Joint Calibration Method of Thermal Infrared-Visible Based on Cross Modal Feature Matching

Abstract. Aiming at the problem of limited imaging quality of monomodal optical cameras in low-light environments, this paper constructs a thermal infrared-RGB binocular stereo vision system and proposes a joint calibration framework for infrared and RGB cameras to provide a high-precision geometric alignment basis for multimodal image fusion. First, a high-precision geometric calibration method is used to eliminate the internal distortion of the infrared camera and establish the mapping relationship between its pixel coordinate system and physical space. Second, a cross-modal extrinsic calibration strategy based on common view targets is designed. A specially designed heated and temperature-controlled chessboard calibration board for thermal infrared is used to enhance the feature contrast in the infrared image through temperature control. Combined with a cross-modal feature matching algorithm, the spatial pose transformation matrix between the infrared and RGB cameras is accurately solved to align multimodal images. Experimental results show that the proposed thermal infrared–RGB binocular calibration method can significantly improve calibration accuracy and robustness, providing effective technical support for visual perception and target recognition in low-light environments.