A Method Of Knowledge Acquisition And Fusion Based On Uncertainty Model In Soil Mapping

Obtain the knowledge of soil environment relationships is the core issue of fine soil mapping, how to extract the knowledge quickly and accurately becomes the focus of the study at this stage. In past studies, there still exist some space to improve the information coverage, precision and accuracy of expression in soil environmental knowledge. The method of soil samples-based knowledge acquisition will be influenced by the number of samples and the sampling process; and soil maps based knowledge acquisition, owning to the "double clear" drawing process that result in some loss of information; the resources knowledge of soil investigative are often not very complete and has low accurate description of knowledge. There will be helpful if we could complement and integrate knowledge with each other, to make up the defect of these three types of knowledge, and form knowledge with high accuracy. Shi et al explored soil- Integration of environmental knowledge, and combined the Global knowledge and Local knowledge in soil mapping, cartographic accuracy can be improved significantly. In view of this, if we can complete the knowledge of independent testing, analysis of its coverage, the expression of the soil foundation precision and accuracy, through statistical analysis, consistency and achieve complementarity and integration of multi-source knowledge, formed with high accuracy-environmental knowledge will significantly improve the accuracy of digital soil mapping, and thus have important scientific value and theoretical significance.This article used Huajia River Town,Huanggang, Hubei Province as study area, using So LIM software to get both the exaggerated and ignored uncertainty distribution map, based on the uncertainty distribution map of high reliability position to resample and sample data mining, access a combination of environmental factors, and then build the correspondence between these combinations and soil types, combined with the original rule, the integration of knowledge, acquire new soil- environment relationship knowledge that is optimized while using the new Drawing inference knowledge, in order to gain new distribution of soil types, and with field samples to verify the accuracy of soil map.This study use the decision tree algorithm and uncertain model as the main method to merge and extract the knowledge of the soil environment relationship. In practice, there are two uncertainties when we assigned one geography entity to one corresponding type: the first entity types with similar sex-related, ignoring uncertainty; second is with the entity inherent type is assigned to the relevant type of deviation, that is exaggerated uncertainty. Used herein So LIM software get both the exaggerated and ignored uncertainty distribution, according to the characteristics and distribution of uncertainty to analyze the accuracy of the study area mapping results qualitatively. By defining the threshold value, decision tree model can be used to predict soil type quickly and efficiently, but with the reduced level of prediction, accuracy rate is gradually reduced. In this study, we can use ignored uncertainty and exaggerated uncertainty to determine the accuracy of our inference graph to some extent and achieve a qualitative assessment of the accuracy of inference graph. Therefore, by coupling the two models, we can not only saves cost and time in the process, but also improve efficiency, achieve scientific retrieve soil and integration of knowledge.Verification the accuracy of inference graph digital soil mapping is an indispensable step, through the verification accuracy, we can effectively assess our methods and processes, and improve methods and processes, in addition it is also the final evaluation of the outcome of the reasoning mapping. This study will use the confusion matrix, the overall accuracy and Kappa coefficient of multiple indexes such as the original reasoning graph to verify the accuracy of reasoning graph after optimization.Application results shows: reasoning figure can show more detailed information in space and attribute when compared with the original soil map. At the same time, the overall accuracy is 86.9% when 253 field samples were used to validate, higher than the existing soil map accuracy of about 13%, more than 10%, and Kappa coefficient value is 0.842, higher than 0.8, indicating degree of consistency is very obvious, and soil properties and spatial distribution of the study area agreement. Therefore, the method used in this paper is to obtain an efficient soil- environment nexus of knowledge, the method of detail these two aspects have improved significantly in the properties of soil maps and spatial accuracy, innovative...