محسن نظري شكيب

Oil spills in seas an‎d oceans are considered one of the major environmental crises, an‎d rapid identification of them is crucial for preventing the spread of pollution an‎d protecting marine ecosystems. In this research, to improve the accuracy an‎d speed of oil spill detection, Sentinel-1 SAR satellite images an‎d machine learning an‎d deep learning algorithms were utilized. Initially, the received images were pre-processed an‎d effective features were extracted using the Principal Component Analysis (PCA) algorithm. Various models, including SVM + PCA combination, CNN, VGG16, an‎d ResNet, were trained, eva‎luated, an‎d tested. This thesis used a dataset comprising 1800 labeled images (oil spill an‎d non-oil spill). The results showed that the CNN model had the best performance among all examined models, with a test accuracy of 92.9% an‎d a validation accuracy of 93.4%. This model also had the lowest loss during training (0.03) an‎d a suitable value in validation (0.30).The VGG16 model ranked second with a test accuracy of 82.4% an‎d a validation loss of 0.42, indicating that increased model complexity does not necessarily lead to improved performance. The ResNet model, despite its advanced structure, only achieved a test accuracy of 66.3%, an‎d its training loss (11.3) was very high, indicating a poor fit between the network structure an‎d the projectʹs data. Additionally, the SVM model combined with PCA, as a more classical algorithm, yielded a low test accuracy of 64.7%. Overall, it was shown that the use of deep learning algorithms, especially the simple yet effective CNN architecture, can be an effective an‎d practical tool for identifying oil slicks in SAR images. High accuracy, suitable processing speed, an‎d simplicity of implementation are among the main advantages of this model for environmental applications an‎d large-scale marine pollution monitoring systems.

تصاوير ماهواره اي SAR , يادگيري ماشين، , يادگيري عميق

Satellite images SAR , Machine learning, , Deep Learning

Mohsen Nazari Shakib

Dr. Nasiri