Remote Estimation Of Yield In Oilseed Rape With Unmanned Aerial Vehicle Data

In recent years, with the rapid development of remote sensing technology, the information acquisition of agricultural remote sensing trend to Air-Ground Integration. The research and application for estimation of yield continue to make a breakthrough. Yield estimations prior to harvest play a key role in the determination of input factors and accurate agriculture management in agricultural production. In addition, bioenergy-and other corn-related industries benefit from these estimations, too. Oilseed rape is one of the major cash crops in the world mostly for its oil-rich seeds, but at the flowering stage and pod stage, the canopy structure of the crop was changed obviously. The bright-yellow flowers lasting more than 30 days, which is approximately 1/4 of its entire growing season. This phenomenon affects the accurate acquisition of remote sensing data and result in parameters estimation errors. The study is closely related to precision agriculture, the ground data and unmanned aerial vehicle (UAV) data was used to make a dynamic three-dimensional monitoring of oilseed rape. In the key growing stage of oilseed rape, real-time spectral information and physico-chemical parameters were obtained by ground measured data and UAV images. These data were used to keep abreast of oilseed rape field conditions and growth status. By analyzing the remote sensing data, we can developed an accurately estimation of physico-chemical parameters and final yield. Through the analysis of the contribution of different growth stages to the final yield, we can provide some quantitative decision-making reference for precision agriculture management. Based on the single-period estimation model, an integrated yield estimation model of multi-period for oilseed rape in central China was proposed. The main research work including:1. The influence factors of canopy spectrum and the effects of different widths on inversion of typical parameters (eg. leaf area index (LAI), chlorophyll content (Chl)) were analyzed. In different growing stage of oilseed rape, with the variation of typical parameters, canopy spectrum shows different changing tendency. In flowering stage, as the chlorophyll content and LAI increased, reflectance of canopy spectrum increase obviously which is quite different from leaf stage. The best band width to retrieval chlorophyll content and LAI is:green, red and near-infrared band should below 30nm and red edge should below 25nm, an appropriate increase in band width can improve the estimation accuracy. According to the bandwidth, we designed the filters for each channel of the MCA camera. Consequently, the parameters and the yield inversion methods based on mathematical combinations of spectral reflectance can be popularized to the UAV platform as well.2. On the ground remote sensing monitoring platform, it is better to estimation the chlorophyll content and LAI according to different growing stage. Based on the good linear or quadratic relationship between chlorophyll content, LAI and yield, the empirical model of vegetation index can be promoted to yield estimation. Based on the characteristics of hyperspectral remote sensing, continuous wavelet transform model and neural network model which can make use of more spectral information are established. Based on the data of ground remote sensing platform, the entropy method and the analytic hierarchy process (AHP) were used to analyze the contribution of different growing stage to the final yield and the combined forecasting models were developed. The key period of yield estimation is ten-leaf stage for the ground remote sensing monitoring platform. The best yield estimation method is the optimized CIrededge vegetation index model, which is based on different planting pattern. The determination coefficient (R2) of the validation is 0.96 and RMSE =169kg/ha.3. On the UAV remote sensing platform, this study developed an approach for remote estimation of Vegetation Fraction (VF) and Flower Fraction (FF) in oilseed rape. Based on reflectance in green and NIR bands, the histogram thresholding method was developed to identify whether a sample contained flowers and then to choose automatically the appropriate algorithm for its VF and FF estimation. The results showed that it was able to predict VF and FF accurately in oilseed rape with RMSE below 6%. The method of estimation typical parameters of ground platform is extended to UAV platform. The best vegetation index for estimation the chlorophyll content and LAI is NDVI and CIrededge in ten-leaf stage, Rgreen in flowering stage and EVI2 and MSAVI in pod stage. According to the characteristics of UAV images, two models based on vegetation coverage and mixture pixels analysis were established for yield estimation. Furthermore, the abundance information was added into the neural network yield estimation model. Based on the data of UAV remote sensing platform, the entropy method and the analytic hierarchy process (AHP) were used to analyze the contribution of different growing stage to the final yield and the combined forecasting models were developed. The key period of yield estimation is also ten-leaf stage for UAV platform. The best yield estimation method is the optimized NDVI vegetation index model, which is based on different planting pattern. Whether on the ground or unmanned aerial vehicle platform, the result of using the spectral information to yield estimation is poor. However, the yield estimation model can be effectively improved by combined with the information of the growth environment (flower coverage, flower abundance) in flowering stage.4. Although the single-period estimation model is simple, fast and flexible, it is lack of comprehensiveness, reality and generalization. Therefore, an integrated yield estimation model for oilseed rape was proposed. The stepwise regression method was used to determine the key variables for estimation of oilseed rape yield at different stages by combining different remote sensing data from different platforms. Based on the combination of entropy method and AHP method, the integrated evaluation model of all period and multi-platform for estimation of oilseed rape yield was developed. The determination coefficient (R2) of the validation is 0.91 and RMSE=225.2kg/ha. On this basis, the model was optimized, and an integrated yield estimation model of multi-period and multi-platform for oilseed rape in central China was proposed. The period of the model was pod stage, ten-leaf stage and eight-leaf stage. The determination coefficient (R2) of the validation is 0.94 and RMSE=190.8kg/ha with an MNB below 10%.