The purpose of trajectory classification is to predict the class labels of unknown trajectories in terms of the trajectory characteristics. Trajectory classification has many real-world applications, for examples: suspicious vehicles identification, illegal fishing vessels detection, transportation mode detection, etc. Currently, most trajectory classification methods only take two movement parameters which are speed and acceleration into account, and only employ simple statistics such as the mean, median and maximum values, thus they can't fully explore the characteristics of trajectories, which leads to relatively low classification accuracy. In order to solve this problem, based on a thorough literature review on movement parameters and quantitative statistics, this paper proposes a trajectory classification method based on the movement characteristics of moving objects. For movement parameters of velocity, acceleration, sinuosity, direction and turning angle, this method employs statistics such as skewness, kurtosis, coefficient of variation and autocorrelation from time series analysis to construct discriminative global features. In addition, this method extracts local features from sub-trajectories after trajectory segmentation. For direction and turning angle, this method incorporates directional statistics to compute their features accurately. The experimental results of this method based on three real trajectory datasets including vessel, wild animal and hurricane datasets, indicate that the classification accuracies of this method are 100%, 80% and 71.43% respectively. The experiments verify the movement features constructed in this paper are discriminative and effective.
The efficiency for querying complex spatial information resources is an important indicator to evaluate the performance of current spatial databases. Traditional single node relation spatial data management is difficult to meet the demand of high-performance in querying large amounts of spatial data, especially for the complex join query on multi-table. In order to solve this problem, we design and implement a spatial database middleware prototype system. This system takes full advantages of the massive parallel processing (MPP) and shared-nothing architecture. In consideration of the characteristics of spatial data, we design the spatial data parallel import, multi-spatial-tables join strategy, spatial data query optimization and other algorithms and models. This paper firstly introduces the development status of parallel database systems in recent years, and then elaborates its MPP architecture and its organizational model, and the strategy of the join query on multi-spatial-table. Finally, we made some query experiments on massive spatial data and analyzed the results of these inquiries. The experimental results show that this system indicates a good performance (nearly linear speedup) in processing the complex query of massive spatial data. Compared with the tradition single node database, this system can fully improve the efficiency of complex querying for large spatial data, and it is a more efficient solution to solve the complex spatial data queries.
Recommendation system has become mature and been successfully applied in many fields since its emergence. Due to the popularization of different types of mobile terminals, spatial information is brought into the recommendation systems. However, the existing researches mainly focus on spatial locations and rarely consider spatial relations. Meanwhile, the existing recommendation algorithms usually consider only the user’s history behaviors but not the influence of future behaviors on current recommendations. According to the activity chain theory, future activities have an impact on the current behavior as well as the past activities did. If a user has two steps of information retrieval, and the second step is based on the result of the first step, he would choose an item from the result which is convenient for him to make the second choice, and thus he can get the best choices at both steps of retrieval. That is to say, the current selection would be affected by the next retrieval of information. In this paper, we model the spatial distribution of the travel targets by considering one’s future intentions as well as the past data, and propose a spatial cascaded model for personalized recommender system. The model is built for situations with a series of continuous choices in the spatial space based on the traditional recommendation algorithm and the influence of future activities. The influence of spatial relation is introduced into the traditional recommendation algorithm as a distance decay function. In order to prove the feasibility of spatial cascaded recommender system, a restaurant recommender system is developed based on the proposed model. Taking into account of user’s preference and distance, a cost-benefit index was proposed to evaluate the result. The result shows that when considering further activities and spatial relations in recommendation, the system can produce a more reasonable result.
The correlation of geographic space has been significantly changed with the emergence of computer networks. In networked geographic spaces, network traffics that related to specific events, phenomena, or messages may lead to a particular group spatial-temporal distribution pattern when they reach a certain level. Therefore, for specific or targeted audiences, this will give rise to an unexpected result. Due to the characteristics of networked geographic space, by combining computer networks with geography, this paper built a cellular automata simulation model for networked geographic spaces. Simulation results show that: (1) there are five types of distributions for the cellular views in networked geographic spaces; (2) cellular spatial aggregation patterns exist in those spaces. The results can help us better understand the impact of information flow, which is related to specific events, phenomena, or messages, on the group spatial-temporal patterns in networked geographic spaces. They also provide the basis for analyzing the group spatial-temporal distributions of specific events.
