فصلنامه سنجش از دور و GIS ایران
سال سوم شماره 4 (پیاپی 12، زمستان 1390)

The proposed semi-automatic road extraction system has been established to store and update road networks in medium and small scaled maps of the RMTO of Iran and to reduce charges and also to keep or increase precision and speed of information extraction in comparison to field work and GPS usage. Furthermore, the capability of the proposed system in extracting different shaped roads such as straight, spiral, junction and square has been investigated. The input data of the proposed road extraction system are Pansharpened and Panchromatic images of IKONOS and IRS-1D (1&4 and 5 meters spatial resolution) and Pansharpened images of ETM+ sensor of Landsat7 (15 meters spatial resolution). In the proposed method, FCM clustering technique is implemented in spectral space. Afterwards, morphological operators, as post- processing tools, are used to improve the road class image by removing noise and filling gaps and then C-Means clustering is performed on the spatial coordinates of the pixels and Key Nodes are attained. These nodes were connected using MST algorithm in graph theory by defining appropriate cost function (Euclidean distances) in order to extract road centerline which can be imported into GIS environment in vector format. Attaining mean overall accuracy (OA) of more than 98% and mean RMS error of about 0.68 pixels, when comparing automatic extracted road centerline with the manual extracted one, is a good criterion for the success of the proposed system in semi-automatic road extraction.

For construction of a power line, technical, economical and environmental factors should be considered. Different costs of a power line, such as the operation, maintenance, and equipment installation costs are affected by its position in the real world. The route can have serious impacts on the environment. Therefore, selection of an optimal route for a new electric power line is a multi-objective problem with conflicting objectives, i.e. a route with minimum cost may have very high and unacceptable environmental impacts. A GIS-based methodology for route selection of new power lines has been tested in this paper. In this approach, a multi-objective genetic algorithm is used for route optimization. Environmental impacts are taken into consideration together with the tower construction and cable lying costs. NSGA-II approach, with two objective functions and several constraints, has been used to handle the trade-offs between the costs and environmental impacts of different power line routes. Efficiency of the proposed approach has been demonstrated for power line (400 KV) routing in Marvdasht located in south of Iran. Usefulness of the approach has been tested for location of three routes with 3 tension towers. By providing detailed information about the conflicts and trade-offs between different objectives, decision-makers would be able to choose one of the given solutions, whereas in most of the previous studies, only one route has been presented as the final solution, resulting in compensation of one objective by the others. By using the proposed method for optimization, there is no need to combine different objective functions.

The effective management of forest resources requires up-to-date and precise data. Aerial photos are one of the suitable tools for this purpose. In this study, we aimed to use texture patterns of aerial photos to provide forest canopy density mapping by Artificial Neural Network (ANN), parallelepiped (PP), Minimum Distance (MD) and Maximum Likelihood (ML) classifiers. The Study area is located in central Zagros in the west of Iran. All aerial photos were referenced to orthorectified, 0.56 meter-resolution with a Root Mean Square Error (RMSE) value of 0.2 pixels. Orthophotomosaic of photos were carried out using radiometric balance between orthophotos in order for a general view of the studied area. Co-occurrence matrix with a window size of 3×3 was used for the purpose of producing different textural bands. In textural pattern, window size of 10×10 pixels was selected for training algorithms due to the homogeneity of training samples in this window size. In this research, the discrimination of four classes of forest canopy density was enabled on textural bands with a higher accuracy than the original band. The achieved results of output maps showed the best result for ANN model with a Kappa coefficient and overall accuracy of o.8 and 85% respectively in analyzed textural bands classification in comparison to the other methods used.

