ENGINEERING JOURNAL OF GEOSPATIAL INFORMATION TECHNOLOGY
Year:2015 | Volume:3 | Issue:2|Papers Count:6

Classification and detection of urban objects have always been a challenge for Photogrammetry and Remote Sensing researchers. Among various urban classes, plant complications due to their high diversity, spectral similarities in visible images and also absence of specific geometrical shapes have remarkable separation intricacies. In previous researches, for separation of the Trees class from "other vegetation" classes, different data sources have been used, which making use of visible images are the most low-cost and accessible data sources. Hence in this research study, an innovative method is developed for classification of Trees and "other vegetation" classes using visible image in urban areas. Therefore, firstly two new vegetation indices: Subdividing Vegetation Index (SVI) and Minus/ Subdividing Vegetation Index (MSVI), which are extracted from blue and green bands, are introduced. Then many textural features using Gray level Co-occurrence Matrix are estimated and then data reduction of Minimum Noise Fraction (MNF) is applied to the obtained textural features and first 5 bands had been selected. The obtained 2 new vegetation indices, 5 first bands of MNF and 3 bands of source images are entered to a Maximum Likelihood (MLL) classifier. The final result includes classification maps consist of Trees, "other vegetation" and "other urban objects" classes. The outcome of the newly proposed classification algorithm has shown the overall accuracy of 98.5 percent and Kappa coefficient values of about 93 percent. Furthermore, obtained results are shown the desirable effectiveness of the introduced vegetation indices in comparison to the other well-known vegetation indices for the classification overall accuracy, Kappa coefficient and average accuracy improvement. Also performance evaluation of the novel indices has shown an improvement about 4 percent in

Today, the high number of the rural road accidents has shown that accidents at intersections organize high percentage of the total number of accidents. Meanwhile, the geographical information system (GIS) is considered as an appropriate tool for doing spatial analysis and analysis of accidents at intersections. Also, considering that the accident data are massive and non-homogenized, the methods of spatial autocorrelation and kernel estimation can present connected and more real models than samples of current spots in rural road accidents. The purpose of this research is spatial analysis of rural road accidents based on rural intersection or utilizing spatial autocorrelation methods and estimation of kernel density. In the first stage, it considers appropriate criteria for spatial analysis of accidents in the old roads of Karaj-Qazvin in the 1388-1392 periods and were weighted by use of fuzzy analytic hierarchy process. Then, to identify accident prone intersections and investigation of their characteristics from autocorrelation functions of Getis-Ord Gi*, Anselin Local Moran's I and the kernel density estimation was used in order to investigate the spatial autocorrelation, each of the useable parameters in five successive years used as a Moran's I function. The results have shown that there are 26 accident-prone intersections towards Karaj-Qazvin path and 10 accident-prone intersections for returning, from the total intersections of two way path. Also, for path towards Karaj-Qazvin none of the parameters and for returning path the only parameter of accident type contained spatial autocorrelation in the five successive years.

In today’s world, rapid increase of urbanization and traffic challenges have led to the profound need of traffic control systems with the highest possible productivity and efficiency. Time loss and increased fuel consumption as well as air and noise pollutions have made traffic control to be one of the most important current issues in the world. One of the best possible methods for reaching this objective is to predict the directions and the final destination of the car. If the future position of a car can be predicted, traffic estimation in an urban zone will be a simple task. Route prediction is possible based on the previous routes of the car as well as parameters such as the starting point, time, day, month, and duration utilizing data mining methods and artificial neural networks. The current paper uses real GPS data obtained from different cars in order to carry out prediction operation of the route and final destination. One of the propoesd methods in this study is to establish a database for the previous routes of the cars using ArcGIS software The high accuracy of recording the previous routes of the cars in this database increases the accuracy of the route prediction process. In this study, two distinct databases were established. The first involves general database, where only the more challenging sections of the roads including intersections and crossroads are considered in the second which is the more complex database. Moreover, in order to carry out the prediction operation, association rules algorithms as well as artificial neural network algorithms have been used. The obtained results indicate the high accuracy of the prediction. Artificial Neutral Network (ANN) algorithms used on the general database and the GRI algorithm used on the more complex one provide better results, respectively. Both algorithms acquires precision greater than 95%. The results obtained from the prediction process can be used for traffic planning and the optimization of car movements.

