Análisis de la sensibilidad en un modelo de calibración cámara-LiDAR

Abstract

Recently the data fusion between a camera and a depth sensor of LiDAR type, has become an issue of major concern in industry and engineering. The quality of the delivered 3D models depends greatly on a proper calibration between sensors. This paper presents a sensitivity analysis in a camera-lidar calibration model. The variability of each parameter was calculated individually by the Sobol method, based on ANOVA technique, and the FAST method, which is based on Fourier analysis. Multiple sets of parameters were simulated using Monte Carlo and Latin Hypercube methods for the purpose of comparing the results of the sensitivity analysis. We defined which parameters are the most sensitive and prone to introduce error into our reconstruction platform. Statistics for the total and global sensibility analysis for each sensor and for each parameter are presented. Furthermore, results on the sensitivity ratio on camera-LiDAR calibration, computational cost, time simulation, discrepancy and homogeneity in the simulated data are presented.