Online Calibration of Binocular Vision Sensor Structural Parameters Based on Kalman Filtering

The structural parameters of a stereo vision sensor directly affect its accuracy. However, they are susceptible to complex environmental interference, and traditional online calibration methods fail to recover scale, thus hindering precise measurement once the parameters are changed. To address this issue, a method for online calibration of structural parameters of binocular vision sensors based on Kalman filtering is proposed in this paper. Based on the continuity inherent in visual measurements, a Kalman Filter is employed to estimate current structural parameters by updating with previous parameters and incorporating temporal context, thereby deriving the object’s scale. The estimated parameters are subsequently refined using the proposed nonlinear optimization based on multiple geometric constraints, yielding the maximum likelihood solution and enabling high-precision online calibration. The effectiveness of this method has been validated through experiments.