Automatic extraction of offshore platforms using time-series Landsat-8 Operational Land Imager data

Offshore oil and gas platforms are rapidly expanding worldwide to meet the growing demand for energy, and up-to-date knowledge of their number and spatial distribution is essential for management and safety maintenance of the marine environment. However, information on the distribution of these platforms over vast areas of the sea is extremely poor - partly because of business secrecy, national interests including security, and partly from bureaucratic inertia. Until recently, the automatic and accurate identification of platforms using optical imagery was challenging due to their subtle characteristics (i.e., low contrast and small size) and prevalent cloud cover. The Landsat-8 satellite, with its increasing Operational Land Imager (OLI) dataset with uniform quality and global coverage, may assist researchers in assessing the worldwide distribution of offshore platforms. Here we describe an automated method, based on time-series of OLI imagery, for extracting offshore platforms (AMEOP) which addresses the aforementioned problems. Specifically, two strategies (time-series and multi-refinement) were designed to address the challenges presented by heavy cloud cover over the sea. Three features documented in the OLI imagery time-series (contextual feature, and position- and size-invariance) were used to accurately discriminate offshore platforms from false positives. The AMEOP was applied to the Gulf of Thailand and 370 offshore platforms were automatically identified using 198 OLI images. Validation using Synthetic Aperture Radar (SAR) images and Chinese high-resolution satellite imagery demonstrated an omission error of 3.8% and a commission error below 1.0%. The AMEOP was also successfully extended to the Persian Gulf and northern Gulf of Mexico, 2346 and 2411 offshore platforms were respectively identified. We believe that the methodology presented herein provides a robust and cost-effective technique for improving the identification of offshore platforms over large areas of the sea; and that the accumulation of OLI data, together with the SAR images, has great potential for mapping and tracking of offshore platforms on a global scale, as well as for providing an early warning of oil spills around platforms and the improved estimation of their greenhouse gas emissions.