1344618323

Results 4 issues of 1344618323

https://github.com/ov2slam/ov2slam/blob/72228fb7b0f117bcb99ce62d68a046fd5c24052e/src/feature_tracker.cpp#L86 `OPTFLOW_LK_GET_MIN_EIGENVALS` has already been set in the above code, so `verr` represents the minimum eigenvalue. This eigenvalue should not be too small, and outliers should not be detected by...

1. 卡尔曼增益后协方差的投影问题: https://github.com/gaoxiang12/slam_in_autonomous_driving/blob/fe9c79bf03cffb4b9e6388015a42c747c318a182/src/ch8/lio-iekf/iekf.hpp#L254 在update P那一步,`cov_ = (Mat18T::Identity() - Qk * HTVH) * Pk` 用得是上一步迭代(最后一个线性化点)后的协方差矩阵`Pk`,那么project P中计算 dtheta时应该用最后一次 `dx_` 里的 $\theta$ 而不该是 `(R_.inverse() * start_R).log()`? 2. 另外,我觉得公式8.5写错了,代码也有问题:书里写的是 $$J_{\theta}=I-\frac{1}{2} \delta \theta_k^{\wedge}, \delta \theta_k=Log(R_k^TR_0)$$...

https://github.com/gaoxiang12/slam_in_autonomous_driving/blob/fe9c79bf03cffb4b9e6388015a42c747c318a182/src/ch3/eskf.hpp#L245

fastlio论文 https://arxiv.org/pdf/2010.08196 中公式18 和 本书中的公式(8.7b)有些区别,多了一项包含 $(1-KH)$ 。想问下是哪里的区别导致了这个差异。 ![image](https://github.com/user-attachments/assets/674872e9-43bd-41ce-9971-464444fe4ca9) fastlio中相关代码为 ![image](https://github.com/user-attachments/assets/901420d3-43c3-4de9-803e-210b20f30a8e) 本书中相关代码 ![image](https://github.com/user-attachments/assets/4eebd072-fd64-4259-b7db-722976beb221) 另外,我注意到https://github.com/gaoxiang12/slam_in_autonomous_driving/issues/97 也提到了相关的问题,论文中套错公式了?