Multi-scale of geospatial data is the cornerstone of cartography, and plays a key role in supporting geographic element analysis and feature recognition. Multi-scale vector data can be generated by selecting, simplifying, aggregating, or other processing of geographic element vector data of a certain scale obtained from remote sensing images. However, a variety of comprehensive processing models and methods will also lead to various levels of information loss in multi-scale vector data. The global coastline is a geographic information element with a wide coverage area, complex curves, various island combinations, and complicated structures of land and water regions. The variation of coastline vector data attributes shows different properties at different scales. For the special coastline vector data, there are multiple influencing factors, and the relationships between them are ambiguous. Therefore, it is impossible to judge the attributes of the elements only based on the combinations of a single or a small number of characteristics of the node or line elements. Meanwhile, using a single mathematical model or algorithm for simplification, the drawing effect often has a large deviation from the actual situation, and it cannot meet the drawing needs of different regions and different scales. Thus, we used Geographic Information System (ArcGIS 10.6) technology to support the automatic comprehensive function of geospatial data mapping, integrated different embedded automatic algorithms and models, and combined human-machine collaboration to build a systematic scale-up method system to achieve different scales of coastline data. Based on fractal theory, the concept of line vector data complexity index was first proposed to characterize the coastline geographic elements and to compare the degree of declination of their information. With the m-scale coastline data interpreted by manual visual interpretation, the scale-up is used to generate coastline data on the scales of 30 m, 250 m, and 1 km, respectively. The information loss assessment was performed on the obtained 30 m, 250 m and 1 km coastline vector data, and the results showed that the mapping integration caused changes in the spatial attributes of land and water. There are significant differences in the fineness of geographic element information represented by different scales. Compared with the m-scale coastline data, the loss of the number of islands on the scales of 30 m, 250 m, and 1 km is 32.07%, 90.46%, and 98.61%, respectively, the information loss of the coastline length is 6.32%, 49.26%, and 75.47%, respectively, and the information granularity of the vector data of the coastline of South America is reduced by 1.97%, 25.33%, and 45.39%, respectively. With the processes of the up-scale of the coastline, it has an increasing trend of the median, mean of the islands area and their complexity index from the m-level to 30 m, 250 m, and 1 km scales. The scale-up method constructed in this paper to combine the computer automatic synthesis model with the artificial processing of the coastline vector data has the potential to efficiently realize the scale-up of the coastline vector data, and describe the information loss of vector data at different spatial scales.