In this research, a new automatic Relative Radiometric Normalization (RRN) method has been developed based on regression, including the following techniques:- Automatic detection of unchanged pixels,- Histogram modeling of subject images, and- Calculation of linear transformation coefficients for different categories of pixels according to their gray values in each band.The proposed method uses a new idea for unchanged pixels detection, which increases accuracy and automation level of the detection process. Also a new method has been proposed for categorizing of pixels according to their gray values. In this method, the number and interval of the categories are determined automatically based on histogram of subject images for each band independently. Thus, various effects of atmosphere and other effective parameters on different gray values can be modeled. This method has been implemented for two images taken by the TM sensor. Evaluation results show that the accuracy of change detection stage has improved from 95.15% to 96.14% as compared with the accuracy of conventional methods. Moreover, Root Mean Square Errors (RMSE) of the normalization has decreased about 26.44%, 50.13%, 38.54%, 5.27%, 3.75% and 3.31% respectively in the bands 1, 2, 3, 4, 5 and 7, as compared to the case where only one linear transformation is used for all area of an image. Compared to using default values for categorizing pixels and linear transformation for each category, the developed method in this research reduces RMSE of the normalization about 30.82%, 45.9%, 30.53%, 4.19%, 2.82% and 0.32% in the aforementioned bands.

Traditional methods of rangeland yield estimation are laborious, costly and time consuming. As satellite data is currently applied for estimation of the yield of rangelands through the world, this research was arranged to evaluate the capabilities of IRS LISS-III data for estimating the yield of rangelands in Tangesayyad Region (Chaharmahal va Bakhtiari Province). Imagery data were acquired on April 28th 2007 and yield samplings were carried out close to image acquisition date. Locations of 30 sampling units were determined randomly in each rangeland vegetation type. Positions of 12 plots were determined at random in each unit and the yield of grass, forbs and bushes were measured in each plot. Raw IRS satellite data in four bands were corrected and 23 currently used vegetation indices were calculated using corrected satellite bands. Correlation coefficient between 23 vegetation indices and yields of grass, forbs, bushes, summation of the three vegetation types and summation of grass and forbs were calculated. The results indicated among applied vegetation indices DVI index had the highest correlation coefficient with yields of grass and summation of grass and forbs while NDVI showed maximum correlation coefficient with total yield. NDVI also had the highest coefficient of determination for yield estimation of different vegetation groups. The results showed possible yield estimation for grass and summation of grass and forbs with reasonable uncertainty using satellite data. Very limited growth of forbs and bushes at the early season are not high enough to impact satellite sensors, hence estimation of these types of plants is accompanied by a lot of uncertainty. This research also indicated that GNDVI index could be applied for estimation of the yield of forbs.

The increasing use of gridded data in earth sciences or in hydrological estimations demands the application of mathematical approaches, especially geo-statistical ones, in obtaining accurate data. Since gridded data are generally generated from the interpolation of point data and/or processed satellite images, the use of neighborhood analysis is unavoidable in attaining more accuracy. On the other hand, in any neighborhood analysis, one is to be mindful of the selection and/or introduction of appropriate neighborhood window in terms of shape and dimension. As square or circular shapes (default shape) are usually used for neighborhood windows, an attempt has been made in the present study to address the acceptance condition of the default shape and to determine an appropriate window shape. To do so, use of image of surface semi-variogram (extracted from a geo-statistical approach) was proposed. The proposed approach was validated by determining the area of the plains in the Roodzard basin of Khozestan in Iran and extracting the channel network of Walnut Gulch Experimental Watershed (WGEW) in the State of Arizona in the United States. The findings of the study showed that the adoption of an elliptical shape for neighborhood window in extracting the Roodzard basin landform layer, led to 28 percent (93 Km2) change in the plain area extent in comparison to a square-shaped window, which will surely affect the scopes of terrain management of the basin. They also indicated that the use of a neighborhood window in WGEW led to 96 percent correspondence of the existing channel network with the extracted channel network in comparison to a circular (or circle-shaped) window.

In this paper, an unsupervised change detection method based on genetic optimization is introduced. With this aim, a cost function based on weighed minimum mean square error (MMSE) considering difference image and a random binary mask produced by genetic algorithm is used. To overcome undesired time-consuming computations of genetic optimization, the image is divided into several regions using a small moving window. In order to investigate the performance of the proposed algorithm, two sets of data are used: Simulated images and two Synthetic Aperture Radar images taken at different times. To increase the robustness of the algorithm against speckle contamination of SAR images, ratio image is used rather than simple difference one. The quantitative and qualitative results show good performance of algorithm in detecting changed pixels of the image along with texture preservation and production of changed map corresponding to true changes between two observed images. The final achieved accuracy using two simulated and real data sets is about 94 percent of all pixels in the image.