JOURNAL OF GEOMATICS SCIENCE AND TECHNOLOGY
Year:2024 | Volume:14 | Issue:2|Papers Count:9

Accurate travel time prediction is one of the important issues in the field of traffic and transportation that can significantly affect the daily life of people and organizations. In this research, four different machine learning methods including linear regression, multivariate regression, random forest and deep artificial neural network were trained to predict travel time. The purpose of this research is to predict travel time for use in intelligent traffic systems and to use and compare several new methods, including deep neural network and random forest regression, as well as considering new parameters in the computations such as weather conditions, traffic flow, travel time, and accidents and the traffic locking points compared to other studies are the innovation and comprehensiveness of this study compared to other studies. In the design and implementation of this research, real traffic data taken from Google map was used and analyzed. This data includes information such as traffic conditions, season, time of day, weather conditions, and route characteristics. The results of this research show that the deep neural network (DNN) model with R2 equal to 0.833 has a very good performance among the investigated models. This model explains 0.833% of the variance of the data and the distribution of the residuals in it is relatively central with a mean of zero and a distribution close to normal. The linear regression model with R2 equal to 0.615 has a poorer performance than DNN and explains 0.615% of the data variance. But the random regression model with R2 equal to 0.955 has one of the best performances in competition with DNN and explains 0.955% of the data variance. MSE and RMSE parameters were also used to evaluate the performance of the models, and as a result, a multidimensional comparison was made between the models, and the random forest model resulted in the lowest error values. Since in the collected traffic data, traffic accidents and consequently traffic locking points are also used in the models, and considering that the random forest model is more effectively adapted to the data despite the presence of noise and anomaly, the R2 value of this model is higher than R2 of Deep neural networks, due to the overfitting nature of Deep Learning methods.

A landslide is a natural hazard that can cause significant damage. Therefore, it is important to identify areas at risk of landslides to minimize their impact. This can be achieved by creating a sensitivity map, which involves selecting relevant factors and assigning appropriate weights. Factors and their weights can be determined using methods based on expert knowledge or machine learning and data-based methods. In this study conducted in the West Azerbaijan province, seventeen data layers related to landslide occurrence, such as geology and TWI, were prepared using GIS. These factors were considered as dependent variables, while the map of previous landslides served as an independent variable. Six different feature selection methods in the field of machine learning were employed to assess the correlation and influence of each factor on landslide occurrence. These methods included correlation method, information gain, gain ratio, CFS, Relief F, and symmetric uncertainty. The research findings indicated that different feature selection methods may yield varying results in determining effective factors. However, some factors, such as geology, were selected across all methods, suggesting a higher level of confidence in their significance. On the other hand, some factors, based on the available data, were not selected by any of the methods. Among the feature selection methods, symmetric uncertainty, information gain, and gain ratio produced similar results in terms of selecting factors. The Relief F method, however, differed from other methods due to its approach to defining neighbors. For instance, distance from the river, which was identified as a significant factor in other methods, was not selected by Relief F, while the type of soil and distance from the road emerged as conditioning factors only in this method. Finally, employing diverse methods for selecting landslide conditioning factors can simplify predictive models based on machine learning.

The water scarcity caused by recent droughts, climate change, population growth, and excessive consumption of water resources has had widespread impacts on the lives of humans, animals, and plants. Iran, due to its geographical location, climate changes, and lack of water resources, is on the verge of a water crisis. Surface water bodies, such as lakes, are also affected by this crisis. Therefore, proper monitoring, control, and management of water resources are essential. This monitoring can be carried out quickly and accurately through the use of satellite images, providing continuous reports on the status of surface water resources. In this research, the water body surface area of Zaribar Lake in Kurdistan Province was determined using radar satellite images through a local thresholding approach. This approach consists of three main steps in its implementation. In the first step, under a feature extraction process, four distinct categories of features—namely: texture, mathematical, geometric, and polarimetric features—were extracted from the primary radar image. Then, a classification process was conducted using four machine learning classification models, resulting in an initial classified image of the area. In the second step, a global threshold was applied to the radar image of the region, resulting in the identification of the primary water cluster in the area. In the final step, to refine and improve the initial water cluster, a local thresholding process was performed. In this process, based on the characteristics of the area, the type, number, and location of existing land uses were considered, and local thresholds for each cluster in the region were determined separately by calculating probability density functions (PDFs). By applying the local thresholds and then imposing a series of hydrological constraints, the final map of surface water was generated. The results obtained from the proposed approach in this research indicate that local thresholding succeeded in detecting and improving the surface water extent with accuracies of 95.44% and 98.27% corresponding to AUC and F1 score criteria, respectively. Additionally, a change detection experiment of Chitgar Lake was implemented to challenge the effectiveness of the proposed approach. The results of this experiment, with accuracies of 94.55% and 96.65% corresponding to AUC and F1 score criteria, respectively, validated the efficacy of the proposed approach.

