Research On Decision-level Data Fusion Methods In Greenhouse Control

The green house environment is characterised by multi-parameter factors,non-linearity,timelag and strongcoupling.In order to accurately determine the environmental conditions in the greenhouse,multiple environmental parameters in the greenhouse need to be integrated,thus providing greenhouse managers with accurate decision support for greenhouse regulation.Therefore,the thesis addresses the decision problem of greenhouse information fusion and proposes a composite framework of decision-level fusion algorithm with improved D-S evidence theory to provide decision support for greenhouse regulation systems.The main work and innovation points are as follows.(1)To address the problem that the existing methods for generating BPA(Basic Probability Assignment functions)in D-S evidence theory do not take into account the uncertain information in the evidence,resulting in the generation of unreasonable BPA,a method for constructing BPA based on FSVM(Fuzzy Support Vector Machine)is proposed.The simulation results show that this method is more accurate and reliable than some classical methods.The method expresses the different degrees of influence of each sample on the category attributes through fuzzy affiliation,which provides a more reliable BPA for the next step of D-S evidence fusion.(2)To address the problem of conflicting evidence in D-S evidence theory,the thesis considers the interdependence between multiple environmental parameter factors of the greenhouse and proposes a method of evidence synthesis with modified evidence weight coefficients.To address the problem of conflicting evidence in D-S evidence theory,the thesis considers the interdependence between multiple environmental parameter factors of the greenhouse and proposes a method of evidence synthesis with modified evidence weight coefficients.The method combines the distance between evidence and correlation coefficients to jointly determine the weight coefficients,and suppresses the influence of conflicting evidence on the fusion results with the modified weight coefficients.This BPA generation method is combined with the modified weights method to construct a composite model based on improved D-S evidence theory.Simulation test results show that the method is effective in reducing conflicts between evidence,accentuating the confidence level of the correct fusion results,and has higher reliability and convergence.Finally,an experimental platform of greenhouse intelligent sensing system is built,and the proposed improved composite algorithm is applied to the experimental system to simulate the greenhouse environment for testing.The system senses and manages the data through LoRa networking,and the paper’s composite decision fusion method is used to make judgments on the state of the greenhouse environment and provide corresponding decision support for greenhouse managers.The experimental results show that the improved evidence-theoretic decision-level data fusion method proposed in the thesis reduces the uncertainty by a factor of 10 compared to the classical method,while outperforming other classical methods in terms of fusion accuracy and stability,and that the decision-level fusion method can be used to achieve accurate judgement of the greenhouse environment.