Calibration And Pose Estimation In Ship Vision Measurement Guidance

Camera calibration and pose estimation are key issues in visual measurement technology and are commonly used in autonomous navigation,autonomous driving,and augmented reality.In view of the problem of aircraft transportation,this paper mainly discussed the methods of calibration,pose estimation and distortion correction.As for the ordinary camera fixedly installed on the ship’s surface,we deduced calibration algorithms based on the point-line combination features,and designed simulation experiments.Experimental results show that the algorithms can make full use of the information in the scene to accurately calibrate the camera parameters.For large-field view cameras,it is difficult to uniformly arrange control points during calibration,so we adopted the method that calibrate the internal and external parameters separately.The results of external field experiments proved that the calibration accuracy of separate calibration is higher than that of simultaneous calibration.As for the pose estimation of aircraft during the transportation process,this paper proposed the pose estimation algorithm based on line features and the combination of point-line features.This algorithm combines the technical advantages of inertial measurement unit and visual measurement,and can quickly calculate the camera pose using a small number of matching points or lines.In the simulation experiments,the result on the noise and inertial measurement error(roll angle,pitch angle)is quantitatively analyzed.In the real experiment,we analyzed the performance of line matching algorithm and plotted the trajectory of the camera motion.The proposed algorithm proved to be effective in solving camera poses.In scenes with low space heights,the range of view field obtained with ordinary cameras is narrow,which is not conducive to the pose estimation.Therefore,in the hangar,we adopted the fish-eye camera,which can significantly expand view field,but its images have severe radial distortion.In the paper,we combined image undistorting algorithm with PNL pose estimation algorithm,and studied the methods based on distortion lines.The results of simulation experiments and real experiments show that the proposed algorithm can effectively eliminate image distortion and accurately solve the camera pose.