Page 380 - NGTU_paper_withoutVideo

New Geomatics Technologies and Applications

To solve these problems, the general solution is to summarize (generalize) the features and then generalize the map.
Model-based and cartography-based methods are the two broad categories of generalization methods [1,2]. The visualization of
spatial data is the subject of cartography-based approaches. Modeling base methods seek to reduce density and volume, as well
as simplify spatial details, depending on the level of detail needed. Summarization of topological relationships, feature geometry,
and semantic information is also included in modeling summarization [3]. For each of these purposes, various methods have
been developed, including the Douglas-Poker and Viswalingam methods. The Douglas-Poker and Viswalingam methods use

different ways to geometrically summarize features, but their main goal is to select and maintain a few mainline points. In other
words, the deleted points are completely ignored in these approaches and have no effect on the summarized final feature. It's
worth noting that the existence or absence of these points has no impact on the results of these algorithms. These points, however,
are part of the main feature and may be significant; ignoring them reduces the geometric similarity of the summarized feature to
the main feature [4]. In this field, a variety of approaches have been implemented. The first methods did not rely on mathematical

relationships or real neighbourhoods. They selected and removed nodes on their own. For example, the random points approach,
after dividing the feature into different parts with consecutive nodes, randomly selected a node from each category [5, 6] and
also the n'th points approach, selecting the n'th node from each category and finally the selected nodes formed as a simplified
geometry [7]. Another categories of algorithms, locally examined the relationship between two or three consecutive nodes in
terms of distance [8]. The Reumann-Witkam algorithm starts from the first point and forms an length by connecting each point
to the next point. Then, in the direction of the specified length, a band with the desired width is considered. By checking the
points sequentially, the point was kept before the first point outside the band and the rest were discarded. Processing started

again from the retained point and continued until the last point of the initial geometry of the feature [9]. The sleeve-fitting
algorithm is another example. This algorithm operates in the same way as Reumann-Witkam does. The distinction is that the
first and last points decide the band's extension at each phase, and for each point that falls beyond the band, the last point from
the previous step is chosen, while the rest, except the first point, is left out [10]. In the Reumann-Witkam algorithm, Opheim
specified the minimum and maximum bandwidth values, as well as the length extension of the minimum bandwidth [11]. Lang
also modified the bandwidth values based on the vertical distance of the line connecting the two points of the original feature

geometry, with the points between them [12]. Douglas and Poker developed a method for summarizing the geometry of features
in 1973, which has since become one of the most widely used algorithms. Hershberger and Snoeyink improved it 19 years later
[13]. This algorithm differs from similar methods in that its search area is global. The Douglas-Poker algorithm operates on the
basis of vertical distance and does not consider the area of the shape. Viswalingam and Whyatt proposed a method that
generalized based on effective area to solve this problem. According to this method, in each iteration, the effective area for each
node (the area of the triangle consisting of each node with its two neighbours) is calculated and the node with the least effective
area is removed. This process continues until it reaches the desired number of nodes [14, 15]. Yilang Shen et al. [16] first

transferred the line into raster space using image processing techniques. As a result, the pixels that the line went through had a
value of 1 while the others had a value of 0. They created polygons with the shortest possible length by connecting a number of
straight lines in the obtained raster channel (pixels with a value of one). In a novel approach, Tinghua Ai et al. [17] performed a
triangulation on the main geometry nodes first. Then, they were given a certain degree of importance based on the number of
triangular neighbourhoods obtained, and this led them to the important nodes. Türkay Gökgöz et al also considered a buffer as
an error band with the aim that the output of the algorithm should be in tolerance of the original geometry of the problem, by

2