Detecting shadows in satellite images

Scientists from the University of Malaga have devised a procedure for accurately identifying shadows in high-resolution images captured by satellites, making it possible to obtain more precise information on streets, buildings, vehicles, crops and other elements detected from space.

The results of this research have been passed to a company in the Parque Tecnológico de Andalucía (Andalusia Technology Park) which is already applying them in satellite image processing and detection of urban changes.

The Professor of the IT Engineering School at the University of Malaga and co-author of the study, Vicente Arévalo, explained to SINC that high-resolution images provided by current satellites and planes “have opened a new era in the field of teledetection, and that resolution enhancement also means that shadows, something inherent in any image, take on special significance”.

For example, shadows fall on buildings, cars or street furniture, the researcher commented, and in an aerial photograph “it is very important to detect what is or is not a shadow to correctly identify the elements that appear in it”.

Arévalo pointed out that the identification of shadows enables the subsequent application of specific information recovery techniques, as well as the preparation of three-dimensional designs. Thanks to the shadows, IT engineers can estimate, amongst other parameters, the height of elements in a landscape, such as a house.

To carry out this study, researchers have used images captured by the QuickBird satellite, also used for capturing aerial photographs of the Google Earth virtual atlas. Images obtained with this satellite have a 60 cm/pixel resolution, i.e. 60 cm of the real terrain captured in a pixel (the smallest unit comprising a digital image, defined by its brightness and colour). Once the image has been taken, its colour components are analysed and the so-called “seeds”, small groups of pixels that have a greater probability of being shadows, are identified through circles. To these “seeds” other surrounding pixels are added which are significant statistically for detecting the shadowy areas of the photo as accurately as possible in a process in which other tools such as edge detectors are also used.

The method developed by Andalusian researchers has been successfully tested in images obtained under different lighting conditions, in both urban and rural areas. For example, in a field of olive trees, it is easy to quantify the trees and see their size more accurately if shadows are properly identified in the high resolution images.

However, scientists’ main line of work is to detect urban changes, seeing how certain areas of cities change over time. Their studies make it possible to detect things ranging from earth movements or changes in the area to discovering whether anyone has built a swimming pool on their land. “We do not judge the legality of these types of actions”, Arévalo said, but the high resolution images of areas do help the work of surveyors.

[via]