多镜头组合型全景相机两种成像模型的定位精度比较

Comparison of Two Sensor Models for Multi-camera Rig System in Measurements

根据多镜头组合型全景相机的严格成像机理,分析了广泛采用的全景理想成像模型系统误差的来源,推导出误差分布规律,并深入比较了严格成像模型和理想成像模型之间的差异。首先,分析了理想模型的单片物方定位误差和像方后方交会误差,理论公式表明其误差分布和物像间距及视角大小相关;其次,分析了全景立体像对的理想核线误差,并证明误差受像对间的旋转角和Z坐标影响。最后,分别采用室内、外数据对两种模型的空间后方交会、核线模型、光束法平差进行比较分析。试验结果表明,室外条件下由于物距接近于全景投影面、立体影像间旋转角较小等因素,两种模型的差异很小,且都达到了相当于1像素左右的测量精度;在狭窄的室内条件下,由于物距小、旋转角大等因素,严格模型在测量精度上明显优于理想模型。

According to a rigorous sensor model for multi-camera rig system,the error sources of the most widely used ideal panoramic sensor model are indicated,error distribution laws are deduced,and comprehensive comparison between the two models are given.First,the resection and 3Dlocalization errors of the ideal model are analyzed respectively in a single panoramic image that shows the correlations both with the object-image distance and the viewing angle.Second,the epipolar errors of the stereo-pairs are analyzed,which are both affected by the rotation angles and z-coordinates of the image points.Finally,the comparative experiments are carried out in space resection,epipolar constraints and bundle adjustment with the two sensor models.The outdoor test shows the difference between the two models is slight,and both models achieve 1pixel accuracy.In contrast,the indoor test shows that the rigorous model is stricter and produces obviously better measurement accuracy than the ideal model.All the test results are consistent with the deduced analytical error laws of the ideal panoramic sensor model.