Research on well blowout volume measurement method based on machine vision and multi-source data feature fusion

Once a blowout accident occurs during oil and gas extraction, obtaining an accurate, real-time measure of the well blowout volume is critical for selecting appropriate wellhead reset parameters, dispatching emergency equipment, and formulating an effective rescue plan. However, the harsh field environment, high temperature, high pressure and high noise makes the traditional flow measurement equipment can not be achieved on-site installation and measurement, forcing operators to rely solely on experience-based estimates. To address the inability of existing techniques to quantitatively identify well blowout volume and the large uncertainties of manual estimation, this paper employs high-speed imaging of wellhead fluid and flame images to propose an automatic, rapid measurement method for well blowout volume based on machine vision and multi-source data feature fusion. First, the characteristic features of the blowout fluid are analyzed and a SURF-FLANN-USAC image-processing model is established. By applying the pinhole imaging principle to paired feature-point coordinates, the displacement of each feature point is computed and, together with the image acquisition frame rate, used to derive the fluid flow velocity. Simultaneously, flame regions are extracted and flame heights identified from the flame images under constraints in the RGB-HIS color space. Next, by fusing the fluid-velocity data and flame-height data via a CFAM-BLOCK model, a multi-source data fusion framework for measuring well blowout volume is developed. Finally, an experimental platform for measuring well blowout volume is constructed; simulated test results demonstrate that the error between measured and actual well blowout volumes remains within 9 %. This method also demonstrated good measurement results in field tests and field experiments with added noise.