Browsing by Author "Curinha, Hugo Miguel Carvalho"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- SAR-to-Optical domain transformation for easier image recognitionPublication . Curinha, Hugo Miguel Carvalho; Marques, Paulo Alexandre CarapinhaAbstract Optical imaging systems work by capturing the light’s photons, that upon interaction with objects, scatter randomly. By sensing and processing the received particles it’s possible to create meaningful representations in the form of images. These systems offer rich spectral information, as they capture the frequencies in the visible band, which makes it desirable when being visually analyzed by humans as these reflect how humans see. Yet one of the biggest limitations is that these are heavily affected by weather and lighting conditions. Some cloud coverage or the lack of light can be enough to block the image acquisition, making these systems unusable unless coupled with other types of sensors such as Infrared. Sythenthic Aperture Radar(SAR) are remote sensing systems, that offer significant advantages in land monitoring, scientific research, security applications, and others. It is an active remote sensing technique that uses microwave signals to obtain high-resolution images of the Earth’s surface. Because of the frequencies that are usually used these are less susceptible to weather conditions, meaning it’s possible to obtain images whether day or night and no matter the cloud coverage. However, inherent limitations, such as speckle noise or geometric distortions, and the fact that it loses some crucial information like color information, make human interpretation of these a complex task. This thesis proposes a Convolutional Neural Network (CNN)-based autoencoder that extracts the relevant information from the SAR image and reconstructs the optical counterpart, therefore enabling image interpretation by non-specialized personnel.