Due to the stratified feature according to the floors in the multi-storey building space, the indoor path analysis is different from the outdoor path analysis, and the impact of floors, such as space location information, should be taken into consideration when the indoor optimal path planning is being carried out. However, the traditional optimal path planning method, which is based on network topology model between nodes, does not incorporate the space concept, which cannot be well applied to indoor path analysis. Aiming at solving the problem of indoor optimal path planning, based on the hierarchical characteristics of multi-storey building space, a structured dynamic network analysis model is put forward and the hierarchical optimal path algorithm is realized in this paper. In the algorithm, the network of each floor and the connections between floors are all regarded as independent structure. Firstly, the set of stops is divided into several sub sets according to the floor-distribution. Then, the stop distribution of the first floor is recorded and the first floor is regarded as the starting floor of the hierarchical optimal path analysis, and the path analysis for two consecutive floors is carried out floor by floor through dynamically constructing the structure network model spanning across every two consecutive floors. After that, the optimal path traversing all stops in the multi-storey building space is obtained. Compared to the traditional optimal path algorithm, the experimental results show that the proposed algorithm can obtain a more reasonable result for path planning, and the time efficiency is significantly improved. In addition, the structured dynamic network analysis model is more flexible, which allows to define specific floor conversion rules based on different requirements. The algorithm can be applied to large public buildings in cities, so that the indoor path analysis can be connected with the outdoor path analysis, making a more comprehensive path analysis.
The identification and reconstruction of karst fracture network has always been a hot and difficult topic in the research of karst groundwater resources and geological environment protection. Based on the original surface fracture data obtained in Zhangfang, Fangshan district in southwest Beijing and the surface fracture records processed through Kriging interpolation method, a construction method of the flow path based on disc model is proposed, focusing on the identification module of flow passage in three dimensional fracture structure and the underground fracture data obtained by Monte Carlo method. The flow path along the fracture generalized by disc model is simulated using directed graph data structure, and is stored through the form of adjacency matrix. In order to meet the demand of fast intersection operation in large scale fracture data and to improve the operation efficiency of karst fracture simulation system, this study gives a three dimension R-tree index algorithm to reduce the time of traversing each sampling point. Finally, based on the sampling surface fracture data of Zhangfang, Fangshan district in southwest Beijing, with the assistance of remote sensing geological survey, fine geological survey of key karst areas, and sampling analysis, this pager commits to the studies of the network distribution model for three dimension fracture space. The algorithm for building the path of karst fractures has also been simulated by computer. It provides a visualized analysis method for the study of karst development mechanism and the numerical simulation of karst water fracture in Zhangfang area.
Based on the three phases of land use and land cover change spatial data sets for the late 1980s, 2000 and 2008 in the Loess Plateau, we calculated the direction and amplitude of land cover change, land cover condition index and land cover change index, and analyzed the temporal and spatial characteristics of land cover and macro-ecological conditions changes in the Loess Plateau since the late 1980s. The results showed that the average land cover condition index of the Loess Plateau was 24.07 in the last 20 years. Land cover condition of the Rocky Mountain Area was the best, the Valley plain area took the second place, and the Agricultural Irrigation Area had the worst land cover condition. From the late 1980 s to 2000, the major land cover change was the conversion from forest and grassland to farmland, with the ecological grade transformed from the high grade to the low grade. From 2000 to 2008, the main land cover change was the conversion from cultivated land to forest and grassland, and the transition from low coverage grassland to high coverage grassland, with the ecological grade transformed from the low grade to the high grade. In the last 20 years, the land cover condition index and land cover change index indicated that the macro-ecological condition had experienced a slip period (1980 s to 2000 period LCCI-1.08), and then a meliorated period (2000 to 2008 period LCCI 2.66). This transformation was firstly driven by the climate change and population growth, and later was superimposed by the effect of ecological engineering.
