Research on Calibration Methods of Long-Wave Infrared Camera and Visible Camera

Abstract

Long-wave infrared (LWIR) and visible (VIS) cameras can image information at different dimensions, but the way to calibrate these two types of cameras while registering and fusing the acquired images is difficult. We propose a calibration plate and a calibration method for thermal imaging and visible imaging to solve three problems: (1) the inability of the existing calibration plates to address LWIR and VIS cameras simultaneously; (2) severe heat interference in the calibration images of LWIR cameras; (3) difficulty in finding feature points for registration due to the different imaging spectra between thermal imaging and visible imaging. Simulation tests and error analysis show the error of outline central point computation is less than 0.1 pixel. Average errors of Euclidean distances from the margin outline scattered point sets of the closed circle and closed ellipse to the outline central points decrease by 10% and 9.9%, respectively. The Mean Reprojection Error in the calibration of LWIR and VIS cameras are 0.1 and 0.227 pixels, respectively. Through image registration design and fusion experiments, the FMIdct, MS-SSIM, Qabf, SCD, and SSIM of the images fused after distortion correction are all higher than those of the images fused before distortion correction, with the highest increases being 4.6%, 0.3%, 3.1%, 7.2%, and 1.4%. These results prove the effectiveness and feasibility of our method.