Research on monocular vision-based reconstruction technology for aircraft fuel liquid surfaces

Fuel measurement is a critical technology for aircraft, currently relying heavily on sensors for data collection. This paper proposes a method for assisting in measuring the fuel surface height in aircraft fuel tanks using monocular vision. The method involves arranging specifically shaped feature points at the bottom of the fuel tank and using a monocular parallel light model for liquid surface reconstruction. Traditional monocular parallel light models introduce cumulative errors due to iterative calculations. This paper analyzes 2D liquid profiles by assessing the degree of fluctuations at different positions. Based on the severity of fluctuations, different points are selected for polynomial fitting to correct the height data, extending the algorithm to 3D liquid surfaces. The factors influencing each stage of the algorithm were analyzed, including the choice of fitting method, number of fits, and data sources for fitting. The reconstruction results indicate that this method offers significant advantages in terms of accuracy and generalizability.