Journal of the Indian Society of Remote Sensing最新文献

Despite recent advances of Deep learning in numerous computer-vision tasks, the possibility of classifying aerial images has not been thoroughly explored. The aerial image classification purely depends on spectral content is an interesting research subject. In this work, a novel Optimized Ghost Network-based Aerial Image Classification (OGN-AIC) approach is proposed to classify the different Aerial images from the dataset. The image is first preprocessed using Gaussian filtering techniques to enhance its quality and remove noise. Consequently, the features are extracted using Ghost Network for classifying the different landscapes. The input images are classified into five different categories namely: Dryland, Forest, Airport, Mountain, and Parking. The classification results are improved by the Slime Mould optimization (SMO) algorithm, which normalizes the parameters of the network. The efficiency of the proposed OGN-AIC model was assessed utilizing precision, F1 score, specificity, sensitivity and accuracy. According to the experimental results, the proposed OGN-AIC model attains an overall accuracy of 98.24%. The proposed OGN-AIC technique enhances the overall accuracy range by 14.2%, 0.77%, 14.5%, 1.08%, and 11.17% better than Artificial Neural Networks, k-nearest neighbor, cutting-edge Deep Convolutional Neural Network (DCNN), semi-supervised Convolutional Neural Network and Cellular neural network respectively. As a result, the classification using a deep learning network is more accurate and effective for classifying aerial landscape images than the traditional DL techniques.

A quintessential component of any type of landslide studies, like susceptibility mapping, risk assessment and identifying the role of influencing parameters, is a landslide inventory map (LIM). LIM helps to analyse the spatial and temporal characteristics of landslides, and is also vital for constructing a landslide early warning system. Thus, LIM plays a vital role in landslide disaster risk reduction processes. As a paradigm work, this study aims at creating a relatively complete landslide inventory dataset for the 2018 rainfall-triggered landslide in a small sector of the south of the Western Ghats, called Kodagu in Karnataka, India. Integration of field investigation, and visual interpretation of pre- and post-landslide images of the Google Earth and Sentinel-2A satellite data were used to construct this LIM. Field investigation was aimed at two components: (i) to verify the created inventory from satellite imageries and (ii) to map those landslides that could not be identified in the images due to non-availability of images or cloud covered images or for any other reasons. The final, newly created LIM comprised 267 landslides: 89 through field investigation, and 178 by image interpretation. Of these, 153 are shallow slides and 114 are debris flow, with major damages attributed to debris flow. The created LIM is uploaded in GitHub and can be freely downloaded by researchers and students for further studies. This LIM was further used to generate a landslide susceptibility map (LSM) using machine learning techniques. This empirical method of LSM was done in Google Colab, and the results show that Random Forest as the best model for the study area. Majority of the landslides are confined within the slope range of 14°-29°, elevation between 970 and 1100 m as well as 1200 and 1700 m, slope aspect corresponding to southwest and west direction, and convex surfaces, especially near roads within 750 m.

Satellite images of different spatial resolutions and separate object classification approaches have been employed for Land Cover (LC) mapping in local and regional projects. Nevertheless, the mapping skills and the attainable accuracy of the LC classification in the current landscape are influenced by the spatial resolution of the datasets utilized and the classification techniques used. In this paper, the effect of the spatial resolution of satellite images (Landsat 8 OLI with 30 m and Sentinel-2 A MSI with 10 m data) on LC mapping accuracy was evaluated by using four non-parametric classification techniques; Random Forest (RF), Neural Network (NN), Support Vector Machine (SVM), and K-Nearest Neighbor (K-NN). The findings showed that SVM could be used efficiently with Landsat 8 (30 m) to classify LC at local and national scale research as it achieved the greatest accuracy utilizing SVM with Overall Accuracy (OA) = 84.44% and K coefficient value (K) = 0.78 followed by RF, K-NN, and NN. SVM has not outperformed other classification methods. Similarly, classification with Sentinel 2-A achieved the greatest accuracy by SVM and RF classifiers, with an average performance for mapping OA = 96.32% with K = 0.956, followed by K-NN and NN, while RF and SVM can be appropriate for classifying LC based on Sentinel-2 A (10 m) images. In addition, SVM and RF have been slightly more efficient than other classification approaches, and Sentinel-2 A-based LC mapping observations were more precise and dependable compared to Landsat 8. Our findings further confirm that both datasets are similar in 88.91% of the outcomes based on the comparison between Sentinel-2 A and Landsat 8 LC maps. Lastly, the spatial resolution of the data has a big effect on how the LC is mapped.

