A novel solution in the simultaneous deep optimization of RGB-D camera calibration parameters using metaheuristic algorithms

Abstract

This article presents a novel method for estimating 19 parameters of RGB and depth camera calibrationsimultaneously. The proposed algorithm is based on applying metaheuristic methods for deep optimization and estimatingall parameters of intrinsic, extrinsic, and lens distortions of cameras. This paper compares four metaheuristic algorithms,i.e. a genetic algorithm, particle swarm optimization, the colonial competitive algorithm, and the shuffled frog leapingalgorithm, with a numerical algorithm called singular value decomposition. The proposed method does not need theinitial estimation for optimization and it can avoid being trapped in local minima. By using nondirect estimation, weachieve middle computing matrices such as the homography matrix, which is used in the pinhole camera model. Bothversions of Kinect sensors are used for the experimental evaluation. The mean square of the reprojection error criteria isdefined as the objective function in the proposed algorithm. The experimental results show that the proposed methodis more efficient and accurate than traditional numerical solutions.