论文部分内容阅读
针对鱼眼相机的特点,首次提出鱼眼相机检校的正则化方法,介绍基于半视场角约束的鱼眼相机检校方法,阐述Tikhonov正则化方法和截断奇异值(TSVD)正则化方法的基本原理,探讨求解正则化参数的L曲线法和广义交叉检验法(GCV)的特点,最后构建5种不同的正则化方案对15组数据进行分别处理,并对得到的相机参数的精度和准确度进行比较分析。试验分析表明,各种正则化方案会引起相机内方位参数约为0.2像素的轻微偏移,但是相对于直接最小二乘解其精度至少提高了1000倍,能够极大地提高相机检校的抗干扰性。
In view of the characteristics of fish-eye camera, the regularization method of fish-eye camera calibration is first proposed. The fish-eye camera calibration method based on half-field angle constraint is introduced. The Tikhonov regularization method and truncated singular value (TSVD) regularization method Basic principle, discussing the characteristics of L curve method and generalized cross test (GCV) for solving regularization parameters. Finally, five different regularization schemes are constructed to process 15 sets of data respectively, and the accuracy and accuracy of the obtained camera parameters Degree for comparative analysis. Experimental analysis shows that various regularization schemes will cause a slight offset of about 0.2 pixel in azimuth parameters within the camera, but their accuracy is at least 1000 times higher than that of direct least square solutions, which can greatly improve the anti-jamming performance of camera calibration Sex.