LEE- Uydu Haberleşme ve Uzaktan Algılama Lisansüstü Programı
Bu topluluk için Kalıcı Uri
Gözat
Konu "deep learning" ile LEE- Uydu Haberleşme ve Uzaktan Algılama Lisansüstü Programı'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeEarthquake damage detection with satellite imagery and deep learning approaches: A case study of the february 2023, Kahramanmaraş, Turkey earthquake sequence(Graduate School, 2023-08-14) Elik, Fatma ; Sertel, Elif ; 705201004 ; Satellite Communication and Remote SensingIn recent years, the fusion of deep learning techniques, remote sensing technology, and artificial intelligence (AI) has profoundly transformed the field of disaster management and damage assessment. The increased availability of high-resolution satellite imagery and advanced computer vision techniques now makes it possible to analyze Earth observation data at a large scale and with unparalleled precision. This thesis investigates the application of remote sensing and deep learning techniques to perform post-earthquake damage classification using computer vision and focuses specifically on the earthquakes that occurred on February 6th, with an emphasis on Kahramanmaraş province. The objective of this thesis is to investigate the potential of a variety of deep learning techniques, evaluate their accuracy in recognizing structurally compromised buildings, and utilize satellite imagery in conjunction with diverse open-source spatial data to enhance research on earthquakes. This master's thesis specifically delves into the integration of remote sensing, computer vision, and earth observation methods within the field of geophysics and earthquake studies. Thus, in this study it is aimed to showcase the application of computer vision in the analysis of post-earthquake damage and underscore the importance of rapid intervention in such critical situations. The thesis places significant emphasis on the use of satellite imagery and pixel-based classification for the classification of images in earthquake damage assessment. The UNet, DeepLabV3, and PSPNet architectures are implemented using the ArcGIS Pro API for Python, an innovative and supportive tool for scientific research. The primary data source for the investigation is RGB images from Maxar Technologies. The research examines three cities that were affected by the February 6, 2023, Kahramanmaraş earthquake sequences: Kahramanmaraş, Hatay, and Gaziantep. Damage-assessed data points are received thanks to Yer Çizenler Non-Governmental Organization (NGO), and recently modified building footprints are taken from Humanitarian OpenStreetMap (HOTOSM), and they are all used to analyze the damage. Labeled polygons are generated within a 5-meter distance of the damage points. However, assigning values for further and closer distances has a negative impact on the model accuracy. The training data, exported based on the satellite imagery and damage level assigned data points, provides a balanced dataset for Kahramanmaraş, where the building footprints match the images most effectively. In Hatay, the damage level assigned data distribution is the most balanced, but the building footprints do not align well with the images. Gaziantep presents a good match between the building footprints and images, but the distribution of the damaged data classes is highly imbalanced. Consequently, the decision is made to focus on training the model for Kahramanmaraş province due to the similarity in roof and building types, which has the potential to adapt the approach to other cities in the region as well as the earthquake-affected region under investigation. Image sizes of 256x256 pixels with 128 strides and 4 batches gave us the optimum model results among other options in the DeepLabV3 ResNet50 encoder. In conclusion, this master's thesis demonstrates the potential of combining remote sensing, computer vision, and earth observation techniques for geophysics and earthquake studies. Also, it is aimed to use different data types from open sources and use these different data types to make damage detection after earthquakes. The utilization of the ArcGIS Pro Python API, satellite imagery, pixel based classsification, and labeled training data provides insights into damage assessment after earthquakes, with Kahramanmaraş Province serving as the focal point for model training. The findings contribute to the development of efficient and accurate disaster management strategies and lay the foundation for further research in this field.
