| The 12th Five-Year program has proposed "Increasing funding for remote sensing to improve the application and development of remote sensing technology; making big breakthrough in key technology; promoting the application of remote sensing technology to higher level; bringing all the expertise together to realize full data mining and effectively extract information from the huge amount of data collected in the past". Remote sensing application in China is stepping into a golden period. Agriculture is one of the most important and wide application areas of remote sensing. In the past decades, remote sensing technology, which is carried on ground, airborne, or satellite platforms, has been widely applied into modern agricultural information management and become an effective tool for undamaged and fast identifying growing plant information. However, any remote sensing platform, sensor, or spectral waveband has its own application scope and limitation. In some circumstances it is hard for us to fully characterize and assess crop growing status from a single data source. At the meantime, the disadvantages of multi-source data are redundancy, complementary and cooperation. Data fusion technology has been demonstrated as an effective way to cooperate these multi-source data. The complementary processing of different data sources in the same area provides the most effective application of remote sensing data. Therefore, it can help reduce the uncertainty, incomplete and errors generated by single remote sensing data source, and maximize the utilization of the multi-source data to make decisions. Using this method, we not only expanded the range of application of remote sensing, but also improved the accuracy of data analysis and practical values.Based on the project "the research based on multi-source data fusion for crop monitoring and control system" proposed by the United States Department of Agriculture (USDA) Southern Plains Research Center, this dissertation makes a deeper research on the multi-source data fusion for estimating crop nitrogen status and the development of a PWM precision spraying controller for UAV. All the related researches have been finished at USDA-ARS in College Station, Texas, USA. The main content of the dissertation involve:(1) Crop nitrogen status was assessed with the nondestructive hyperspectral measurements acquired with a modern scientific spectroradiometer from crop canopy. Both linear and nonlinear mathematical models were built on the data from the single data source to characterize the spectral properties of three types of crops. The component analysis linear regression model (PCR), the partial least-squares regression model (PLS), the feedforward neural network model (PCA-BP) and the RBF model (PCA-RBF) were established. The correlation coefficient (R), root-mean-square error correction (RMSEC), and root-mean-square error prediction (RMSEP) were used to verify the accuracy of the models. The results showed that the more various the properties of the samples were, the more nonlinear the relationship between the spectral reflectance and the components were. Thus, the PCA-RBF model gained the highest accuracy of prediction.(2) The relationship between spectral reflectance of crop canopy and canopy leaf SPAD readings were studied. The models were developed to test the linear relationship between SPAD readings and the different nitrogen levels. The results showed that crop canopy leaf SPAD values and crop nitrogen status highly correlated with a strong linear relationship. With the analysis of correlation and regression coefficients between spectral reflectance and SPAD values of crop canopy, the feature wavebands which contain significant information were found. The linear and nonlinear models for SPAD value prediction were built with the full range of wavebands and the selected feature wavebands. The results indicated that all the models, no matter with full wavebands or feature wavebands, have the high-coefficient with nitrogen status and low prediction standard error. In other words, crop canopy nitrogen status could be detected by spectral reflectance measurements and the feature wavebands were extracted to be related to SPAD values.(3) The optimized method for selecting feature wavebands based on Uniform Desigen Particle Swarm Optimization (UPSO) algorithm was investigated, and the design method of uniform test table based on the theory of uniform experiment design method was studied. The running parameters of UPSO was given and the mathematical model based on PLS algorithm using ( ) as an object function was established. The results showed that not only could UPSO-PLS model ensure the selection of feature wavebands, but also the mathematical model was optimized by UPSO with high accuracy of prediction and low prediction standard error. (4) The modern remote sensing platform was applied to experimental research. Nondestructive multispectral remote sensing measurements were taken to assess crop canopy nitrogen status, and the mathematical model was developed. With the analysis of regression coefficients and the PLS model between spectral reflectance and nitrogen status of crop canopy, the feature wavebands were found. With algorithms of data fusion, the linear and nonlinear multi-source data fusion models were established for predicting crop nitrogen status. The data fusion model optimized by UPSO was obviously improved with high-coefficient of prediction and low prediction standard error. These studies will help solve the problem of the complicated spectrum analysis, high cost, long testing time, low accuracy, and so on.(5) A precision spraying control system based on Pulse Width Modulation for the UAV was developed. The design criterion for the UAV spraying system was analyzed, and the precision spraying system based on the PWM technology was developed. Based on the experiments conducted in laboratory, system calibration parameters were set up and a decision system based on the multi-source remote sensing data fusion model for the spraying system was established. According to the prediction results of crop nitrogen status, the spraying gradient information of the field was produced and then experiment was conducted to test the precision spraying system. The result demonstrated that the precision spraying system could be controlled by PMW technology accurately and reliably. The nozzle sprayed uniformly along the flying direction of UAV without missing any spray or spraying too much.
|
PDF Full Text Request