Study On Growth Monitoring Based On Fusing Multi-Source Remote Sensing Information In Wheat

As the frontier modern information technology, the remote sensing can collect the nitrogen and growth status of crop in the field rapidly and accurately in large scale, which offers important technical support for implementation of precision farming and realizing high yield, good, quality and high efficiency in modern agricultural production. Multi-source remote sensing information can obtain more abundant and accurate information than single remote sensing data, and can improve the accuracy of analysis and extraction of remote sensing information in monitoring crop growth.In this study, the technology based on fusing multi-source remote sensing information was applied to growth monitoring in wheat. A series of field experiments with different nitrogen levels were carried out, multi-plat wheat spectral reflectance were obtained with different scales and temporal images and ground hyperspectral spectroradiometer, field sampling and testing synchronously implemented. Then, based on the technology of fusing multi-source remote sensing information and analysis of spectral reflectance and mathematical statistic, the models that monitoring wheat nitrogen status and growth characters based on fusing multi-source remote sensing information were established in this paper. The result can offer a technological support for monitoring wheat growth status with remote sensing in large scale.A pure pixel spectrum extraction method was proposed based on spectral response function and pixel unmixing by coupling SPOT-5, ground-spectrum and field measured data of different wheat ecological zones in this paper. The quantitative relationships between leaf nitrogen concentration (LNC) and leaf nitrogen accumulation (LNA) of wheat and simulated, measured and pure pixel spectra have been developed. The results showed that the sequence of estimation accuracy was simulated, pure and measured pixel spectra respectively. But leaf nitrogen status monitoring model based on simulated pixel spectra couldn’t be extrapolated directly to regional level. In addition, the model testing results based on independent data indicated that the monitoring model based on pure pixel spectra performed well in different wheat ecological areas. Which maybe contribute to pixel unmixing based on integrating ground-and space-remotely sensed data. Therefore, the pure pixel spectrum extraction method can be applied to remotely sensed data with different spatial and spectral resolutions and estimate wheat nitrogen status in regional scale.The performance of different scale remote sensing data was evaluated in terms of the accuracy of monitoring model based on coupling ground-spectrum, SPOT-5, HJ-CCD and field measured data of different wheat ecological zones, years and nitrogen level in this paper. The results showed there are differences in the reflectance of same bands based on different scale remote sensing data, but their near-infrared band were linearly related significantly to LAI and leaf dry weight. The sequence of estimation accuracy was ground-spectrum, simulated, pure and measured pixel spectra respectively. In addition, SPOT-5and HJ-CCD images have a highly uniform spatial distribution pattern of wheat LAI and leaf dry weight, but the former was higher than the latter in the estimation accuracy. Nevertheless, these results offer a technological support for regional quantitative monitoring of wheat.By combining the techniques of linear pixel unmixing and data assimilation, the LAI based on SPOT-5image with high spatial resolution was used to adjust the time-series LAI based on HJ-CCD image with high temporal resolution, and LAI series covering the whole winter wheat growth period and with high spatial and temporal resolutions were generated. The effects of pixel purity and the number of high spatial image on the performance of fusing method were analyzed by comparing the LAI with fusing method and LAI from SPOT-5image or observed LAI. The results showed that the estimated LAI with fusing method has high consistency with observed LAI and the pixel purity is main obstacle factor. The fusion results based on two scenes of SPOT-5images are better than that based on one image. Meanwhile, leaf dry weight, leaf nitrogen concentration and leaf nitrogen accumulation series with high spatial and temporal resolutions were generated based on the data fusion.Quantitative relations between multi-temporal remote sensing data in main growth period of wheat and yield and protein content were analyzed based on the time series data fused by fusing high spatial and temporal resolution remote sensing data. On this basis, the optimum period predicting yield and protein content was selected and predicting models were constructed. The results showed the optimum period predicting wheat yield was initial filling, secondly anthesis. Cumulative value of spectral parameters from jointing to initial filling was highly correlated with grain yield and the predicting accuracy was higher than mono temporal. Meanwhile, the results also showed the optimum period predicting wheat grain protein content was anthesis. Cumulative value of spectral parameters from jointing to anthesis could reflect dynamic growth information of wheat main growth stages, which predicting result was more stable than mono temporal spectral parameters.