-
ÖgeExplainable deep learning classification of tree species with very high resolution VHRTreeSpecies dataset(Graduate School, 2025-01-24) Topgül, Şule Nur ; Sertel, Elif ; 705211017 ; Satellite Communication and Remote SensingForests are among the most vital natural resources, playing a significant role in regulating the climate, maintaining ecological balance, and supporting biodiversity conservation and sustainable forest management. Additionally, they contribute to various applications, such as hazard management and wildlife habitat mapping. Understanding the spatial and temporal distribution of forests and forest stand types is a prerequisite for gaining deeper insights into their role within the Earth's systems. In this context, remote sensing data is widely utilized for forest stand type classification. However, traditional classification methods are often time-consuming and typically limited to specific areas and species, which significantly restricts their applicability to different regions and diverse tree species. With the increasing availability of high-resolution satellite imagery, deep learning methods have emerged as a powerful tool for forest management and tree species classification, offering enhanced efficiency and broader applicability compared to conventional approaches. Remote sensing (RS) applications, which serve as an essential spatial data source in forestry practices, have emerged as an effective solution for field studies due to their cost-efficiency and rapid data acquisition capabilities. Remote sensing systems provide valuable spatial, temporal, and spectral resolution data to cover forest areas at the required scale and within the necessary temporal intervals for data collection. High-resolution remote sensing data are frequently preferred for deriving detailed tree-level information, particularly for tasks such as individual tree detection or damage assessment necessary for maintaining tree health. Satellite systems such as Sentinel-2 and Landsat are frequently preferred due to their open-access nature, which allows for the collection of data across broad spectral bands and the provision of continuous data access. Nevertheless, the spatial resolution limitations of these satellites may render them inadequate for particular applications. Tree species with varying structural and morphological characteristics exhibit distinct spectral properties. Trees within the same environment but at varying developmental stages or health conditions can display significant differences in their spectral characteristics. In this regard, the application of remote sensing data is essential for achieving precise and reliable classification of tree species. Over the past decade, considerable progress has been made in the identification of tree species, encompassing a spectrum of approaches from fundamental image processing techniques to sophisticated machine learning (ML) and deep learning (DL) methodologies. Nevertheless, traditional classification algorithms, such as Random Forest (RF) and Support Vector Machines (SVM), have shown limited effectiveness in identifying tree canopies within dense and complex backgrounds. However, the time-consuming nature of traditional methods and their typical application to only specific areas and tree species substantially constrain the usability of these models across different regions and diverse species. Conversely, with the increasing availability of high-resolution satellite imagery, deep learning methods have emerged as powerful tools in forest management and tree species classification. DL-based models possess the potential to accurately extract more intricate information structures. Nevertheless, the effective application of these models generally requires a larger number of reference data samples to enable sufficient learning of the model parameters. As part of this thesis, a new benchmark dataset for forest stand type classification, called VHRTreeSpecies, is introduced. This comprehensive dataset includes very high-resolution RGB satellite imagery of 15 dominant tree species from various forest ecosystems across Turkey. The input images and their corresponding labels were generated using Google Earth imagery and forest stand maps provided by the General Directorate of Forestry (GDF). The dataset was curated by selecting pure species and masking raster images using vector data. High-quality images captured during the summer months (late July to mid-August) from the past five years were prioritized. The dataset was further diversified to represent different forest stand development stages (youth, sapling, thin, medium, and mature trees) and canopy closure levels (open, moderately closed, fully closed). The dataset was analyzed using various CNN architectures, including ResNet-50, ResNet-101, VGG16, VGG19, ResNeXt-50, EfficientNet, and ConvNeXt. Additionally, explainable artificial intelligence (XAI) methods, such as Occlusion, Integrated Gradients and Grad-CAM, were applied to examine the decision-making processes of the models. Evaluation metrics, including Max-Sensitivity and AUC-MoRF, were employed to comprehensively assess model performance not only in terms of classification accuracy but also in terms of the interpretability and reliability of their decision-making mechanisms.