The production of land cover maps (LCM) provides essential information about land types and their characteristics, playing a significant role in updating urban maps, managing natural resources, environmental protection, and sustainable development. In this context, the use of image processing techniques and free remote sensing data is considered an optimal method for generating land cover maps (LCM). In this study, various artificial intelligence approaches, including machine learning (ML) and deep learning (DL) algorithms, were used to produce the LCM. The ML approach includes two stages: feature extraction and classification. In the feature extraction stage, texture features extracted from the gray-level co-occurrence matrix (GLCM), including mean, variance, homogeneity, contrast, and entropy, were used. For classification, the logistic regression (LR), decision tree (DT), random forest (RF), and support vector machine (SVM) algorithms were employed. In the DL approach, deep learning semantic segmentation models, including U-Net, U-Net++, ResU-Net, and MRU-Net, were used. To evaluate the accuracy of the ML and DL algorithms in producing the land cover map, Sentinel-2 images from two areas located in the west of Tehran were utilized. The results of this study were examined in three different sections: ML, DL, and their comparison. In the ML section, the RF model, which used a combination of the image's primary bands and texture features, performed better than other models with an overall accuracy of 95.21% and a Kappa coefficient of 92.62%. In the DL section, the MRU-Net model produced the most optimal LCM with an overall accuracy of 95.33% and a Kappa coefficient of 92.73% compared to other deep models. The MRU-Net model, without using texture features, improved overall accuracy and the Kappa coefficient by 0.53% and 0.82%, respectively, compared to the RF model using a combination of primary image bands. Furthermore, compared to the RF model, which used a combination of primary bands and texture features, the MRU-Net model's overall accuracy and Kappa coefficient were 0.12% and 0.11% higher, respectively.

Despite the fact that the energy sector has contributed to the destruction of the environment through its emissions of greenhouse gases, human societies have paid more attention to renewable energy sources such as small hydropower plants in response to this issue. For the purpose of determining zoning, 12 layers of environmental, technical, and geographical criteria have been used. In order to achieve the results of this study, we combined multi-criteria decision-making (MCDM) with fuzzy algorithms. Basically, the fuzzy-AHP method of weighting is used as a method for evaluating the technical and environmental criteria that do not have an internal relationship with one another. Using the fuzzy-ANP method, we are able to weight geographical criteria that are related to one another and are evaluated in terms of their relative importance. In order to identify the zoning map, layers have been overlaid by using the gamma operator 0.9 in order to identify the zoning map. In order to produce the final zoning map, the zoning and the physiographic maps are combined using the Sum operator. Therefore, 13 suitable sites were selected for the construction of power plants, resulting in 22084.69 megawatts of energy being generated per year and 5.8 tons of greenhouse gases being prevented from being released into the atmosphere. During the course of this study, a watershed located in Iran, known as the Karoun watershed, was studied. Furthermore, the methods applied in this research could be performed in other watersheds and evaluated the potential of other power plants such as solar and wind power plants.

The surface reflectance captured by satellites is exposed to the atmosphere gases and aerosols, and its value changes due to colliding with those particles. The change of surface reflectance can also effects on other uses of this quantity. Atmospheric correction using different correction methods and algorithms could lead to different results. In addition, different values of atmospheric parameters such as water vapor, visibility, aerosol optical depth and CO2 which are extracted from ground measurements, or from products of other sensors, or based on similar studies and the users’s guess can achieve different results. Furthermore, even small changes in reflectance can lead to significant uncertainty in parameter retrieval or other remote sensing applications. The main purpose of this study is to evaluate the effect of atmospheric parameters on the accuracy of parameter retrieval from reflectance. For this regard, after implementing the FLAASH atmospheric correction on the Landsat-8 image by changing the values of water vapor and visibility, FCOVER was modeled by Neural network and Regression tree algorithms. Afterwards, effects of the uncertainty related to each atmospheric parameter is evaluated by paired sample T-test. Results indicated changes of water vapor and visibility influences FCOVER retrieval, and causes more than 5 percent error. Also, the neural network and regression tree have shown relatively similar and suitable performance for FCOVER modelling despite the uncertainty in the input parameters of FLAASH model.

