| With the widely use of road network data in the field of Smart Navigation and Location-based Services, the demand for the data is increasing. In the same time, the requirement of the data's currency and quality becomes higher. Road network data updating help a lot in tracking the current situation of the data. An essential part of the updating with the newest data is the discovery speed of changes in data which determines the accuracy and efficiency of updating. Data matching, a key part in discovering changes in road network data, helps a lot in the first step of the updating to build connections among road entities which form the basis of rapid updating.Data matching today is still weak in its accuracy and automation, leading to the low efficiency and accuracy of road network data updating. Aiming at solving those problems, a practical road network matching algorithm based on multi-characteristics and its application to updating are discussed in this paper.As similarity characteristics and matching model are key elements in data matching, four road network similarity characteristics estimation methods are researched in this paper. In order to improve the efficiency and accuracy of road network matching, this paper proposes two road network matching methods based on multiple Logistics regression model and SVM model respectively. Also, the application of those matching methods to the process of multi-scale road network data interaction update is discussed in this paper. The main research works and contributions of this dissertation can be summarized as the following:(1) Current research status is summarized, and some road network matching methods and rapid data update methods based on multi-characteristics are proposed in this paper.(2) The organization and storage mode, reasons for data variation of multi-scale spatial data and the causes of differences of multi-scale road network data are discussed. Also, the process of data matching and the evaluation measurement of that matching method are proposed, which lays the foundation of a new road network matching algorithm.(3) After the analysis of traditional methods of describing road network entity similarity and their disadvantages, four optimized similarity characteristics algorithm are proposed in this paper:Aiming at solving the difficulties of describing direction differences based on angle chain in selecting encryption points and comparing when the differences in length are large, a method of describing direction differences based on tangent angle, which uses Vernier principle to compare the direction differences among road network entities, is proposed to realize an effective measurement of the direction differences. Aiming at solving the difficulties of the middle area shape description in selecting dual points and realizing the method, a method based on describing shape difference through area accumulation, which uses definite area integral of closed regions of two line entities to describe the shape differences between the two. Aiming at solving the difficulties of shorter median Hausdorff distance in measuring effectively when a pedal exists in the extension of a long line, mixed media Hausdorff distance description, which is a mix of Euclidean Distance and Vertical Distance, is proposed. Median distance is selected in this description method through making trade-offs, calculating and ordering the directed synthesis distance from a shorter line to a longer line. This method works well in measuring the distances among line entities and has a little amount of calculation, low complexity and obvious effect. Also, a semantic similarity description method based on global weighted distance is proposed in this paper. This method is able to make a comparison between a character string and a number. Semantic differential judgment. The coverage of semantic differences analysis is wide in this method and semantic characteristics can help in assistance identification. Those four characteristics are compared in description ability with some existing similar characteristics, which provide the feasibility and validity of the method.(4) According to the low degree of automation in data matching caused by manual intervention in determining similarity characteristics weights and matching thresholds, a road network matching method based on multiple logistic regression model is proposed, which makes use of the advantages of multiple logistic regression model in quantitative statistical analysis. The multiple logistic regression model in this method is built in training samples with similarity characteristics. Also, this model is utilized to realize automatic prediction of matching probability, and then, get the categories of matching result. After an experimental analysis with road network data in different areas, this method is proved to be stable, adaptable and precise in classification when dealing with different data, and the high classification accuracy.(5) Due to the excellent classification performance of SVM training in high dimensional characteristics and small sample data, a road network matching method based on SVM model is proposed. This method selects and combines a variety of effective similarity characteristics to construct a multidimensional vector. A small sample of road network data is used in SVM training to get a practical model to category other data, which contributes to the efficiency and accuracy of this method. Meanwhile, a better characteristics combination is obtained through the comparison among a series of multiple characteristics combination. Also, through a lot of experimental comparisons, this method is proven to be able to improve the accuracy and efficiency of road network data matching.(6) The application of road network data matching in data updating is researched to propose a multi-scale road network linkage updating method in this paper. This method utilizes road network data to realize multi-scale change detection, discover the changing entities, update the corresponding small-scale data and update data at the same scale. At last, the multi-scale data linkage updating is realized. |
PDF Full Text Request