Performance Enhancement of Spatial Big Data Map Matching by Using Linear Transformation of Coordinates and Distinctive Spatial Index Structures

• Main Authors: Mu, Ching-Yun, 穆青雲
• Other Authors: 周天穎
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/37237774224069186183

博士 === 逢甲大學 === 土木及水利工程博士學位學程 === 103 === Because Google and other international software vendors have launched free map services as well as mobile devices with positioning capabilities are increasingly popular in last decade, applications of spatial information technology have been rapidly commercialized and popularized in recent years. The public can easily purchase software or professional services of spatial information technology to solve their problems at work or in daily life. In response to this trend, governments and private sectors around the globe have started to establish the infrastructure of spatial information, expecting to provide complete spatial information contents and meet the public demand for spatial information applications. In the domain of spatial information technology, map matching is one of the basic functions. Map matching basically solves a major problem: mapping a set of spatial coordinates to a particular object or feature on the map, such as an address (point), a road (polyline), or an administrative district (polygon). With this mapping, we can then perform spatial analysis and application afterwards. Employing computers to perform map matching seems to be a simple task, but in fact it consists of many complex mathematical procedures. Puncture Method or Angle Method that map matching used in the past are based on Vector Data to process the above-mentioned mapping between coordinates and map features. When applying these methods to process large amounts of data, we would often encounter a bottleneck of performance. For this reason, this study is aimed at mapping the spatial coordinates to polygon features of map by first using linear equations to rapidly transform latitude and longitude coordinates into planar projected coordinates, followed by adopting raster as spatial data structure as well as data compression technology to effectively improve the efficiency of map matching. The Puncture Method with spatial index is also added for benchmarking. Experiment results show that, with a 2-minute interval, the linear transformation formula proposed by this study can successfully transform a WGS84 coordinates into TWD 97, a commonly-used projected coordination in Taiwan. The transformation errors are within one meter longitudinally and latitudinally; meanwhile, the linear transformation can also enhance traditional coordinate transformation by about 37% in terms of processing time. The Puncture Method with the use of spatial index improves the efficiency by 46.4 times compared with the one without spatial index. This indicates that spatial index can effectively enhance the efficiency of map matching. Map matching with Raster as the data structure, data compression technology, and table look-up method even improves the average processing time by 7.89 times compared with the above Puncture Method with spatial index as well as by 365.9 times compared with the traditional Puncture Method without spatial index. Although the spatial index can improve efficiency of map matching, the efficiency decreases as the complexity of road network. In terms of processing time, Puncture Method with spatial index is not as good as Map matching with Raster as the data structure, data compression technology, and table look-up method. In addition, the number of coordinates to be matched will not greatly affect efficiency of map matching by raster method. Although both Puncture Method with spatial index and raster method can obtain higher efficiency, their drawbacks include the huge data capacity and a long lead time. Whenever the road network changes, the former method should update the spatial index while the latter method should rasterify the road, re-coding and compress the data before process the map matching. In other words, these two methods make the use of preparing spatial index or raster data beforehand to earn the efficiency of map matching afterwards. Overall speaking, it is very worthwhile.