Drought is a prevalent complex natural disaster due to its environmental extent and can severely impact global ecosystems. For the purpose of monitoring droughts and assessing their effects on a regional and global level, usually remote sensing data with an appropriate temporal and spatial resolution can be accessed. This research utilized the Moderate Resolution Imaging Spectroradiometer (MODIS)-based normalized difference vegetation index (NDVI), land surface temperature (LST), gross primary productivity (GPP) and vegetation health index (VHI) to investigate the historical duration, severity and recovery period for drought in selected districts of Balochistan. The Pearson correlation was used to determine the local link between the duration and severity of the drought between 2001 and 2021. The results showed that 2001, 2002, and 2004 were the driest years in which extreme to mild drought occurred with severity of 36%, 48% and 48% respectively. On the other hand, the drought duration result revealed 80–275 days, 160–275 days, and 176–275 days for 2001, 2002, and 2004 respectively. The result also indicated that crop land, water bodies, grass land and forest land, were positive correlation while shrub land was negative correlation with drought severity. On the other hand, crop land, water bodies, grass land and forest land, were negative correlation while shrub land was the positive correlation with drought duration. The drought recovery period analysis resulted in 16–66 days, 18–67 days, and 17–66 days for the years 2001, 2002, and 2004 respectively. With every aspect considered, the study offers insightful information on drought resistance for improved management.

Building footprint extraction is crucial for various applications, including disaster management, change detection, and 3D modeling. Satellite and aerial images, when combined with deep learning techniques, offer an effective means for this task. The Multi-scale Aggregation Fully Convolutional Network (MA-FCN) is an encoder-decoder model that emphasizes scale information, producing the final segmentation map by concatenating four feature maps from different stages of the decoder. To enhance segmentation accuracy, we propose two novel deep learning models: Attention MA-FCN and Residual Attention MA-FCN. Attention MA-FCN incorporates attention gates in the skip connections to emphasize relevant features, directing the model’s focus to essential areas. Residual Attention MA-FCN further integrates residual blocks into the architecture, using both attention mechanisms and residual blocks to improve stability against gradient vanishing and overfitting, thereby enabling deeper training. These models were evaluated on the WHU, Massachusetts, and Jinghai District datasets, showing superior performance compared to the original MA-FCN. Specifically, Residual Attention MA-FCN outperformed MA-FCN and Attention MA-FCN by 3.6% and 0.92% on the WHU dataset, and by 5.51% and 0.91% on the Massachusetts dataset in terms of the Intersection Over Union (IOU) metric. Additionally, Residual Attention MA-FCN surpassed MA-FCN, Attention MA-FCN, Mask-RCNN, and U-Net models on the Jinghai District dataset. Due to the significance of building footprint extraction in various applications, the results of this study indicates that the proposed methods are more accurate than the MA-FCN model with better performances in IOU and F1-score metrics.

This research was carried out to monitor the NO₂ and SO₂ (Nitrogen and Sulfur IV oxides) levels around the Oben gas flow station in Edo State, Southern Nigeria, using remote sensing data. Secondary data was collected from the Sentinel 5P satellite and processed using Google Earth Pro, ArcMap, Google Earth Engine, and Microsoft Excel to determine the concentrations of the pollutants of interest in the study area for 2019 and 2020. In 2019 and 2020, the maximum mean values for SO₂ were 3.99 E-5 mol/m² and 4.26 E-5 mol/m², respectively, and for NO₂, the maximum mean values were 6.63 E-5 mol/m² and 6.88 E-5 mol/m², respectively. For the seasonal variations in concentrations of pollutants, there was p > 0.05 (no significant differences) in the seasonal variation between the concentrations of SO₂ in 2019 (t (5) = 1.410) and 2020 (t (5) = 2.399). There was a significant difference (p < 0.05) between NO₂ concentrations in the wet and dry seasons for 2019 (t (5) = 5.719) and 2020 (t (5) = 5.991). Also, there was a significant variation between the concentrations of NO₂ in 2019 and 2020 at p > 0.05, but not for SO₂. Based on the findings from this study, it is recommended that stricter enforcement of already existing legislation on gas flaring and finding cleaner or alternative sources of energy like biofuels and biogas, are highly needed to reduce any unforeseen health and environmental impact in this zone.

