Decision-making reasoning in land-use patch generalization

The purpose of this research was to settle the complex issue of handling land use patches in land use data generalization and putting into use the principles and methods of the Agent-based technology through the analysis of the features of the land use patch itself. These patches include the shape, width, size, land use types, and the context of the land use patch which contains the space, semantic neighboring relationship, and the importance in the cartographic region. It also establishes a decision-making mechanism to judge the importance of the land use patch while taking into account a variety of criteria, and considers the geometry and semantic indicators integrated into the reasoning process which can be controlled and operated via an automated procedure from judging the importance of the land use patch to the selection of the scale transformation operators. The researchers conducted an experiment to generalize the land use data of a town from 1:10000 to 1:20000, making use of an agent-based software which was developed for land use data generalization. According to the decision-making mechanism and generalization process mentioned above, the results show that the amount of the land use type and the sum area in the cartographic region remain the same before and after generalization, and the area of the data in which the land use type is road or river is narrow polygon reduced considerably because of the use 'CollapseToLine' operator. Using the 'Eliminate' operator to generalize the data in which the land use type is pits cannot meet the threshold of the minimum area, resulting in loss of the area. In addition to the other area of each land use type data changed a little. The generalized data have no topology errors after using ArcGIS software topology to check, and keep the feature of full coverage with no overlap and no gap, and shortens the time from one week in the past to one day now which used to generalize this data, greatly improving the efficiency of generalization. Because the selection of operators was a computer process based on decision-making reasoning mechanisms, reducing the human participation, the generalization efficiency has been improved for the same data mapping area and equipment and the time was shortened from one week in the past to one day now. The experiments proved that the decision-making reasoning mechanism in land-use patch generalization in this paper is reasonable, and can be applied to actual production data.