Zhanlong Chen

Building Extraction in Very High Resolution Remote Sensing Imagery Using Deep Learning and Guided Filters

Very high resolution (VHR) remote sensing imagery has been used for land cover classification, and it tends to a transition from land-use classification to pixel-level semantic segmentation. Inspired by the recent success of deep learning and the filter method in computer vision, this work provides a segmentation model, which designs an image segmentation neural network based on the deep residual networks and uses a guided filter to extract buildings in remote sensing imagery. Our method includes the following steps: first, the VHR remote sensing imagery is preprocessed and some hand-crafted features are calculated. Second, a designed deep network architecture is trained with the urban district remote sensing image to extract buildings at the pixel level. Third, a guided filter is employed to optimize the classification map produced by deep learning; at the same time, some salt-and-pepper noise is removed. Experimental results based on the Vaihingen and Potsdam datasets demonstrate that our method, which benefits from neural networks and guided filtering, achieves a higher overall accuracy when compared with other machine learning and deep learning methods. The method proposed shows outstanding performance in terms of the building extraction from diversified objects in the urban district.

A Geospatial Information Grid Framework for Geological Survey.

The use of digital information in geological fields is becoming very important. Thus, informatization in geological surveys should not stagnate as a result of the level of data accumulation. The integration and sharing of distributed, multi-source, heterogeneous geological information is an open problem in geological domains. Applications and services use geological spatial data with many features, including being cross-region and cross-domain and requiring real-time updating. As a result of these features, desktop and web-based geographic information systems (GISs) experience difficulties in meeting the demand for geological spatial information. To facilitate the real-time sharing of data and services in distributed environments, a GIS platform that is open, integrative, reconfigurable, reusable and elastic would represent an indispensable tool. The purpose of this paper is to develop a geological cloud-computing platform for integrating and sharing geological information based on a cloud architecture. Thus, the geological cloud-computing platform defines geological ontology semantics; designs a standard geological information framework and a standard resource integration model; builds a peer-to-peer node management mechanism; achieves the description, organization, discovery, computing and integration of the distributed resources; and provides the distributed spatial meta service, the spatial information catalog service, the multi-mode geological data service and the spatial data interoperation service. The geological survey information cloud-computing platform has been implemented, and based on the platform, some geological data services and geological processing services were developed. Furthermore, an iron mine resource forecast and an evaluation service is introduced in this paper.

Shape similarity measurement model for holed polygons based on position graphs and Fourier descriptors

ABSTRACT In geographic information retrieval and spatial data mining, similarity is used to resolve shape matching and clustering. Many approaches have been developed to calculate similarity between simple geometric shapes. However, complex spatial objects are common in spatial database systems, spatial query languages and Geographic Information Science (GIS) applications. With holed polygons, many similarity measurement approaches are restricted to address the relationships between holes or between the holes and the entire complex geometric shape. A successful method should remove the restrictions due to these complex relations and retain invariant during geometric translation (rotation, moving and scaling). To overcome these deficiencies, we utilize position graphs to describe the distribution of holes in complex geometric shapes by storing invariants, such as angles and distances. In addition, Fourier descriptors and the position graph-based method are used to measure the similarity between holed polygons. Experiments show that the proposed method takes into account the relationships in an entire complex geometric shape. It can effectively calculate the similarity of holed polygons, even if they contain different numbers of holes.

Quality assessment of building footprint data using a deep autoencoder network

ABSTRACT Volunteered geographic information (VGI), OpenStreetMap (OSM), has been used in many applications, especially when official spatial data are unavailable or outdated. However, the quality of VGI remains a valid concern. In this paper, we use the matched results between OSM building footprints and official data as the samples for training an autoencoder network, which encodes and reconstructs the sample populations according to unknown complex multivariate probability distributions. Then, the OSM data are assessed based on the theory that small probability samples contribute little to the autoencoder network and that they can be recognized by the higher reconstructed errors during training. In the method described here, the selected measures, including data completeness, positional accuracy, shape accuracy, semantic accuracy and orientation consistency between OSM and official data, are used as the inputs for a deep autoencoder network. Finally, building footprint data from Toronto, Canada, are evaluated, and experiments show that the proposed method can assess the OSM data comprehensively, objectively and accurately.

A Knowledge-Driven Geospatially Enabled Framework for Geological Big Data

Geologic survey procedures accumulate large volumes of structured and unstructured data. Fully exploiting the knowledge and information that are included in geological big data and improving the accessibility of large volumes of data are important endeavors. In this paper, which is based on the architecture of the geological survey information cloud-computing platform (GSICCP) and big-data-related technologies, we split geologic unstructured data into fragments and extract multi-dimensional features via geological domain ontology. These fragments are reorganized into a NoSQL (Not Only SQL) database, and then associations between the fragments are added. A specific class of geological questions was analyzed and transformed into workflow tasks according to the predefined rules and associations between fragments to identify spatial information and unstructured content. We establish a knowledge-driven geologic survey information smart-service platform (GSISSP) based on previous work, and we detail a study case for our research. The study case shows that all the content that has known relationships or semantic associations can be mined with the assistance of multiple ontologies, thereby improving the accuracy and comprehensiveness of geological information discovery.