Based on the highly accurate classification results of 6 years extracted from the remote sensing images from 1980-2012, a detailed research of the long cycled and multi-period of the land use layout and structure evolution caused by urban expansion in the expansion areas of Nanjing was implemented. The expansion intensity equal fan analysis method and transfer matrix data mining method were applied in this paper. During the application of the transfer matrix, the meaning of land use net change, swap change and total change was analyzed, the calculation method was summarized, and the calculation model of land use dynamic degree was improved to better depict the dynamic change of land use. The results show that the direction of the urban expansion was extremely uneven among the 5 periods in the 32 years. The main direction of expansion had shifted from northeast to southeast and then to southwest, and the expansion intensity had expanded considerably after the year of 2000. In the constitution of the total changes, the proportion of the net change in the arable land was close to the proportion of the swap change, and the arable land area was under a net reduction. In the meantime, the transfer of its spatial position was also significant. The net change played a main role in the urban construction land, urban green land, mining land and bare land, while the scope of water and rural construction land was mainly represented by the swap change. After the year of 2000, the average annual total change in the arable land and urban construction land of each period is significantly higher than before, and reached its highest value in 2004-2008 while dropped afterwards in 2008-2012. The land use dynamic degree in each kind of land use is relatively high over the past 32 years, and the dynamic degree values of the urban green land, rural construction land and urban construction land were higher than 90%. Studies of the total land use dynamic degree reflect that the urban expansion in Nanjing over the past 32 years has experienced four main stages: the accelerated evolution, the gradually slow in changing, the rapid evolution, and the integration and reconstruction.
Based on the land cover data in 2000, 2005 and 2010, we used the Shanxi Province administrative boundary and the digital geomorphic database of China at a scale of 1:1 000 000, the methods of the dynamic degree analysis, the transformation probability matrix and the frequency distribution of area in different land cover types to analyze the land cover dynamic change from 2000 to 2010 on basic geomorphic types in Shanxi Province. The results were indicated as follows: (1) The main types of land cover were arable land, woodland and grassland, which accounted for over 95% of the area in Shanxi Province. The woodland and construction land revealed an increasing tendency; however, the arable land and grassland indicatd the opposite tendency. The primary change tendency of land cover types was similar for the two periods of 2000-2005 and 2005-2010, which were resulting from the mutual transformation between woodland and grassland, arable land and construction land, as well as grassland and construction land. (2) The percentages of arable land and construction land gradually reduced with the increase of relief amplitude. Inversely, the percentage of woodland increased as the relief amplitude increased. In terms of the area change, the main changes of arable land, woodland and grassland occurred in the medium relief mountains; change of water area mainly occurred in high relief mountains; change of construction land area mainly occurred in plains and tablelands; and change of unused land area primarily occurred in small relief mountains. In the view of dynamic degree, arable land, grassland and water area exhibited the greatest changes in high relief mountains. And the most evident changes of woodland, construction land and unused land occurred in plains, hills and small relief mountains respectively. (3) The major transformations between land cover types were different with respect to basic geomorphic types. In the plains, the main occurrence was that the grassland evolved into woodland. The major transformations of tablelands were similar to that of the hills. Between 2000 and 2005, woodland was mainly degraded to grassland, while between 2005 and 2010, there was a mutual transition between grassland and woodland. In the small relief mountains, the dominating change trend was the mutual transition between grassland and woodland. However, the proportion of the grassland converted into woodland was much higher than that from the woodland into grassland. While in the medium relief and high relief mountains, the main transitions were that other land cover types transformed into woodland. It can be noted that the National Grain for Green Project and Afforestation Policy mainly occurred in the mountain area, while deforestation occurred in the gentle relief area.
Shoulder line is one of the most effective terrain structure line used to describe the loess landform. It plays an important role in the study of the spatial distribution and landform evolution of the loess landform. Shoulder line lies on the boundary of the positive and negative terrain, where the elevation and slope of the loess surface reveal obvious changes. Many efforts have been done for the extraction of shoulder line, but these methods have disadvantages, of which the experimental process requires human intervention or the shoulder line is discontinuous. P-N terrain method can effectively extract the shoulder line, but while using this method to segment the positive and negative terrains, it tends to produce large amounts of broken polygons and classification errors, which affect the accuracy of shoulder line. This paper investigates a region growing algorithm to improve the P-N terrain method. Using the highest elevation of the local area as a growing point for positive terrain and the outlet as a growing point for negative terrain, four-neighborhood growth were carried out until they reached the boundary of the positive and negative terrains or a slope threshold. Then the edge detection method was used to extract the critical boundary. Finally, the morphological image processing method was used to eliminate burrs to get the final shoulder line. In order to verify the result, this paper used the 0.6 m resolution remote sensing image to get the relatively accurate shoulder line by visual interpretation. And then different results were compared using overlay analysis. It is revealed that the shoulder line extracted using the improved method is closer to the visual interpretation results. This method is an automatic way to extract shoulder line, which solves the inaccurate location problem of P-N terrain method. Meanwhile, this method keeps the integrity and continuity of shoulder line and avoids the emergence of broken shoulder line and closure shoulder line. The use of morphological image processing method in burrs removal also ensures the accuracy.
