针对在轨条件下因观测条件匮乏导致巡视器导航相机标定精度受限的问题,提出了一种基于直线特征的导航相机快速在轨自标定方法.首先,利用SAM(segment anything model)对导航影像太阳能板区域进行图像分割,并对太阳能板OCR(optical control room)直线特征进行提取;然后,构建附加距离约束的导航相机自检校模型;最后,利用Levenberg-Marquardt方法对标定模型进行求解.开展了模拟实验及在轨实验.分析结果表明,相比光束法自标定、CAHVOR及Vanishing Points等标定方法,所提出方法标定参数重建的坐标精度平均值分别提升了20.60％、30.82％和35.92％,对太阳能板OCR距离恢复距离偏差范围在-1.4～1.5 mm之间.该方法可为巡视器导航相机在轨标定技术提供一定的价值参考.
A Fast Calibration Method for Navigation Cameras with Additional Distance Constraints
Aiming at the problem of limited calibration accuracy of the rover navigation camera due to the lack of observation conditions under on-orbit conditions,a fast on-orbit self-calibration method of the navigation camera based on straight line features is proposed.Firstly,image segmentation of the solar panel region of the navigation image is carried out by using SAM(segment anything model),and the linear features of the solar panel OCR(optical control room)are extracted by using the linear extraction method combined with luminance weighting.Then,the self-calibration model of the navigation camera with additional distance constraints is constructed based on the linear features.Finally,the calibration model is calibrated by using the LM method.Simulation experiments and on-orbit experiments are carried out,and the results show that compared with the calibration methods such as bundle adjustment,CAHVOR and Vanishing Points,the average values of coordinate accuracy reconstructed based on the calibration parameters of the proposed method are improved by 20.60％,30.82％and 35.92％,respectively.Meanwhile,the OCR distance of the solar panel is recovered by combining the calibration results of this paper and the triangulation method,and the recovered distance deviation ranges from-1.4～1.5 mm,indicating that the proposed method can provide a certain value of reference for the on-orbit calibration technology of the rover navigation camera.
NETL
NSTL
万方数据
