.246 and rRMSE = 10.20 %) further improved the LAI estimation of maize, with R2 increasing by 6.55 % and 14.48 %, respectively. In conclusion, the combination of MS and TIR data can effectively improve the estimation accuracy of maize LAI, and also provide a feasible method for monitoring crop growth. [ABSTRACT FROM AUTHOR]