A solar cycle linkage is investigated on wind, temperature and surface pressure throughout 1981 to 2021 using The Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA2) data. Sunspot data were obtained from the Royal Observatory of Belgium (solar cycles 21 to early 25). It is determined from the analysis that solar cycle intensity decreased gradually in the last four decades; wind data at six stations (metropolitan cities) over India showed a consistent decrease by an amount of 0.3 to 0.6 m/s, on average 0.5 m/s. A strong association is evident between the solar cycle and wind variability, this is more evident while approaching closer to the equator from northern tropics. Wind speed declined more clearly near and around 10°N latitudes. This consistent decline assumes a strong significance during winter in Northern India i.e., the climatic trend is unfavourable for dispersing the pollutants and will harm the air quality in future. On the other hand, temperature and pressure data showed a climatic increasing trend (~ 0.9 °C and ~ 1.5–2.0 mb) most prominently seen from 2000 to 2021 over the tropical region; which became slightly weak in extra tropical region (Delhi). Temperature and pressure data did not show a relationship with sunspot numbers. It is determined that solar cycle variability has influenced windspeed (positive correlation ~ 0.5, with 95% confidence level) near ground level.

In this study, the aim was to determine the relationship between tectonic lineaments derived from satellite data and springs. The study area covers Gümüşler (Nigde) and its surroundings in the northern Nigde Massif. The study investigated the connection between this area’s tectonic lineaments and natural water resources. Remote sensing methods used in mineral exploration and the determination of geothermal fields have also been applied in this study, supported by intensive field studies. The high-grade metamorphic rocks of the massif exhibit faulted and fractured structures due to polyphase deformation, giving these rocks important aquifer characteristics. Numerous springs have formed due to the effects of faults and fractures. The study definitively established the relationship between 82 natural water resources (springs) and tectonic lineaments. Almost 87% of the identified natural water resources are located on lineaments. In addition, other springs were determined to have discharge due to discontinuities in formation boundaries.

In the era of smart computing, edge computing, and machine intelligence, the Internet of Things (IoT) is playing a greater role in establishing hyper connective, cost-effective infrastructure for monitoring the environment. With the increase in the level of urbanization and population density in very fast-growing cities like Nagpur, the increase in air pollution needs to be monitored. This requires a network of Pollution monitoring systems for carrying out spatio-temporal analysis of pollution in the city on a real-time basis. Such established networks can be a key to understand the sources of pollution under various city conditions. To monitor and manage air pollutants, it is essential to put in place monitoring stations at multiple places. Although commercial pollution monitoring stations exist, they are limited in number. In this study, an attempt has been made to develop and implement a network of IoT devices using cost effective Metal Oxide Semiconductor based gas sensors integrated with ATMEGA 328P Microcontroller. Commercial systems are found to be space, energy and cost expensive. The developed pollution monitoring system can be replicated easily since they are compact in size, cost-effective, network and energy independent. This study discusses the development and implementation of a network of 10 smart IoT sensors in the Nagpur metropolis. The developed smart air pollution monitoring system combines IoT technology with real-time pollution monitoring systems. It measures and monitors temperature, humidity and pollutant concentration of Carbon Monoxide, Ozone, Carbon Dioxide, Sulphur Dioxide and PM2.5 and Nitrous oxides simultaneously. The study envisages to support Sustainable Development Goals – SDG11 which aims to reduce the environmental impact of cities by improving air quality.

The extraction of structural lineaments was conducted on a portion of the Zagros orogenic belt in western Iran, using filters applied to Landsat satellite imagery (ETM) and a digital elevation model (DEM). The study area was divided into internal (Sanandaj-Sirjan) and external (Zagros) subzones by the Main Zagros Thrust. To extract lineaments, Edge Detector, Spectral Rationing, Principal Component Analysis (PCA) filters, and color combinations were applied to the ETM satellite imagery, while a directional filter (Sobel) was applied to the DEM for enhanced visual interpretation. The analysis identified 350 fault lineaments with a total length of 3689 km. The majority of these lineaments were shorter in length, with 188 lines measuring over 5 km, 110 lines between 5 and 10 km, 39 lines between 10 and 20 km, and 10 lines between 20 and 30 km. Only three lineaments exceeded 30 km in length. Statistical analysis of the lineaments, presented in Rose diagrams, revealed a predominance of NW and NE trends, with less frequent WNW, NNE, and E-W trends. The most dominant trend observed was NW. These findings suggest that the extracted lineaments are largely consistent with the faults in some inner subzones identified in previous studies of adjacent areas. However, differences in lineament orientations and densities, when considering subzones, were attributed to the likely reactivation of basement faults.