Facilitating autonomous movement is essential for individuals coping with mobility impairments in their everyday activities. However, their movement is frequently restricted by environmental and social factors. This limitation is exacerbated during adverse weather conditions, such as precipitation, which poses challenges to the accessibility of sidewalks, particularly for wheelchair users. This article delves into the examination of sidewalk accessibility for people with mobility impairments, with a specific focus on wheelchair users, during the rainy season. The primary objective of this study is to explore the impact of environmental factors, including dynamic elements like precipitation, on the independent movement of individuals with mobility impairments. The proposed methodology involves assessing the confidence level of users when encountering different sidewalk components under varying precipitation intensities. A fuzzy decision-making model is then employed to evaluate the overall accessibility of sidewalks. In addition to providing accessibility maps tailored to different rainfall intensities, this research utilizes a similarity index to compare the effects of varying precipitation intensities on different sidewalk components. The results obtained underscore the significant influence of rainfall intensity on the accessibility of each segment of the sidewalk. Therefore, it is imperative to factor in precipitation intensity when considering and planning for the accessibility needs of individuals with mobility impairments. While most sections of the sidewalk remain accessible to wheelchair users during light rainfall, increased precipitation significantly diminishes accessibility, reducing it to an average level across most areas. Furthermore, during extreme rainfall, accessibility in sections with narrow widths and poor surface quality becomes severely limited.

Information on land use and cover needs to be gathered due to the growing urban population, city growth, and urbanization. Applications for this data include environmental protection, urban planning, planning for urban infrastructure, and strategic planning to guarantee the sustainable growth of urban areas. The primary source of data on land cover and land use at the moment is remote sensing imagery. Information about land cover and land use can be retrieved from remote sensing images using image classification techniques. In terms of classification accuracy, deep learning techniques recently outperformed other methods for classifying land use and cover. Convolutional neural networks (CNNs), which are quite popular in this field, are one of the significant deep learning classification architectures frequently used in land cover and land use classification. Recently, the convolutional neural network technique known as ResNet has been applied to remote sensing applications, particularly for the classification of land use and cover. ResNet models are an effective choice for classifying land cover and land use because they can handle the vanishing gradient issue. The primary objective of this study is to assess the performance of the Glorot Uniform and Random Uniform weight initializers in the ResNet50, ResNet101, and ResNet152 architectures for extracting the land cover and land use of the EuroSat dataset. The weighted F1 score, IoU indexes, overall accuracy, and kappa coefficient were used to evaluate the accuracy of the results. ResNet101's corresponding values for these indexes were, in turn, 0.8869, 0.7951, 0.8871, and 0.8743. These results indicate that, in terms of classification accuracy, ResNet101 has outperformed the ResNet50 and ResNet152 methods.

The ionosphere, the upper layer of the earth's atmosphere, contains free ions and electrons, which greatly impact radio and satellite signal transmission. Correcting the observations of global navigation satellite systems (GNSS) to increase the accuracy of positioning, improving the design and performance of radio communication systems, correcting the effects of the ionosphere to improve the accuracy of radar data, better understanding atmospheric and spatial changes due to storms, solar and geomagnetic activities are some examples which demonstrate the considerable importance of ionospheric electron density (Ne) prediction. Electron density prediction is technically superior to total electron content (TEC), providing more accurate information regarding the distribution of electrons at different heights of the ionosphere. This research aims to predict the ionospheric electron density by predicting the parameters of the continuity equation using an artificial neural network. This study, which was conducted in three broad phases, predicts the electron density of the ionosphere for Iran on the 129th day of 2016. Firstly, the parameters of the continuity equation are computed using the electron density obtained from the International Ionosphere Reference (IRI), a restricted linear regression procedure, and the sun's radiant flux on grid points spaced 0.5 degrees apart. Together with the input data from the 123rd to 128th days of 2016. In the second phase, these values are used in the artificial neural network to train a feed-forward neural network. In the third stage, the ionosphere's electron density is predicted by inserting the artificial neural network's predicted parameters into the continuity differential equation. After analyzing the data, it was determined that the average RMS value of the difference between the IRI electron density and the predicted electron density on the 129th day was 1.0943×1011 for three hours, 2.3733×1010 for two hours, and 1694.5×1010 for two hours.