Recent decades has witnessed the use of multi- version in multi- scale spatial databases. Multi- scale spatial databases developed using these method have disadvantages, such as storage and time issues, and economic problems. As multi- scale spatial databases are among basic components in Spatial Data Infrastructures and play a ruling role in digital cities construction, the abovementioned method cannot ensure economic beneficiaries. Today, in order to resolve its disadvantages, a generalization process is used. Due to the incorporation of individuals' favours in choosing and adjusting the parameters, rules, operators, and algorithms, different spatial databases are generated that the selection of the best alternative becomes a significant problem.
Considering the importance of generalization process in building multi- scale spatial databases and the ability of generating various outputs through different generalization algorithms and operators, the selection of best and most similar output is necessary. The main aim of this thesis, which is the innovation of this research, is to present an integrated approach in order to calculating spatial similarity degree of the output of generalization algorithms of linear topographic databases in multi- scale space using the combination of models and criteria indicating the spatial similarity degree individual linear objects and the matching methods for groups linear objects. In this paper, Road networks are used as representatives of linear topographic databases. First, in order to produce different outputs of road networks, the Douglas- Peucker generalization algorithm have been used. In this research, two road networks were used in the scale of 1:25000 and 1:50000, and the Douglas- Peucker generalization algorithm were implemented on multiple thresholds on the road network of 1:25000 scale. Then, the matching process has been conducted using five geometric criteria, namely tangent function, direction, median Hausdorff distance based on length, buffer common area, length on the road network of 1:50000 scale and each of the different outputs generated by the generalization algorithms. For each of the two matched road networks, the F-  score value, which represents the degree of accuracy of the matching process, was calculated and those with an F- score value of over 95% were chosen to calculate the spatial similarity degree.
After the matching and selection of road networks with an F- score value of over 95%, the spatial similarity degree between the selected road networks was calculated using four relations of distance, direction, topology and attribute. The distance relations include density, length, number of lines, straight- line distance, sinuosity, complexity, linear object area, curvilinearity, and tangent function. The direction relations include difference direction and average angularity. The topological relations include difference topology, number of points, and degrees of points and the attribute relation include significance value. In the end, the total spatial similarity degree was calculated for each pair of compare road networks. The results show that the Douglas- Peucker algorithm with 3 meters tolerance and spatial similarity degree 77.106 percent has the highest degree of spatial similarity among different outputs.
In future researches, in order to increase the accuracy of matching, it is suggested in addition to the geometric criteria used in this study, topological criteria and attribute criteria are also used in the matching process. It is also recommended that in addition to the criteria used in this study to calculate the spatial similarity degree between two road networks, topological criteria relate to the relationship between lines in two road networks, as well as descriptive criteria such as the width of roads and the degree of roads. Also, in addition to the Douglas- Peucker algorithm, other generalization algorithms are recommended for the production of road networks at different scales, in order to identify the most similar smaller scale road networks using the large scale road networks.