ENGINEERING JOURNAL OF GEOSPATIAL INFORMATION TECHNOLOGY
Year:2020 | Volume:7 | Issue:4 |Papers Count:12

Estimation of forest biomass has received much attention in recent decades. Airborne and spaceborne (SAR) have a great potential to quantify biomass and structural diversity because of its penetration capability. Polarizations are important elements in SAR systems due to sensitivity of them to backscattering mechanisms and can be useful to estimate biomass. Full Polarimetric Synthetic Aperture Radar (SAR) data used in this research was acquired by SETHI over Remningstorp, a boreal forest in south of Sweden. A new method based on Polarimetric indicators from covariance and coherency matrixes by changing the polarization basis using transformation matrix in the boreal forests at L and P-band is presented. The presented method showed its capability to improve forest biomass estimation. The correlation between biomass and extracted Polarimetric indicators is investigated before and after changing polarization basis. Particle swarm optimization in binary version is used to select optimum Polarimetric indicators and afterward biomass is estimated based on these optimum parameters. Results indicated that maximum correlation between biomass and Polarimetric indicators was in HV and HH-VV polarizations before changing polarization basis. After changing the polarization bases, the results show significantly higher correlation of biomass with the extracted polarization variables. The results have been improved approximately about 6% and 2% in L and P band respectively, after extraction of optimum parameters by particle swarm optimization and using linear regression model for estimation of forest biomass.

Today, the global positioning systems (GPS) do not work well in buildings and in dense urban areas when there is no lines of sight between the user and their satellites. Hence, the local positioning system (LPS) has been considerably used in recent years. The main purpose of this research is to provide a four-layer artificial neural network based on nonlinear system solver (NLANN) for local positioning problem. To evaluate the performance of artificial neural network, three methods of gauss-newton (GN), genetic algorithm (GA) and hybrid particle swarm optimization (HPSO) have been used. The results indicate that the proposed model has high accuracy. The accuracy of the artificial neural network on the simulated data is 0. 05 m, while the best accuracy in other algorithms is about 0. 45 meters. In the data of Italy's GPS network, the artificial neural network has been reached to accuracy below 10 cm in one minute. Also, artificial neural network has better accuracy in different dimensions of study area and different signal to noise ratio (SNR), and by increasing the number of stations, it has achieved good results in less time. Whereas other algorithms have not get well accuracy. However, the HPSO has better results related to GA and GN algorithms.

Access to current and future water resources is one of the concerned problems for managers and policymakers around the world. Because of the communication between water resources and land use, these two topics had come together in different researches. Scenarios designed in regional land planning provide the basis for analyzing the existing opportunities and making the right decisions for managing these natural resources. In this research, a combination of the Markov Chain model and multilayer perceptron network (MLP) were used for predicting the land-use changes in Sarab watershed and the SWAT model was used for hydrological modeling of the watershed area. Using the land use map in 2015, soil map, digital elevation model and meteorological data during the period (1987-2015), the hydrological model of the area is formed and also calibrated. According to the land-use changes in the past (1987-2015), three scenarios defined and three land use maps have been predicted for 2030 by modeling the land-use changes and calculating the conversion probability matrix using the Markov chain model. The watershed hydrological response to the first scenario with the title of conversion of grassland to the irrigated agriculture was observed an increase of 0. 7% of the annual average run-off and a 4% decrease in the river flow. In the second and the third scenarios, the surface run-off has been increased by 1% and 2. 5% respectively by conversion of the grassland to the rain-fed agriculture and bare lands. Flow changes in these two scenarios show an increase of 1. 8%. According to the results of this research, grazing and conversion grassland to bare lands will have the greatest impact on underground water resources in the Sarab basin. Furthermore, the expansion of irrigated agriculture lands, by increasing the harvesting of surface water and underground water resources will result in a significant reduction of these resources.

Since Iran plateau is located in the Alpine-Himalayan Orogenic belt, it is recognized as a region with a high seismic risk. Thus, investigation of geodynamic activities of the faults, their slip rates and corresponding deformation fields is very important for quantification of possible seismic risk in this region. The aim of this study is to analyze the tectonic features of eastern part of Iran plateau and determine the long-term slip rates of active faults in this part. To do so, the velocity vectors of geodynamic stations, the directions of principal stresses from global models and fault slip rates derived from different geological sources are assimilated using kinematical finite element model (Neokinema) to derive the optimum slip rates on the fault surface. Neokinema uses three controlling parameters to run the model. These parameters are determined by sensitivity analysis. The final slip rates using this model are determined with the error of 1 mm/yr. To validate the results of Neokinema model, the slip rates of KouhBanan, Dasht-e Bayaz and Nayband faults computed from the model are compared with those of geological observations which illustrate a good consistency between model prediction and geologic observations.

