EGU General Assembly 2020
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Automatic extraction of modern reefs satellite images geometries using Computer Vision

Grisel Jimenez Soto1, Mirza Arshad Beg1, Michael C. Poppelreiter2, and Khaidi Rahmatsyah3
Grisel Jimenez Soto et al.
  • 1Universiti Teknologi PETRONAS, Institute of Hydrocarbon Recovery, Geosciences, Malaysia (
  • 2Shell Kuwait
  • 3Baker Hughes Netherlands

Three selective, global – scale areas for modern’s reef sites have been selected on exploring the automatic metrics extraction for further studies among the relationships between reef morphology and the surrounding oceanographic conditions that can potentially be used as an input parameter for training images for Multiple Point Statistics simulations.

Obtain geometric features from satellite images is a very laborious task and hard work when making manually. Nevertheless, automatic geometric features detection is a challenging problem due to the varying lighting, orientation, and background of the target object, especially when analyzing raw images in RGB format. In this work, a robust algorithm programmed in python is presented with the purpose of estimate automatically the geometric properties of a set of coral reef islands located in South East Asia. First, a python code load massive satellite imagery from a specific folder RGB format, then each raw coral reef island image is resized, converted from RGB band to gray, smoothed and binarized using Open Computer Vision Library available in python 3. The island edge information contains very prominent geometric attributes that characterize their behavior, thus morphological transformations were applied to define the contour of the island. Moreover, a structural analysis and shape descriptors were made in a set of images in order to numerically calculate the characteristics of the island.  Finally, a total of 27 satellite images were processed by the algorithm successfully, only two images were not segmented correctly because the illumination and intensity of the predominant colors, especially blue, were different from the rest of the images. This dataset was exported from python to Microsoft Excel spreadsheets and CSV format.

How to cite: Jimenez Soto, G., Arshad Beg, M., Poppelreiter, M. C., and Rahmatsyah, K.: Automatic extraction of modern reefs satellite images geometries using Computer Vision, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13893,, 2020.