ENGINEERING JOURNAL OF GEOSPATIAL INFORMATION TECHNOLOGY
Year:2015 | Volume:3 | Issue:1|Papers Count:6

Seismotectonic provinces are defined as areas that have similar seismological characteristics, active tectonic fault system similar to each other. Identification of these areas is one of the most important subjects in seismic hazard analysis. Clustering methods are known as techniques for identifying information associated with the huge number of data. Since these methods are appropriate visual representations and provide reliable results, they can be an important tool for determining the seismotectonic provinces. In this study, three partitioning clustering methods were used including: K-means, CALARA, and Fuzzy C-means. For assessing the validity of these methods, five clustering validity indexes including: Davies-Bouldin, Silhouette, SD, S_Dbw, and Xie-Beni were used. Parameters including the earthquake epicenters, magnitude, gravity, magnetism, and fault density, as input clustering methods, were considered for modeling seismotectonic regions. After determining the optimal number of clusters, in order to evaluate the results of clustering model IRCOLD seismotectonic model was applied. The results showed that compared with other methods, CLARA have a greater compatibility with seismotectonic model in Iran. Finally, after comparing the number of optimized clustering methods with seismotectonic model it is concluded that the best number of seismotectonic provinces is 13.

Forest biomass and estimate its value has a significant role in climate change. Because of land constraints and time-consuming methods to estimate biomass, using remote sensing is an effective alternative to terrestrial methods. In this study, in order to improve the accuracy of estimates of forest biomass to earlier research, optical image AVNIR-2 and PALSAR radar satellite ALOS images used with data from ground-based College of Agriculture, Tehran University of North region Kheiroudkenar. This stude procedure respectively 1 - features extraction from images, 2 - select features using genetic algorithms, 3 - Biomass estimated with features selected by regression analysis and neural networks. Evaluating the results of the application of neural networks and regression analysis on the features selected by genetic algorithms, neural networks represent the accuracy over 70 percent and regression analysis represent the accuracy to about 15 percent. For this reason, the use of neural networks in a way that has been used in this study for the northern forests and the complex structures is recommended.

When the beam of a laser scanner because of an obstacle (such as trees or walls of a building) is not able to pass the complication, the points are on the backside, not measured. These points appear as occlusion parts of the point cloud that study of them is a key issue for the next station selection. Nowadays, occlusion area analysis study manually that this work is time-consuming and erroneous. In this paper new technique provide, which is used to determined occlusions of the laser scanner data, automatically. As we know, the edges in an image showing the drastic difference between adjacent pixels values. In the range image, the difference is due to the difference in depth in the presence of occlusions; its value will be high. Therefore, in this article to determine the occlusion borders automatically, made an image from the data collected by laser scanners and edges are extracted. After that, the borders defined on a reference plane and are connected to each other up to specified occlusion area. To implement this, the ground laser scanner station point cloud was used in an urban area. Canny technique was used to extract edges. The findings show that the proposed method can be used in the absence of complicating factors (such as trees) and occlusions of point cloud automatically can be identified.

Hyper spectral remote sensing technologies have many applications in land cover classification and study their changes. With recent developments and create images with high spatial resolution, it is necessary the use of both spatial and spectral information in hyper spectral image classification. In this paper, we have evaluated the effect of dimensionality reduction using genetic algorithm on spectral-spatial classification of hyper spectral imagery. So far, among the various algorithms spectral-spatial classification of hyper spectral images, three segmentation algorithms, watershed, hierarchical and Minimum Spanning Forest (MSF) based on markers, combined with Support Vector Machines (SVM) to achieve the best results. In the proposed approach, the dimension of hyper spectral images is first reduced by using genetic algorithm. Then, the three mentioned segmentation algorithms are applied on the resulting bands. Finally, the obtained segmentation maps are combined with SVM classification map using majority voting rule. The proposed approach was implemented on three hyper spectral data sets, the Pavia dataset, the Telops dataset, and the DC Mall dataset. The obtained experimental results indicate the superiority use of reduced bands in MSF based on markers algorithm and all bands in watershed and hierarchical based on markers algorithms.

Recognition and study of snow reservoirs as the supplier of the base flow of rivers and main outset of freshwater resources in snowy and high basins play an important role in planning and management of water resources usage. However, one of the main problems in snow phenomenon recognition using optical satellite images is to separate clouds and snow. To supper this problem, we use the fact that the cloud does not have a stable geolocation compared to snow. A temporal filter is designed by the combination of Modis Terra and Aqua to remove the cloud pixels. Moreover, different spectral behavior of the cloud in different wavelengths makes it possible to separate it from the snow. A normalized difference cloud index is defined using Modis data to detect and remove the cloud pixels from the image. The pixel-based method is used to extract the snow coverage map of the Northen area of the Fars province using the daily Modis data spanning between 1392 and 1393. In order to evaluate the final results, the data from 14 ground stations as well as Landsat8 OLI image are used as ground truth. The accuracy of 100% was achieved using the first method while the accuracy of the second method by corresponding the pixels of snow coverage maps is estimated as 98.58%. According to the results and accomplished evaluations, the snow maps generated using the threshold-based method without or with the cloud coverage removed by the application of the proposed method has a high precision. The results can then be easily used in the snowmelt run-off modeling in the water resource and reservoirs management.

The increasing availability of high resolution satellite images is an opportunity to detect urban objects such as roads. In order to increasing the precision a new image analysis using object-based approaches has been proposed. In this paper, designing steps of knowledge based of road detection has been presented. In this field, an important challenge is the use of knowledge for automatic road objects identification, and a major issue is the formalization and exploitation of this knowledge. At first, optimum features, including spectral, texture and structural features, are detected using a genetic algorithm with a k-nearest neighbor classifier. After that a rule based road detection strategy has been developed using prior knowledge and optimum features interpretation. The method is designed and validated by IKONOS images of the urban areas of Hobart, Kish and Shiraz. The validation results highlight the capacity of the proposed method to automatically identify road objects using the knowledge based proposed system.