Versatile unsupervised multi-resolution classification method based hydrocarbon extraction from satellite images [articol]

Abstract

Satellite image processing is a developing innovation for the oil and gas industry. The look for seafloor hydrocarbon detection is as often as possible an essential block of deep-water investigation programs. On the off chance that thermo genic hydrocarbons are found at the seafloor, investigation hazard can be moderated by the leaks giving proof of a working oil framework. The recuperated leaked hydrocarbons would then be able to be utilized to observe the substance of the subsurface supply and provide pieces of information about their inception. This work intended to enhance the spatial and spectral data of satellite images by detection hydrocarbon in Ramanathapuram district (Tamilnadu state) utilizing Versatile Unsupervised Multi-Resolution modeling classification approach. The proposed Versatile Unsupervised Multi-Resolution based method that automatically classifies the hydrocarbon regions from spatiotemporal remote sensing images. First, a kernel has designed according to the structure of multi-spectral and multi-temporal remote sensing data. Secondly, the Versatile Unsupervised Multi-Resolution framework with fine-tuned parameters has intended for training region samples and learning spatiotemporal discriminative representations. The performance of the proposed hydrocarbon detection method is validated through simulation using the Matlab software. Compared to conventional hydrocarbon extraction methods with a proposed Versatile Unsupervised Multi-Resolution method achieving the best results for accuracy is 97.72%, sensitivity is 98.25 %, and specificity is 94.02%.