← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9051.c · 生成模型:GPT-6

增量重建会积累漂移

理解等级:中级

逐步加入图像时,局部姿态和三角测量误差可能累积,使尺度或位置发生漂移。缺失共同观测会让约束变弱,但漂移不是“看不到第一帧”必然产生的独立尺度重置。共享轨迹、闭环和联合调整能提供更多全局约束,仍需检查误差。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      In the previous description, we are making the assumption that every point is visible in every view (so one can run fundamental matrix estimation, triangulation, and other tasks of each of the M – 1 cameras with respect to the first view). However, if the visibility criteria isn't met, for each new image we are adding in to SFM, we will need to triangulate it

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第44章《多视图几何与运动恢复结构》(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/c044.xhtml) - 结论依据:摘录中关于增量重建会积累漂移的定义、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:In the previous description, we are making the assumption that every point is visible in every view (so one can run fundamental matrix estimation, triangulation, and other tasks of each of the M – 1 cameras with respect to the first view). However, if the visibility criteria isn't met, for each new image we are adding in to SFM, we will need to triangulate it