Advanced Techniques and Application Areas in Remote Sensing Images: Integration of Deep Learning and YOLOv5 Algorithms

Remote sensing is a multidisciplinary research field that focuses on evaluating remotely acquired data through advanced image analysis techniques. This study aims to examine various advanced techniques and various application areas of these techniques within the framework of research focusing on remote sensing images. Advances in image analysis and processing techniques stand out as important issues that allow remote sensing images to be used more effectively and meaningfully. In this context, the advanced techniques to be examined in the article and various application areas will be discussed to reveal the broad perspective in the remote sensing discipline. The study provides a broad perspective within a research framework focusing on remote sensing images, highlighting the importance of advances and usage potential in this field. Advanced techniques such as deep learning indicate significant potential to obtain sharper and more accurate results in remote sensing. Advances in this field may enable the development of innovative remote sensing systems with potential for application in various sectors. Within the scope of the study, our original dataset was created, and our dataset was enriched as missing data classes were identified. The Roboflow platform was used to view the health of the dataset used. Additionally, data augmentation operations were also carried out on this platform. In addition, synthetic data was produced with Generative Adversarial Networks to diversify the object classes that were found insufficient while creating the original data set. By using super-resolution methods during the training phase, the image quality was increased, and higher performance was achieved. While target objects were detected using the most up-to-date object recognition algorithms (YOLOv5, YOLOR, YOLOX), the SAHI algorithm was tested as a hybrid model to detect small objects. In experimental results, it has been observed that YOLOv5 gives better results.