Research on semantics of entity space similarity measure based on artificial neural networks

Semantics plays an important role on spatial scenes building and similarity contrast. Based on the description logic knowledge base (ontology) and multi-layer neural network, this paper simulates the procedure of human perception, measures the semantic similarity between spatial entities. In the Knowledge Base, spatial concepts are built by some description of space, time, and properties, most of these properties are representative, such as structure, shape and function and so on. This paper will describe the spatial entities semantics by function, part and attribute. Semantics description of similarity is calculated by each category. Then, introducing the artificial neural network algorithm during calculating the similarity, establishing the learning rules, optimizing the problem of weight value in similarity calculation process. This paper regard the waters as research object, train the artificial neural network by the calculated result and human subject, to mine knowledge, and verify the results. The result shows that this model can simulate cognition of human better, and calculate similarity of semantics easily and accurately.

Road Extraction from High-Resolution Remote Sensing Imagery Using Deep Learning

The road network plays an important role in the modern traffic system; as development occurs, the road structure changes frequently. Owing to the advancements in the field of high-resolution remote sensing, and the success of semantic segmentation success using deep learning in computer version, extracting the road network from high-resolution remote sensing imagery is becoming increasingly popular, and has become a new tool to update the geospatial database. Considering that the training dataset of the deep convolutional neural network will be clipped to a fixed size, which lead to the roads run through each sample, and that different kinds of road types have different widths, this work provides a segmentation model that was designed based on densely connected convolutional networks (DenseNet) and introduces the local and global attention units. The aim of this work is to propose a novel road extraction method that can efficiently extract the road network from remote sensing imagery with local and global information. A dataset from Google Earth was used to validate the method, and experiments showed that the proposed deep convolutional neural network can extract the road network accurately and effectively. This method also achieves a harmonic mean of precision and recall higher than other machine learning and deep learning methods.

A novel computation method for line-line topological relations based on conformal geometric algebra

Topological relations computation is an important component of spatial analysis and reasoning. In the framework of Euclidean space, topological relations computation is mainly based on computational geometry methods, and it is hard to unify the computation for spatial objects of different dims. But with the birth of conformal geometric algebra (CGA), on one hand, classical geometry can be unified to a simple homogeneous algebraic framework, and with the concise and general algebraic language of CGA for geometric modeling, the classical geometry objects can be simply unified represented by the novel CGA. On the other hand, with the CGA providing fast and robust algebraic processing method for the geometric calculation, it is very easy for classical geometrys computation. This paper aims to provide a novel computation method for line-line topological relations based on CGA. It can be divided into the following two steps. First, using CGA to represent the classical geometry of line. Second, using CGA to calculate line-line topological relations. The computing process shows that the CGA can simplify the computation of line-line topological relations, and the calculation is robust and effective.

Building damage assessment approach based on seismic spatial information grid

Earthquakes are a sudden natural disaster that seriously endangers the safety of peoples lives and property. Rapid and accurate assessment of earthquake disasters, especially the building damage assessment, is an effective way to reduce the risk of disasters. Compared with traditional methods of GIS-based building damage assessment, which takes the entire stricken region as a whole assessment area, lacks in considering the uneven space-time distribution of damage factors, the seismic spatial information grid (SSIG) method fully considers the spatial and temporal characteristics of seismic data and can give a more accurate and detailed assessment results. SSIG is an earthquake thematic spatial information grid, which has integrated numerous kinds of spatial information technology. In this paper, SSIG has been introduced into building damage assessments. It describes the model and method of building damage assessment based on SSIG. The proposed method was exemplified after the Yushu earthquake in Qinghai province of China. The case application shows that the SSIG-based method can get more accurate assessment results than single GIS-based method.

Polygon aggregation algorithm using the simple data model in multicore environment considering spatial cluster

In this paper, Aiming at the characteristic of simple feature class polygon aggregation in multi-core environment, the improved algorithm divides spatial data for the polygon in simple feature class using STR (Sort-Tile-Recursive) tree index, it reduces the number of reading disk during searching polygons in the spatial database. The algorithm is based on the feature of STR-tree index, first of all it traverses the STR tree using the middle order traversal and returns the results, then do cascaded mergence for the traversal results of the STR index tree. That is for the built STR-tree index, recursively traverse each node using the middle traversal method, start from the root node of every layer, traverse the current STR tree branch in the order of starting from the beginning to the left, root, right. When the traversal of this layer is finished, we store the traversal results into the father node in the form of a pointer. We will get the middle traversal table of the STR tree. And this method fully takes into account the spatial aggregation characteristic of distributed polygon. At last, cascaded mergence is taken for the STR index tree. As the logical structure of traversal results remains the STR tree essentially, and merger operation is from the traversal mergence from the nodes of the first layer in the STR tree. Namely, firstly merge the leaf nodes of STR, then merge the nodes in the upper layer of the leaf nodes, traverse like that until the root of STR tree. Finally the STR tree was parallelized using parallel programming model OPENMP, we fully use the CPU computing power of multi-core computing environments. For the access features of massive data, we have designed and implemented the algorithm of data organization and building approach. At the same time, we compare the realization in this paper with general traversal polygon aggregation algorithm. The experiments show that this realization has high efficiency for large amounts data during the polygons aggregation. Functions based on this algorithm developed for practical problems can solve the polygon aggregation efficiency of large-scale and complex polygon data layer.