This article estimated the precision of the precipitation simulated by 15 IPCC AR5 (the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, IPCC AR5) GCMs (Global Climate Models) and the multi-model ensemble (MME), based on the observed precipitation from 660 stations in China during 1996 to 2005. We firstly extracted the model simulation value at the corresponding position of the meteorological station, using the bilinear interpolation method, and took the average value of different models at the same station as the multi-model ensemble simulation value, then estimated the precision of the precipitation simulated by 15 IPCC AR5 GCMs and MME based on the observation of meteorological station. There were four evaluation parameters, including Corr (correlation coefficient), Bias, MRE (Mean Relative Error), and RMSE (Root Mean Square Error). Results show that the biases of the average daily precipitations simulated by IPCC AR5 GCMs present a gradually downward trend from northwest to southeast, and the RMSEs show a gradually increasing trend from northwest to southeast, while MREs in the east are less than those in the west. 82.3% of the average daily precipitations simulated by MRI-CGCM3 have relatively small biases, ranging from -0.5 to 0.5. The precisions of average daily precipitations simulated by BNU and MIROC-ESM are lower than that of others. Compared with other models, the MME simulation has the largest percentages of which the correlation coefficients are more than 0.5, MREs are less than 0.5, and RMSEs are less than 4mm, which accounted for 64.8%, 25.8% and 86.4% respectively. And the percentage of the biases ranging from -0.5 to 0.5 is relatively large, which is 56.7%, indicating that the simulation precision of MME is better than that of any other GCMs, and the MME can reduce the uncertainty of a single GCM simulation in future scenarios. Therefore, it is more scientific and reasonable to select the precipitation simulated by MME as the climate change condition, while studying subjects related to climate change.
Considering the limitations of DRASTIC model and the effect of uncertainties on the groundwater resource evaluation, combining with RS technology, a DRASTICL model based on fuzzy pattern recognition was established. The model was applied to assess the groundwater vulnerability in the lower reaches of Liaohe River Plain. The sub-watershed information was extracted by DEM using the hydrologic analysis tool of ArcGIS. According to the uncertainty characterization of the parameters, the stochastic and fuzzy parameters were simulated under different α-cuts of the triangular fuzzy parameters by Monte Carlo. According to the simulation under different α-cuts by the DRASTICL model based on the fuzzy pattern recognition and the cumulative distribution, the different groundwater vulnerable values under different α-cuts and percentiles were obtained. In order to analyze the groundwater uncertainty and vulnerability, the groundwater vulnerability distribution map under different α-cuts of the lower reaches of Liao River Plain was visualized by ArcGIS. Finally, the sensitivity analysis was used to identify the actual contribution of each parameter making to the simulation results. The results show that: (1) the fuzzy pattern recognition model generates a continuous vulnerability index and describes the groundwater vulnerability of contamination transit continuously from the easiest to the most difficult by the nonlinear form. (2) Adding the parameter of land use type could better reflect the groundwater vulnerability degree, which is higher in the paddy field area than in the dry land. (3) This study deals with the uncertainty issues of parameters effectively from three categories: different alpha levels, different percentiles, and different sensitivity coefficients. This article reflects the vulnerability degree of groundwater in different regions and under different possibilities and combines the subjectivity of decision makers with the objectivity of the actual hydrogeological condition for the research region, which has great significance to local groundwater development and protection.
With the progress of sustained urbanization in the past ten years, information of accurate urban vegetation cover is turning to be essential for the study of both regional climate and urban energy balance. High spatial-resolution remote sensing imagery provides an important tool for automatic mapping and monitoring of urban vegetation cover due to its broad coverage and high-spatial resolution. We propose an automatic urban vegetation extraction methodology, named as hyperplanes for plant extraction methodology (HPEM), based on the vegetation spectral feature analysis with the ZY-3 multi-spectral imagery over different cities in Yangtze River Delta. The results showed that: first, the vegetation pixels and non-vegetation pixels with low NDVI value can be well separated in the false color composite reflectance space, while the vegetation pixels and non-vegetation pixels with high NDVI value can be well separated in the true color composite reflectance space; second, HPEM could effectively suppress the errors of commission that come from built-up pixels which was often misclassified in NDVI method. HPEM’s performance was better than NDVI at the optimal threshold, with kappa coefficients increased from 0.85 to 0.90 and the total errors of omission and commission reduced from 21.15% to 14.18%. Compared to NDVI method, HPEM also avoided the tedious trial-and-error procedures for searching the optimal threshold. Therefore, HPEM can effectively improve the accuracy of automatic urban vegetation mapping. Moreover, the urban vegetation products are more reliable for further urban environment research.
