Study On Crop Heavy Metal Pollution Stress Information Extraction From Remotely Sensed Data

The problem associated with farmland soil heavy metal pollution has been of interest to researchers for a long time. Heavy metal pollution has significant impacts on the growth of crops, and it may cause reduction of crop yield. The concentrations of pollutants will be increased as toxins are passed up the food chain and ultimately threatening the health of human beings. Estimation of crop heavy metal stress in large scale farmland is essential for managing and protecting agricultural environment. The goal of this research is to derive a new model for crop heavy metal pollution stress level estimation based on remote sensing data.This paper reviews the application of fuzzy theory, artificial neural network, and fuzzy-neural network for remote sensing information extraction. A dynamic fuzzy neural-network model is presented for crop heavy metal pollution stress level assessment based on remote sensing data. Hierarchy self-organized learning algorithm and pruning algorithm are applied for its training progress to optimize network structure. Hyperspectral vegetation indices, NDVI,MTVI2 and MCARI/OSAV, are used as input variables in this model for the purpose of enhancing weak information of crop heavy metal pollution stress.Both simulation experiment and crop heavy metal pollution stress level classification experiment were carried on to verify the performance of this network. By comparing its result of non-linear system simulation with FNS, RBF-AFS and OLS, it is confirmed that this model can achieve better performance with a less complex architecture.In the experiment of crop heavy metal pollution stress classification, 250 data samples, which contained values of hyperspectral vegetation indices and heavy metal stress levels, were prepared for the training process. At the end of the training process, this dynamic fuzzy neural-network model generated a total number of seven fuzzy rules. Another dataset, with 60 testing samples, was applied to evaluate the performance of this trained system. The result of this experiment indicated that this model was capable of extracting stress information with reasonable accuracy, which is over 95%, and thus it could be used as an effective tool in monitoring and managing agricultural environment. There are six chapters in this paper. Chapter one is an overview of previous research work on the usage of remote sensing technology in vegetation health monitoring area. Chapter two introduces the biochemical reaction mechanism of crop heavy metal pollution stress, and the principle of vegetation remote sensing. It also explains the selection standard of vegetation indices used in this model. Chapter three gives out a detailed introduction of fuzzy-neural network. Chapter four specifies the structure of the dynamic fuzzy-neural network model, along with its algorithms. Chapter five contains analysis of the results obtained from both simulation experiment and stress level classification experiment based on outfield-collected remote sensing data. And chapter six is the conclusion of this research work.