知识卡片
奇异值投影落实几何约束
若线性估计矩阵F=U diag(s1,s2,s3)V^T,可将s3置零恢复秩2。对E还需将前两个奇异值替换为同值,如平均值。这只是投影到结构约束的步骤,不能纠正错误匹配或代替几何误差优化,退化数据仍须识别。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 立体视觉
原书出处 · 1 处
- · 查看原书依据
As both matrices are rank 2, the minimization the linear set of equations can be done using a constrained minimization or directly optimizing a rank 2 matrix. You can also delete the third singular value F, and, for E, you can delete the third singular value and set the first and second singular values to be equal.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第40章《立体视觉》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c040.xhtml)
- 结论依据:摘录中关于奇异值投影落实几何约束的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:As both matrices are rank 2, the minimization the linear set of equations can be done using a constrained minimization or directly optimizing a rank 2 matrix. You can also delete the third singular value F, and, for E, you can delete the third singular value and set the first and second singular values to be equal.