Spherical radial basis functions (SRBFs) have been extensively used in regional gravity field modeling. Determining the optimal of the SRBF parameters including their shape and locations is one of the most challenging tasks in SRBF approximation of the Earth's gravity field. In this paper, an optimization strategy is suggested to solve the problem of gravity field modeling using SRBFs. For this purpose, the potential gravity anomaly is expanded into a linear combination of the SRBFs, and then, the system of observation equations is set based on gravity anomaly data. The unknown modeling parameters are consisted of two steps: 1- the 3D position of SRBFs, namely SRBF centers and SRBF depths are determined utilizing the genetic algorithm, and 2- the scaling coefficients in SRBF expansion of the gravity anomaly are determined using the Tikhonov regularization algorithm. In this approach, a chromosome population which includes the 3D position of the kernels is generated and those with more competence are chosen. Furthermore, new chromosomes are produced based on crossover, mutation and migration processes. Therefore, since the kernel positions are obtained via the genetic algorithm, the non-linear problem convert into a linear problem which the coefficients of the expansion for each chromosome can be solved using the Tikhonov regularization algorithm. The performance of the proposed optimization scheme is assessed based on synthetic gravity anomalies provided by EGM2008 up to degree and order of 2160. Finally, an accuracy of 1.08 mGal in gravity anomalies and 0.78 m2/s2 in anomaly potentials is obtained. The numerical experiments reveal that the proposed optimization algorithm provides an appropriate SRBF distribution which improves the gravimetric models' accuracies.

A fully polarimetric synthetic aperture radar (POLSAR) image can provide important polarimetric features for land cover classification. These features can be the parameters obtained from scatering, covariance and coherency matrices, parameters extracted from target decomposition methods or both of them. In this paper, many polarimetric features are extracted from a POLSAR image. Then, with the use of Genetic Algorithm (GA) and Decision Tree (DT), a feature selection method based on the classification is presented. Afterwards, a comparative analysis is accomplished between DT classification with features selected from the proposed method and DT classification with all features. Moreover, the proposed method should be compared with the feature selection method of GA and Support Vector Machine (SVM). The results indicated that the accuracy of the proposed method (DT classification with the features selected from GA-DT algorithm) is nearly 3% higher than the ones of the DT classification with all features and it is approximately equal with the ones of the DT classification with the features selected from GA-SVM algorithm. However, the performance speed of the proposed method is approximately 5 times more than the ones of DT classification with the features selected from GA-SVM algorithm. As an another result, the features selected from the proposed method have a more success than the ones of two other methods at classifying the urban areas and vegetation classes.

To reach a more accurate prediction of future of a city, modeling must be done for all land-uses of the town. Logistic regression only can model a bivariate urban growth, i.e., urban and non-urban. Also, this method cannot consider the neighborhood effects in the allocation process. Due to this issue, the aim of this paper is to provide a method for modeling multiple land-use changes and applying the neighborhood parameter in allocation process, and thereby increasing the accuracy of modeling. So, in this article, we predicted the land-use map of the year 2014, using the land-use maps of the years 2002 and 2008 by considering the effects of the neighborhood parameter and by comparing the predicted land-use map of the year 2014 with the reference map of 2014, the accuracy of the model was obtained. Reference land-use maps were obtained using classification of Landsat images of 2002, 2008, and 2014 using the Support Vector Machine (SVM) method. In the proposed method, the first modeling was performed separately using the Logistic Regression method for each land-use. Then the results of the Logistic Regression as a Competency Map for allocation process were combined with the Markov Chain and a combined method of MOLA-Neighborhood to obtain the land-use map of 2014. The procedure was performed in 4 different scenarios. In three of them, the neighborhood effects was considered as 3´3, 5´5, and 7´7 kernel and in the last one, modeling was performed without considering neighborhood effects. The accuracy of 4 scenarios was compered using the reference map of 2014. In the best scenario the accuracy of method was obtained using overall accuracy, kappa index and location about 84.26% and 76.35%, and 79.3%. Finally, the accuracy of each land-use category was evaluated separately using the ROC, which indicates the capability of the proposed approach of this paper. Finally, the land-use map of the year 2020 was predicted in two different scenarios.