Applications Of Shape Recognition In Map Generalization

As an important characteristic of spatial cognition and a cognition result of graphic structure, pattern and distributing condition, shape plays a significant role in map generalization. Structural recognition of geographical features is essential in map generalization, which includes the analysis in distributing pattern and shape characteristic. Therefore, map generalization overall strategy could be established on the basis of the acquired distributing rule of geographical features and phenomena.Secondly, whereas a feature should be kept shape similarity in multi-scale representation, shape fidelity takes priority in generalization algorithm. Thirdly, shape similarity is an important index in map generalization evaluation. As an aspect in the field of cognition, shape recognition carries subjectivity and uncertainty. Therefore, shape analysis and recognition in map generalization is challanging and practical as well. This dissertation focused on the following:(1) Theoretical bases supporting shape recognition are systematically summarized. From the start of the definition of shape, the dissertation introduces three models about pattern recognition——template matching model, prototype matching model and feature matching model in a systematical way and expatiates on human’s cognitive regularity to shape. From the angle of the science of computer and information, the method or model about shape recognition and characteristics of shape representation of features in cartography are introduced and summarized.(2) The shape characteristic value of area feature in GIS and the model of calculating shape similarity distance are interpreted in detail. The Fourier shape descriptor and shape number to calculating similarity distance of shape are placed great emphasis.The convergence speed and descriptive accuracy are analyzed when the Fourier descriptor applied in describing different kinds of shape. The method to obtain shape number with three invariances which is proper to building is devised and that the shape number with different scale exists influence on shape description is researched.(3) Under the elicitation of prototype matching model of pattern recognition in cognition psychology, the dissertation presents the method of simplification of building based on dynamic template matching. The theory and operational processing of the method are provided from the aspects of prototype template designing, dynamic template creating, template setting and detecting validity of simplification result. The method has been proved practical by data experiment.(4) shape characteristic unit——bend of line feature and its shape characteristic parameters are analysed. The partition, hierarchy of bend of line feature are explained. Simple shape characteristic parameters and composite shape characteristic parameters and their geometrical meaning are summarized.(5) On the basis of the research of two dimension attributes of shape of line feature, the line feature is devided into different classes and is partitioned into many segments which have same shape characteristics, and the segments are simplified through different simplification algorithms and simplification threshold.(6) The dissertation expatiates on the possible losses as a result of the simplification of line feature——position accuracy, topology consistency and shape similarity. The relations among the above-mentioned losses are also presented. The shape deformation of different line simplification model is analysed and the conclusion is drawed that Douglas-peucker simplification model is the most proper in position accuracy and fidelity of shape.