Recognition of building group patterns is of great significance for understanding and modeling the urban space. However, many current methods cannot fully utilize spatial information and have trouble efficiently dealing with topographic data with high complexity. The design of intelligent computational models that can act directly on topographic data to extract spatial features is critical. To this end, we propose a novel deep neural network based on graph convolutions to automatically identify building group patterns with arbitrary forms. The method first models buildings by a general graph, and then the neural network simultaneously learns the structural information as well as vertex attributes to classify building objects. We apply this method to real building data, and the experimental results show that the proposed method can effectively capture spatial information to make more accurate predictions than traditional methods.

The automatic extraction of valley lines (VLs) from digital elevation models (DEMs) has had a long history in the GIS and hydrology fields. The quality of the extracted results relies on the geometrical shape, spatial tessellation, and placement of the grids in the DEM structure. The traditional DEM structure consists of square grids with an eight‐neighborhood relationship, where there is an inconsistent distance measurement between orthogonal neighborhoods and diagonal neighborhoods. The directional difference results in the extracted VLs by the D8 algorithm not guaranteeing isotropy characteristics. Alternatively, hexagonal grids have been proved to be advantageous over square grids due to their consistent connectivity, isotropy of local neighborhoods, higher symmetry, increased compactness, and more. Considering the merits above, this study develops an approach to VL extraction from DEMs based on hexagonal grids. First, the pre‐process phase contains the depression filling, flow direction calculation, and flow accumulation calculation based on the six‐neighborhood relationship. Then, the flow arcs are connected, followed by estimating the flow direction. Finally, the connected paths are organized into a tree structure. To explore the effectiveness of hexagonal grids, comparative experiments are implemented against traditional DEMs with square grids using three sample regions. By analyzing the results between these two grid structures via visual and quantitative comparison, we conclude that the hexagonal grid structure has an outstanding ability in maintaining the location accuracy and bending characteristics of extracted valley networks. That is to say, the DEM‐derived VLs based on hexagonal grids have better spatial agreement with mapped river systems and lower shape diversion under the same resolution representation. Therefore, the DEMs with hexagonal grids can extract finer valley networks with the same data volume relative to traditional DEM.

Line simplification is an important component of map generalization. In recent years, algorithms for line simplification have been widely researched, and most of them are based on vector data. However, with the increasing development of computer vision, analysing and processing information from unstructured image data is both meaningful and challenging. Therefore, in this paper, we present a new line simplification approach based on image processing (BIP), which is specifically designed for raster data. First, the key corner points on a multi-scale image feature are detected and treated as candidate points. Then, to capture the essence of the shape within a given boundary using the fewest possible segments, the minimum-perimeter polygon (MPP) is calculated and the points of the MPP are defined as the approximate feature points. Finally, the points after simplification are selected from the candidate points by comparing the distances between the candidate points and the approximate feature points. An empirical example was used to test the applicability of the proposed method. The results showed that (1) when the key corner points are detected based on a multi-scale image feature, the local features of the line can be extracted and retained and the positional accuracy of the proposed method can be maintained well; and (2) by defining the visibility constraint of geographical features, this method is especially suitable for simplifying water areas as it is aligned with people’s visual habits.

One important classical research area in automated cartographic generalization is simplification. Over the past few decades, numerous scholars have proposed various methods for polygon and line simplification, most of which have focused on vector data. However, with the rapid development of computer vision technology, unstructured image analysis and processing has provided a plethora of information, as well as new challenges. Therefore, in this article, we propose a new method for simplifying polygonal and linear features: a superpixel segmentation (SUSS) method specially designed for image data. In this method, polygonal boundaries are first divided by a superpixel algorithm called simple linear iterative clustering. Then, three types of curves – convex, concave, and flat – are globally simplified by comparing and selecting superpixels. Finally, uneven local features are removed by Fourier descriptors. In addition, the proposed SUSS method is extended for linear features, and it maintains topological relationships. To demonstrate the effectiveness of this approach, we use contours and water area data to perform experiments. Compared with the classic Douglas–Peucker and Wang and Muller algorithms, the proposed method is able to properly simplify the curves of polygonal and linear features while maintaining their essential shapes, and it maintains a steady change in area for large-scale applications while effectively avoiding self-intersection issues. Compared with the typical smoothing and Raposo algorithms, the proposed SUSS method can simplify lines at different scales and guarantee effective smoothing while maintaining displacement.