-
ÖgeSingle-frame and multi-frame super-resolution on remote sensing images via deep learning approaches(Graduate School, 2022-07-29) Wang, Peijuan ; Sertel, Elif ; 705172003 ; Satellite Communication and Remote SensingAs a quite significant computer vision task, image super-resolution (SR) has been widely applied in remote sensing (RS), medical imaging, video surveillance, and biometrics. Image SR aims to restore high-resolution (HR) images by enhancing the spatial, spectral, or temporal resolution of the low-resolution (LR) inputs. In recent years, great efforts have been made for improving the SR approaches. One of the approaches is to design deeper networks. Nevertheless, this greatly increases computation and memory consumption. As a result, some mechanisms (such as cascading networks, attention mechanisms, and back projection) are proposed to improve the performance and the training process of the complex networks. Satellite imagery can be seen in various fields, namely Land cover/Land use classification, road and building extraction, observation of climate, and earthquake prediction. However, in some cases, the resolution of satellite images can not meet the application requirements due to the technology and cost limitations during the satellite design; therefore, the improvement of image resolution might be necessary. Since it is not possible to upgrade the equipment onboard the launched satellite, the software-based SR algorithms are deserved to be explored in RS fields. This thesis aims to strengthen the spatial resolution of optical satellite imageries by using deep learning (DL) methods. Generally, image SR algorithms can be categorized as single-frame image SR (SFSR) and multi-frame image SR (MFSR). The inputs of the SFSR can be a single LR image. While multi-frame image SR aims to restore HR image by using multiple LR images, which can be obtained under different conditions and at different angles. Recently, great contributions have been made to improve the SR methods including two aspects: (1) increasing the value of PSNR (Peak-Signal-Noise-Ratio); (2) improving the image quality perceptually. Nevertheless, some algorithms obtain a high PSNR but with a low perceptual quality which is more important to human perception. Therefore, this thesis has the following objectives: (1) Explore a perceptual-driven approach to enhance the SR image quality visually on single-frame and multi-frame RS imageries; (2) Explore Generative Adversarial Network (GAN)-based models for single-frame and multi-frame RS imagery super-resolution task to fulfill the multi-scale problem and blind to the degradation model; (3) Explore an image fusion method that can generate an arbitrary size of the super-resolved image rather than a small patch. This thesis firstly gives an overview of single-frame and multi-frame RS image SR methods. The single-frame RS image SR methods are briefly classified into supervised and unsupervised methods. The former mainly includes Convolutional Neural Networks (CNN)-based, GAN-based, attention-based, and Back-projection based methods. In addition, the commonly used attention mechanisms including self-attention, channel attention, spatial attention, mixed high-order attention (MHOA), non-local attention (NLA), and non-local sparse attention (NLSA) are also introduced. Moreover, loss functions including pixel-wise loss, perceptual loss, adversarial loss, and cycle consistency loss are presented. For the single-frame RSISR, firstly, an attention CNN-based SR method is proposed. Although CNN-based algorithms have made outstanding achievements in computer vision tasks, the traditional CNNs methods treat the abundant low-frequency information included in the LR inputs equally across channels. Attention-guided algorithms play a vital role in the informative features extraction in various tasks including image SR. With the application of the attention mechanism, the proposed CNN-based method can further learn the deeper relationships among the different channels. Instead of simply integrating the attention module with the residual blocks, a Layer Attention Module (LAM) and Spatial Attention Module (SAM) are proposed to further learn the relationships among the Residual Groups (RG). Moreover, the perceptual loss function is adopted in the training process to enhance the generated image quality perceptually, and Random down-sampling is applied to strengthen the model's generalization ability. Secondly, an attention GAN-based super-resolution method is explored for the single-frame RS images. CNN-based methods have made great contributions to increasing the value of PSNR/SSIM. Nevertheless, the generated outputs tend to be overly smooth and blurry. GAN can generate more realistic images than normal CNN-based methods and has been introduced to single image super-resolution (SRGAN, ESRGAN, EEGAN). Standard GANs only function on spatially local points in LR feature maps. The attention mechanism can directly learn the long-range dependencies in the feature maps both in the generator and discriminator in a powerful way. By applying the attention mechanism, the network allocates attention based on the similarity of color and texture. Therefore, based on ESRGAN, an attention GAN-based method is for the single-frame RS image SR. The ESRGAN was mainly improved from two aspects: (1) we further improved the architecture of the residual blocks by adding more skip connections; (2) we add attention modules to the residual blocks for further feature extraction. Moreover, instead of working on aerial photographs or low-resolution and medium-resolution satellite images, we are focusing on the Very High-Resolution (VHR) satellite imageries, such as the Pleiades, and Worldview-3. The spatial resolutions of the multispectral images for the Pleiades, and Worldview-3 are 2m, and 1.24m, respectively. Furthermore, for the attention CNN-based method, we evaluated the method on the Pleiades and Worldview-3 datasets with scaling factors of 2, 4, and 8. For the attention GAN-based method, we evaluated the method on the Pleiades and Worldview-3 datasets with a scaling factor of 4. The experimental results show the attention-based method can provide better perceptual results both quantitatively and qualitatively. At last, we proposed an attention GAN-based method for the multi-frame RS image SR. Firstly, we introduced an attention mechanism to the Generator and proposed a space-based network that worked on every single frame for better temporal information extraction. Secondly, we proposed a novel attention module for better spatial and spectral information extraction. Thirdly, we applied an attention-based discriminator to enhance the discriminator's discriminative ability. Finally, the experimental results on the SpaceNet7 dataset and Jilin-1 dataset exhibit the superior of the proposed model both quantitatively and qualitatively.