Urban green space, as the main element of urban structure, plays a very important role in urban studies. In recent years, remote sensing technology has been applied in various industries with its rapid development. It has become one of the major techniques for city information (especially, green space) extraction with the emergence of high-resolution remote sensing images. Compared to the low-resolution images, the high-resolution images have many advantages such as clear texture detail and rich information. However, the existence of shadows in the urban district has a great impact on image classification, interpretation, mapping and so on. Therefore, the building shadows in the images have become one of the important limiting factors for green space extraction. In order to effectively extract city (green space) information from remote sensing images, it is necessary to extract and eliminate shadows from remote sensing images. This study is based on the QuickBird image of Hohhot city. Firstly, by utilizing the methods of band ratio process and band recombination, the image shadow information was enhanced; and the optimal band combinations for shadow extraction were derived using optimum index factor (OIF). Secondly, the image mask was established and the shadow information was extracted based on shadow threshold values (the minimum values and maximum values) of near-infrared band. At last, the shadow was removed by combining color space transformation with homomorphic filter and gamma correction; and then the effects were compared with other methods. The results show that, for shadow extraction, the optimal band combination is a combination of band 3/4, 4 and 2; and the best DN range is between 70 and 165. The method combined with gamma correction and color space transformation can effectively eliminate shadows and retain color information of QuickBird image. It is the best solution for shadow elimination in this study. The results can provide theoretical basis and technical support for efficient city (especially, green space) information extraction based on high-resolution remote sensing images.
A novel sparse representation classification model with spectral-spatial sparsity properties is presented to improve the classification accuracy of hyperspectral images. Firstly, this method uses the wavelet dictionary as the core dictionary to extract spectral domain sparse information, and then the spectral dimension sparse representation classification is transformed into the wavelet domain (WSRC) by inverse wavelet transformation. After that, we actually extract the sparse spectral features of the hyperspectral images and increase the recognition of the original dictionary. Secondly, considering the unity and diversity of the spatial adjacent object, we realize the sparse coding of the neighborhood pixels, and then accumulate the sparse codes. At the same time, we classify the hyperspectral images using a linear classifier that is based on the accumulated sparse codes. This method ensures that we extract the main sparse signal of the neighborhood pixels on the basis of the personality features of sparse encoding, and it performs better than the joint sparse representation model (JSRC) which is directly based on the neighborhood pixels. Finally, two commonly used hyperspectral images are utilized to validate the proposed model. The experimental results demonstrate that the proposed algorithm outperforms other models in terms of overall accuracy and kappa coefficient measurements.
Ground Penetrating Radar(GPR) has been more and more widely used in archaeological investigations, because it can be a non-destructive, cost-effective way to locate buried structures in archaeological studies. Compared with the conventional geophysical tools used in the shallow explorations, the electromagnetic method, ground-penetrating radar (GPR), is more economical and is capable to produce large amounts of continuous, high resolution subsurface data. GPR canextend the exploration range of remote sensing (RS) to subsurface. However, because of the non-uniqueness of inversion, an anomaly could be raised by the archaeological interest or the inhomogeneity of underground matrixes, therefore studying the typical anomalies of diferent archaeological targets on GPR images is helpful to distinguish the “true” anamolies from the “fake” anamolies. Furthemore, some experiences and references could be provided. The following experiments have been carried out: firstly, in order to analysize how the small targets of different materials and rammed earth will raise anomalies on the GPR maps, GPR was emploied to detect five pre-buried targets that are equivalent to the archaeological interest and a beacon tower in a integrated experiment station of remote sensing. The first experiment of GPR prospection was designed to simulate the buried-enviorment of the archaeological structure in the northwest region of China whose climate is predominantly arid. Secondly, the authors applied GPR in detecting the residual city walls of Xuanquanzhi ruins, then analysized the response features of the walls, and found that the detecting results well fitted the excavation. The engineering practice indicates that the ground penetrating radar technology is successful and effective in invetigating the archaeological remains which are of small scale, buried shallowly and very analogical with the matrixes in electromagnetic nature. The response models of different archaeological targets, which are respectively considered as the point, line and surface shape, have been proposed and explained according to the principle of rectilinear propagation of electromagnetic wave.
