Research On The Homonymous Road Polygon Elements Based On Fourier Shape Similarity Models

There is a certain limitation on the representation of objects by single data,while there are some complementary features to represent objects by multi-source data.It is more accurate that integrating multi-source data to represent objects.The homonymous elements matching is the key technology for registration which is the premise of integrating multi-source.There are many problems in current homonymous elements matching model of multi spatial data such as requirement the scale closely,and single feature similarity index with strongly one-sidedness while fuzzy distribution of weight and complexity of computation by multiple indexes.Accordingly,the applicability of Fourier shape similarity models respectively based on position function,polygon approximation and turning function,are introduced into homonymous elements matching of road network.And homonymous polygon elements matching of road network have been realized.The following works have been carried out:(1)Extraction of road network polygon elements based on multi-source remote sensing data.The road network data are extracted from the high resolution remote sensing image and airborne LiDAR point cloud data of the same city to provide data support for the homonymous polygon elements matching experiment based on Fourier shape similarity models.(2)The Fourier shape similarity model based on position function.Using the coefficient of the Fourier series expansion of the complex coordinates of the contour points at equal intervals to construct the Fourier shape description model;utilizing the Min-Max normalization method to address the description model and constructing the shape similarity model based on Euclidean distance,and then the characteristics change of translation,rotation,scale and starting point of the model are analyzed.(3)The Fourier shape similarity model based on polygon approximation.The approximate polygon of the contour curve is constructed and the Fourier shape description model is constructed by using the coefficient of the Fourier series expansion of the arc length function of the approximate polygon;next,the normalized shape description model and shape similarity model are constructed,and then the characteristics change of translation,rotation,scale and starting point of the model are analyzed.(4)The Fourier shape similarity model based on turning angle function.The Douglas-Peucker algorithm is used to thin the points on the contour curves in order to extract the salient feature points;the Fourier description model and shape similarity model are constructed by the coefficients of Fourier series expansion of the salient feature points' turning function;and then the characteristics change of translation,rotation,scale and starting point of the model are analyzed.(5)The homonymous road network polygon elements matching has been achieved by programming.The characteristics of Fourier shape similarity models respectively based on position function,polygon approximation and turning function have been analyzed by experiments in translation,rotation and scale and the change of start point,and the applicability of the three models in matching the homonymous polygon elements have been summarized.It shows that the three models have the translation,rotation,scale invariance and start point independence;it is different from other two models that the first model requires equal sampling and uniform number of the points which makes it more suitable for matching the homonymous polygon elements with closely scale.The first two models have a certain one-sidedness but better anti-noise capability than the third;although affected by the noise extracted from objected and increased the shape difference between the objects by thinned,there is a better stability of the third model;when the similarity is greater than 0.9,it can be determined that the matching pair is the homonymous elements.Therefore,the third model is more suitable for homonymous polygon elements matching of registration in multi-data.The research results of this paper have positive reference value for further registration of multi-source data.