Research On Intelligent Early-warning System In Agricultural Based On Multi-attribute Information Fusion Decision

Early-warning for agricultural acts as the first link of agricultural risk management. Sofar the early warning in terms of agricultural risk has mainly adopts the leftover methodof ordinary risk pre-warning method, making a breakthrough from one single aspect in mostcases, however, failed to combine with the multi-factor features in agriculture system to buildmodels based on data fusion analyzing platform. Therefore, it is of great theoretical valueand practical significance for agriculture development to establish intelligent early warningplatform capable of effective multi-information fusion.The thesis, supported by the Open Project Program of Key Laboratory of Digital Agri-cultural Early-warning Technology (No. 2010-DAEW-01), aims to do research on differ-ent uncertainty puzzles in agricultural system, establish a multi-property information fusiondecision-making model and then develops an agriculture early warning system. The early-warning model is based on the HHM method, Analytic Hierarchy Process (AHP) methodcombined with D-S evidence theory. The research contents and creative achievements arelisted as follows:(1) In accordance with the complexity of agriculture system, we set up a indicatorsystem in view of agricultural risk early warning based on Hierarchy Holographic Mod-eling(HHM).(2) According to the distinct effect each factor has on the agricultural system, this articleadopts the Analytic Hierarchy Process (AHP) method to distribute importance to indicators,as also can be said weight confirming.(3) To avoid the subjectivity and the difficulty in processing the quantitative index inthe fuzzy risk analysis method, we use D-S theory to process the multi-property data. Thisessay puts forward a method of converting property data into basic probability assignment(BPA), rendering it the right form for D-S theory. The data conversion, in fact, is exceptingfor the effects of dimensions of quantitative and qualitative information.(4) In overall consideration of various factors of in?uence on agricultural system, thispaper introduces a collision information fusion of modified evidence source based on weightcoefficient, by comparing main advantages of D-S theory's combination rule, the method used in this paper is domestic and foreign. This method can well solve the problem ofcollision evidence fusion in a rather high convergence speed.(5) Establish a BPA function transformation model to evaluate agricultural risk.(6) In this paper, we develop the software of agriculture early warning system basedon IEDSS design notion, under Matlab2009b programming and GUI interface designingenvironment.