-
ÖgeVessel detection from very high-resolution satellite images with deep learning methods(Graduate School, 2022-06-22) Büyükkanber, Furkan ; Yanalak, Mustafa ; 705181008 ; Satellite Communication and Remote SensingVessel detection from remote sensing images is becoming increasingly important component in marine surveillance applications such as maritime traffic control, anti-illegal fishing applications, oil discharge control, marine pollution and marine safety. Increasingly, very high and medium resolution (VHR and MR) earth observation satellites both significantly increase the detectability of many terrestrial objects and shorten recurring visit times in orbit like never before, making the use of this technology attractive for a variety of maritime monitoring missions. However, the difficulty and complexity of object detection in huge satellite images that cover hundreds of square kilometers and derive results under near real-time constraints cause traditional methods to face many difficulties when processing satellite images of this size. Processing these images and applying them to deep learning methods makes it possible to minimize unforeseen errors that can be made by analysts, and to save labor, time and cost. In order to create the artificial neural network and make it successful by determining the deep learning method, it is necessary to train using as much as possible examples of the objects targeted to be detected. By using the designed convolutional neural networks, it is possible to detect more than one object in a given test input image and perform change analysis as well. The weights are updated in each layer for the input image processed in the multilayer convolutional neural network, and the error rate is found by looking at the difference between the detected value and the actual ground truth value. Many vessels for commercial, military and civil purposes are observed in international maritime areas, usually in areas close to ports and coasts. High resolution satellite images, which provide wide field of view and altitude monitoring, are very useful for vessel detection. Vessel detection from satellite images plays a significant role for inspecting maritime areas, controlling maritime transport traffic and applications for defense purposes. Open source datasets are widely used in object detection applications, since it takes a substantial amount of time and cost to build a dataset for object recognition and detection from satellite images. Within the scope of this thesis, models developed using convolutional neural networks including single-stage and two-stage deep learning methods were used by applying our own dataset images that we build with the open source DOTA dataset selected for vessel detection. For the purposes of the experiments in this research, three separate datasets were built. All the images were labelled with YOLO annotation format, then in accordance of use for various models, they have been converted to COCO and Pascal VOC annotation format. Both inshore and offshore vessel images have been collected with having wide variety of scales, shapes, orientations and weather conditions (fuzzy, cloudy, sunny, etc.). Experiments were performed by using Faster R-CNN, YOLOv3, YOLOv5 and YOLOX deep learning models on all three different datasets. Any dataset containing various examples of the target object considerably improves the accuracy of outcomes in deep learning applications by implementing various data augmentation techniques, such as mosaic, mixup, and rotating images, are utilized for remote sensing. In some experiments, more than one augmentation approach is being used simultaneously to improve the accuracy of the results. Not all data augmentation approaches had the same effect on the experiment outcomes. As a result, there is no logical answer to the question of which data augmentation strategy is the most effective. The outcomes of the studies were compared using the mean average precision metric (mAP), and the YOLOv5 model achieved on top results. All of the experiments have yielded the same result: raising the depth of the network by increasing the size of the input images. mAP value results improved as the input sizes were increased, however this caused the selected models longer to train. Experiments in deep learning studies are made easier by machines that have powerful graphics cards. Faster R-CNN, YOLOv3, YOLOv5 and YOLOX model trainings were conducted on a local machine workstation equipped with NVIDIA GeForce RTX 2080Ti graphics card and Intel® Core™ i9-9900K 3.60 GHz CPU processor. Deep learning applications were carried out using Python programming language and PyTorch framework deep learning library.