Nowadays, ground vehicle monitoring (GVM) is one of the areas of application in the intelligent traffic control system using image processing methods. In this context, the use of unmanned aerial vehicles based on thermal infrared (UAV-TIR) images is one of the optimal options for GVM due to the suitable spatial resolution, costeffective and low volume of images. The methods that have been proposed for vehicle extraction from thermal infrared imaging often experience problems such as low accuracy in detection, segmentation (e. g. HOG+SVM) and also the need for big data training (e. g. deep learning methods). In the present study, a new model, called SegRBM-Net, based on deep learning (DL) and the restricted Boltzmann machine (RBM) is being presented. One of the features of the SegRBM-Net model is the improving accuracy of vehicle detection and segmentation from thermal infrared images by using both convolutional layers and the features of the Gaussian-Bernoulli restricted Boltzmann machine. This structure has led the algorithm to find the target faster and more accurately than other DL methods. To examine the performance of the proposed method, we performed a controlled benchmark (e. g. high density of vehicles scene, and difference in viewing angle) of SegRBM-Net and other DL models on four UAV-TIR image datasets. The results showed that the SegRBM-Net model with a mean accuracy of 99% and improved processing speed compared with similar methods have a good performance.

In recent years, the development of the use of location-based tools has made it possible to produce geometric trajectories from the user's movement paths. In this way, users' goal of traveling and related activities can be considered in addition to the geometry and route shape. the user activity trajectory represents the sequence of the visited activities and its related analysis as presented in many studies. In the meantime, the most important analysis is the identification of frequent patterns to predict future activities. In previous studies, only one user’ s data was used, as well as previous activities of the user were not taken into account in presented frequent patterns. For this purpose, this paper presents a framework such as After encoding the activities and forming a sequence matrix, the frequent patterns using all users' trajectories are identified considering previous activities. The proposed method also offers the ability to identify frequent patterns for the origin, destination, or interesting activity. The results of comparing the proposed method with two methods based on tree structure represent a mean of 60% reduction in computing time to form the database.

This paper presents a new method for detecting the features using LiDAR data and visible images. The proposed features detection algorithm has the lowest dependency on region and the type of sensor used for imaging, and about any input LiDAR and image data, including visible bands (red, green and blue) with high spatial resolution, identify features with acceptable accuracy. In the proposed approach, detecting the features by using the object-based analysis theory as the main approach has been performed. Also two different approaches and innovations in order to increase “ Level of Automation” (LoA) and level of accuracy and precision in detecting process have been proposed and performed. The first approach uses visible and LiDAR data independently in order to resolve the problem of high-dependencies between data in the existing algorithms. The second proposed method has been suggested in order to the detection of vegetation regions. Among the characteristics of this method it can be mentioned that there is no need to use the infrared band in the image data and also there is no need to intensify information of the laser returns. By assessing the results of available data classification, the determined overall accuracy of the proposed method on average, about vegetation regions is 98 % which shows the highest value compared with other features. The proposed method about other features also achieves acceptable accuracy.

Over the past decades, urban growth has been known as a worldwide phenomenon that includes widening process and expanding pattern. While the cities are changing rapidly, their quantitative analysis as well as decision making in urban planning can benefit from two-dimensional (2D) and three-dimensional (3D) digital models. The recent developments in imaging and non-imaging sensor technologies, such as airborne Light Detection and Ranging (LiDAR) system, lead to a huge amount of remotely sensed data which can be employed to produce 2D/3D models. Although much of the previous researches have investigated on the performance improvement of the traditional data analyzing techniques, recently, more recent attention has focused on using probabilistic graphical models. However, less attention has paid to Conditional Random Field (CRF) method for the classification of the LiDAR point cloud dataset. Moreover, most researchers investigating CRF have utilized cameras or LiDAR point cloud; therefore, this paper adopted CRF model to employ both data sources. The methods were evaluated using ISPRS benchmark datasets for Vaihingen dataset on urban classification and 3D building reconstruction. The evaluation of this research shows that the performance of CRF model with an overall accuracy of 89. 06% and kappa value of 0. 84 is higher than other techniques to classify the employed LiDAR point cloud dataset.

