Flaresat: método de segmentação da atividade de queima de gás em chaminés de instalações petrolíferas utilizando deep learning

Abstract

The process of controlled burning of gas due to pressure in pipelines, known as gas flaring, is a daily routine, mainly used in the oil and gas industries for decades. The impact of this process is a global concern due to environmental, health and economic issues. To meet the global goal of controlling greenhouse gas emissions, the detection of gas flaring activity in oil industry facilities is essential information for measuring the volume of gases being flared. This work presents a method for the segmentation of gas flaring activity in oil and gas facilities using remote sensing images and deep learning techniques. For the development and validation of the method, a new dataset specifically designed for gas flaring segmentation was created using Landsat 8 satellite images. The dataset consists of processed and labeled images, representing a significant contribution to future research in the field. The use of deep learning achieved better results compared to currently used methods, which are based on image processing techniques, highlighting the contribution of the proposed method to the identification and segmentation of gas flaring activity.