The estimate of the damage caused by the earthquake and other natural disasters in the first days after the occurrence of these events can provide a quick damages assessment and help to manage the crisis. Several methods are available to investigate the extent of earthquake’ s damage. Optical remote sensing, photogrammetric methods (UAVs and LIDARs), radar interferometry (InSAR) and field observations are examples of the operational methods. Today, InSAR technology has become a powerful yet inexpensive tool for monitoring deformation and changes in the Earth's crust. The Coherence Product is derived from SAR imagery. The lack of coherence in radar images can be due to several factors such as vegetation, changes in the dielectric coefficient in the master and slave images, high gradient slop areas, soil erosion, damage degradation and etc. In this paper, we have tried to estimate the extent of the damaged area by focusing on InSAR technique and eliminating cells that have lost their coherence due to vegetation, dielectric coefficients, and high mountain range areas. In this regard, Envisat Advanced SAR images of Bam, Iran, that were acquired before and after the 2003 earthquake were used. The coherence of cells with a mean value of 0. 2 in the area of Arg-e-Bam with its high degradation level indicates the ability to use this criterion in the rate of destruction. The results of the Bam earthquake investigation indicate that about 23. 5% of 14290 hectares of the study area collapsed and about 31% of it had high degradation.

The drought can be described as a natural disaster in each region. In this study, one of the important factors in drought, vegetation, has been considered. For this purpose, monthly vegetation cover images and snow cover data of MODIS and TRMM satellite precipitation product from 2009 to 2018 were used for the study area of Iran. After initial preprocessing, we have used artificial neural network method and hybrid neural network and wavelet transform method to predict the normalized difference vegetation index (NDVI). After training the two algorithms using the time series of (NDVI) index as well as the time series of snow cover and precipitation from 2009 to 2017, the (NDVI) index is predicted for twelve months from 2018, which is finally estimated with real values. The results and prediction accuracy for these two algorithms are different and in general the combined neural network and wavelet transform method has higher accuracy compare to the neural network method so that the twelve average of 2018 is equal to the root mean square error of 0. 055 and coefficient of determination was 0. 804. The results also show that in both methods the accuracy of the index in the early months of 2018 is better than the end months. Therefore, this method can be used to predict this index, as one of the drought parameters.

Today, medicinal plants have a special place in the economy and health of a society. Due to the natural growth of many of these products, the necessity of zoning them for optimum and optimal utilization seems necessary. Traditional zoning solutions are not efficient due to their low accuracy and speed, therefore a new approach is needed. Remote sensing data have many applications in various fields including target detection because of their spectral, spatial and temporal information of land surface phenomena. In this paper, target detection methods including Constrained Energy Minimization (CEM), Matched Filtering (MF), Adjusted Spectral Matched Filter (ASMF) and Adaptive Coherence Estimator (ACE) are used to detect Amygdalus Scoparia in Sentinel-2 satellite time series images. In this process, firstly, the filtering of undesirable effects (unlikely areas of plant growth) is eliminated from the time series of images. Then, with the help of hyper heuristic optimization, the optimal features from time-series were identified to reduce the dimension from one hand and increase the detection accuracy from the other hand. The final detection map is generated by weighting the results obtained from each training sample with a different share of the target. The generalizability of the proposed solution was evaluated using the selected optimal features elsewhere, using the ground truth map. The ROC and its subarea (AUC) are used to evaluate the results. In the optimization phase for feature selection, the AUC index for all detection methods used was greater than 0. 99. The best results in this process were obtained by the CEM detection method, which achieved the accuracy of 0. 993 and 0. 846 in the optimization and independent evaluation, respectively. The results of this study indicate the ability of Sentinel-2 multiplexed time series images to detect targets such as medicinal plants.

Surface soil moisture is an important variable that plays a crucial role in the management of water and soil resources. Estimating this parameter is one of the important applications of remote sensing. One of the remote sensing techniques for precise estimation of this parameter is data-driven models. In this study, volumetric soil moisture content was estimated using data-driven models, support vector regression (SVR) and multi-layer perceptron artificial neural network (ANNMLP) method. The parameters of the two models are optimized by the Genetic optimization algorithm. Estimation of volumetric soil moisture content with the two top models was performed using two types of radar image (Sentinel 1) and optics image (Sentinel 2), in which optimized optics image bands were identified by the Genetic optimization algorithm. After estimating the volumetric soil moisture map, four outputs of the two methods are compared. The best estimate of the volumetric soil moisture content has been achieved by the support vector regression (SVR) method with the Sentinel 1 image. The worst estimate of the volumetric soil moisture content has been achieved by the multi-layer perceptron artificial neural network (ANN-MLP) method with the Sentinel 2 image. The accuracy of this study was calculated by the square of correlation coefficient of the measured volumetric soil moisture content and the estimated volumetric soil moisture content, which the best and worst correlation coefficients, respectively, 0. 659 for Sentinel1 image using support vector regression method and 0. 409 for Sentinel2 image using multilayer perceptron neural network method have been calculated. The root mean square error (RMSE) is also used to calculate the error of the methods. The lowest and highest errors were calculated by 0. 291( 𝑚 3 𝑚 3) for Sentinel1 image with support vector regression and 0. 4612( 𝑚 3 𝑚 3) for Sentinel2 image with Multilayer Perceptron Artificial Neural Network.